CN1094223C - Automatic testing system for driver pile examination process based on image sequence analysis - Google Patents

Abstract

The invention belongs to the field of computer and digital image processing. The system includes: a gantry unit, a plurality of pile poles for examination; video camera; LED large-screen display and loudspeaker; multiple sets of infrared receiving/emitting sensor units; computer image processing unit, wireless transmission engine flameout measurement unit and Switch value acquisition unit. The invention recognizes the vehicle trajectory by real-time digital image analysis of the image sequence captured by the camera, which can completely replace the manual invigilation, improve the accuracy and objectivity of the examination results, and has higher accuracy compared with similar products. Flexibility, reliability and practicality.

Description

Automatic test system for driver pile test process based on image sequence analysis

technical field

The invention belongs to the field of computer technology and digital image processing technology. In particular, it involves the design of an automatic test system for the driver's pile test process.

Background technique

Personnel who learn to drive motor vehicles must undergo training in field pile examination subjects. In the traditional pile test training, the pile rod is made of wood, inserted on a disc base, and placed on the test site before the test. During the test, the examiner sits in the control room in front of the venue and observes the driving process of the vehicle. Through visual inspection and experience, it can be judged whether the vehicle has passed the line, rubbed against the pile pole, failed to move into the warehouse, stopped halfway, turned off the engine, and whether it is driving according to the prescribed route. When there are many vehicles, in order to speed up the test, there are generally three vehicles on one site to take the test at the same time. This has several disadvantages:

(1) The artificial random placement of the piles is inaccurate, resulting in errors in the size of the garage.

(2) Visually inspect the error of the body's outgoing line. Since there is no actual line marked in the garage area, the sideline of the garage can only be judged by the position of the pile poles, and the distance between the examiner and the garage is more than ten meters, and it is not in a straight line with the sideline of the garage, so the judgment of the body crossing the line is all Visual inspection is combined with estimation, and for the rear exit line in the process of posting and moving the warehouse, the estimated component is even more important. Such judgments are often inaccurate, especially when the exit line is not obvious, it often causes controversy.

(3) The judgment of engine flameout is completely visual. Since the examiner is indoors and the test site is relatively noisy, it is impossible for the examiner to hear the sound of the engine ignition. If the student re-ignites successfully within 2 or 3 seconds after the engine is turned off, the examiner may not be able to judge that the engine has been turned off.

(4) It is impossible to take into account all the test vehicles, resulting in missed judgments.

(5) The standards mastered by each examiner on the syllabus will be inconsistent.

At present, the computer pile test system has been used in the pile test process, which has changed the pile test subjects in the driver's field test process, which has always been visually tested by the examiner on the field, so as to avoid the above-mentioned limitations caused by subjective and objective factors such as the examiner's observation angle. The shortcoming of not being timely and accurate enough can reduce the occurrence of disputes by standardizing the venue examination standards.

Among such products, the main features of the products of Nanjing Duolun Precision System (Far East) Co., Ltd. are:

(1) Install a set of mechanical servo mechanism on the test car, and brake immediately if the students violate the regulations.

(2) Install a wireless system on the test vehicle to transmit control signals and detection signals to the computer room.

(3) Install a set of sensors on the transmission shaft, and judge the forward, backward and stopped states of the vehicle by detecting the forward rotation, reverse rotation, and speed of the transmission shaft.

(4) Install a sensor in front of the engine fan blade to measure the operation of the cooling fan blade to determine whether the engine is turned off.

(5) On the site, 6-8 pairs of infrared sensors are used to calibrate the boundary of the warehouse, and the switch signal is connected to the machine room through the embedded cable.

(6) Two front and rear gantry frames are used on the field, and 6 pile poles are suspended by suspension springs. Hall-type proximity switches are installed at the bottom of the pile poles, and permanent magnetic steel is embedded in the ground. When the test vehicle hits the benchmark pole, the change control of the magnetic circuit The switch signal notifies the host.

Nanjing product design is relatively mature, and the system is reliable in operation. The two key technologies of system design: vehicle driving route and motion state discrimination; engine flameout discrimination both require special devices to modify the vehicle, and install various sensors and mechanical actuators on the vehicle. The disadvantages are that the trainees are not able to use the familiar training vehicles, which will affect the test results (most of the trainees are very dependent on the familiar vehicles); in addition, the driving route of the vehicles cannot be directly measured, but can be obtained through indirect conversion simulation, and the judgment of the route is not necessary. Inaccurate. There are also certain problems in flameout discrimination.

Compared with Nanjing products, the products produced by a company in Beijing have the following characteristics:

(1) Through the suction cup on the roof of the test vehicle, a set of mechanical servo mechanism is involved to transmit the vehicle position information to the machine room to detect the forward, backward and stopped status of the vehicle.

(2) A sensor is installed on the ignition key to monitor the ignition action, and the engine flameout is judged through the re-ignition start process.

(3) Mark the boundary of the warehouse by 6-8 pairs of infrared sensors on the field.

(4) Two front and rear gantry frames are used on the field, and 6 pile poles are suspended by suspension springs. Hall-type proximity switches are installed at the bottom of the pile poles, and permanent magnetic steel is embedded in the ground. When the test vehicle hits the benchmark pole, the change control of the magnetic circuit The switch signal notifies the host. Beijing products install a set of mechanical servo mechanism on the gantry, which can move the gantry to set the size of the garage according to the test model.

The Beijing product design cancels the special test vehicle, and there is no special mechanical actuator on the car suitable for students to use. The driving route of the vehicle involves a set of mechanical servo mechanisms to detect the position information of the vehicle through the suction cup on the roof of the test vehicle, and the test site can be set according to the vehicle model. The biggest problem reflected in practice is that the test time is too slow, and the driving route is still measured with a mechanical structure, which has low technical content and complicated maintenance.

Contents of the invention

The purpose of the invention is to overcome the deficiencies of the prior art, to propose a driver's pile test process automatic test system based on image sequence analysis, combining computer technology with digital image processing technology, through the camera Real-time digital image analysis of the captured image sequence to identify the vehicle trajectory can completely replace manual invigilation, improve the accuracy and objectivity of the test results, and have higher flexibility, reliability and practicality compared with similar products sex.

A kind of driver pile test process automatic test system that the present invention proposes is characterized in that comprising the following several parts:

(1) The gantry unit used to mark the size of the garage on the test site is composed of two or more gantry units composed of triangular beams and support frames;

(2) A plurality of pile poles used for examination on the beam of the gantry frame suspended by suspension springs, which consist of a pole body, a connecting screw, an up and down adjustment device and a Hall switch installed on the pole body;

(3) The camera bracket installed on the examination site and the camera fixed on the bracket;

(4) LED large-screen displays and speakers installed on the test site to prompt test information;

(5) Multiple groups of infrared receiving/emitting sensor units set on the test site to mark the garage boundary;

(6) be connected with this video camera and be arranged on the computer image processing unit in the computer room, it is made up of the image acquisition card installed in the computer equipment and the image processing program software stored in the computer storage unit;

(7) The wireless transmission engine flameout measurement unit consisting of two parts: transmitting and receiving, wherein the transmitting part is composed of an electromagnetic sensor and a signal shaping circuit connected to it, and a radio transmitter, which is adsorbed on the roof of the test vehicle; the receiving unit consists of Composed of radio receivers and signal processing circuits connected to them, installed in the machine room;

(8) The switching value acquisition unit is placed in the computer room and consists of a switching value acquisition card, an adapter, and a power supply. All the switching value signals collected on the site are connected to the acquisition unit by buried cables.

A kind of automatic test method that the present invention proposes adopts the automatic test system of driver's stake test process as mentioned above, it is characterized in that, combine with digital image processing and pattern recognition method, without mechanical contact, use computer to automatically judge the driver Whether to drive and operate according to the route stipulated in the test syllabus;

Said digital image processing method specifically comprises the following steps:

(1) Obtain the image information of the vehicle by using a camera and an image acquisition card;

(2) Use image difference, image binarization, edge erosion and smoothing filter to eliminate the noise interference of image information;

(3) Finally, use the coordinates of the center of gravity to describe the trajectory of the vehicle, and then obtain the digital motion parameters through direction coding;

Said pattern recognition method specifically comprises: adopting said digital image processing method to carry out parking discrimination; Reversal point discrimination in motion; Reversed direction and route in combination with infrared sensor signal to judge whether motion route is completed in the first warehouse as stipulated The identification method of entering into the B library after two entries and two entries;

Said using the computer to automatically judge whether the driver is driving and operating according to the route specified in the test outline specifically includes: using the vehicle pile test automatic measurement system software based on image sequence analysis stored in the computer to measure each pile test that occurs during the pile test. This situation is measured in real time and the test information is prompted through the LED large-screen display and the speaker in a timely and accurate manner.

The image difference method mentioned in the present invention can be used to extract moving target information by using the gray level difference of two frames of images; for the fixed part in the image sequence, the gray level difference value is zero, while the moving part then presents a certain gray level difference, so the difference image only has the information of the moving part; specifically, the following three methods can be included:

(1) Use each frame in the image sequence to make image difference with a static reference frame;

(2) Differentiate two adjacent frames in the image sequence, and then binarize the grayscale difference image to extract motion information;

(3) Two difference images are calculated by using three frames of images, and then a second difference is performed on the two difference images.

The said image binarization of the present invention can adopt fast fixed threshold value method to carry out segmentation, selects binarization threshold value T according to experiment, the grayscale value exceeds the 255 grayscale values of target area all to be set as the pixel above this threshold value, at threshold value The gray value of all the following pixels is set to zero of the background area.

The said edge erosion method of the present invention can be: the background of each frame of image produces shot noise due to the change of light direction and illumination, utilizes the characteristic that the area occupied by it in the image is much smaller than the area occupied by the vehicle, through Edge erosion eliminates these shot noise pixels; according to the size and distribution of other object images, perform one-dimensional or two-dimensional erosion to eliminate the width of any pixel; the smoothing filtering method is: the coordinates of the center of gravity of each frame of image constitute The sequence is the trajectory of the vehicle. Due to various disturbances on the scene, the calculated trajectory has sudden changes in details. The trajectory must be smoothed and filtered, that is, the current (x, y) coordinates are obtained by recursively recursively based on the weighted average of the four sample points before and after. .

The direction coding of the said vehicle trajectory of the present invention can adopt image pixel and 8 convolution kernels of Kirsch edge operator to do convolution, can obtain the maximum response of each checking specific edge direction, all 8 directions The maximum value is used as the output of the edge amplitude image, and the sequence number of the maximum response kernel constitutes the edge direction code, and the so-called center of gravity motion trajectory is coded with 8 degrees of freedom to obtain the direction change trajectory.

The method for discriminating the parking mode in the present invention can be: within a limited threshold, two adjacent frames of images are differentiated and then binarized to see if there are pixels with gray scale changes, and if not, the difference is made after 4 frames. If there are still no changing pixels, it can be determined that the vehicle is stopped.

The turning point discrimination method of said driving route of the present invention can comprise:

(1) When the vehicle enters the bottom of the garage and moves upward from the bottom of the garage (vehicle advances), record the maximum value y _max of the center of gravity y value;

(2) After calculating the vehicle center of gravity of each frame image, compare the y value with the y _max value, if y≥y _max , then set y _max =y, if y<y _max , then look at y _max minus y Whether the difference Δy is greater than the preset threshold T _h , if Δy≤T _h , it is considered that the vehicle has not turned back _; Then:

(3) When the vehicle reaches the roof of the warehouse and moves downward from the roof of the warehouse (the vehicle reverses), by recording the minimum y _min , if y≤y _min , set y _min =y, if y>y _min , then see Whether the difference Δy between y minus y _min is greater than the preset threshold T _h , if Δy≤T _h , it is considered that the vehicle has not turned back from the bottom of the reservoir; if Δy>T _h , it is considered that the vehicle has turned back from the bottom of the reservoir, and the Movement to the top of the warehouse (vehicle advances).

According to the present invention, the method of judging entering into the second warehouse after finishing the second entry and the second download in the A warehouse can be: from the route inversion point combined with the current garage position of the vehicle, it can be analyzed whether the current vehicle sticking operation conforms to the two Enter the following standards:

(1) The vehicle has entered the warehouse;

(2) The vehicle advances towards the top of the warehouse and detects a turning point (one up);

(3) The vehicle reverses towards the bottom of the warehouse, and a turning point is detected (one click);

(4) The vehicle advances toward the top of the warehouse, and detects a turning point (second upper);

(5) The vehicle reverses toward the bottom of the warehouse, and a turning point is detected (secondary);

(6) At this time, if the vehicle is already in the second warehouse, the posting to the warehouse is completed. If the vehicle presses the middle line of A and B warehouses (blocking the infrared rays in the middle), the paste library will fail.

Features and effects of the present invention:

(1) The present invention detects the outgoing line of the vehicle body with an infrared sensor, detects the rubbing pile pole with a magnetic induction sensor, detects the working state of the engine engine with an electromagnetic sensor, and transmits the moving image of the vehicle to the computer with a CCD camera, and detects the parking and parking of the vehicle through image processing. Whether the driving route meets the outline requirements. The process of pile inspection is automatically tracked and measured through the processing of vehicle motion images captured by the camera. Infrared sensors, gantry and pile poles assist the system to calibrate the garage boundary. The ignition signal detection uses the electromagnetic sensor that senses the engine ignition signal to automatically measure the engine working status throughout the process. .

(2) All test subjects do not need to modify the vehicle.

(3) Real-time measurement without preparation time, the test process is completely determined by the running time of the tested vehicle.

(4) The system can detect various situations in the process of pile testing (such as not following the prescribed route and sequence, rubbing pile poles, failing to move into the warehouse, vehicle body going out of the line, stopping twice halfway, turning off the engine, etc.) Timely and accurate computer monitoring instructions:

(5) Realized the automatic tracking and identification of vehicle pile test movement trajectory mode under various weather conditions in the field environment, focusing on solving the differential image processing and center of gravity extraction of moving vehicles under the background interference of strong wind, sunshine, clouds, rain and snow, etc., and then After the noise is removed by spatial smoothing and filtering, the target motion trajectory is matched according to the set 8-degree-of-freedom template, and the automatic recognition of the vehicle motion mode is realized.

Description of drawings

Fig. 1 is a schematic structural diagram of an embodiment of the examination form library site of the present invention.

Fig. 2 is a block diagram of the overall structure of the system circuit of this embodiment.

Fig. 3 is a schematic diagram of the distribution positions of infrared sensors on the site and Hall sensors on piles in this embodiment.

Fig. 4a is one of the schematic diagrams showing the positional relationship between the field and the camera in this embodiment.

Fig. 4b is the second schematic diagram of the positional relationship between the field and the camera in this embodiment.

Fig. 5 is a structural block diagram of the flameout detection unit in this embodiment.

Fig. 6 is a circuit schematic diagram of the transmitting unit of this embodiment.

Fig. 7 is the logical structure of the receiving unit in this embodiment.

Fig. 8 is the principle of the signal conditioning circuit of this embodiment.

Fig. 9 is the signal connection relationship of the switching value sensor in this embodiment.

Fig. 10 is the image acquisition process of this embodiment.

Fig. 11 is the logical structure of the acquisition card in this embodiment.

Fig. 12(a) is a fixed reference frame image in this embodiment.

Fig. 12(b) is a frame of image in the motion sequence of this embodiment.

Figure 12(c) is the difference image between (b) and (a).

Fig. 12(d) is the binarized image of (c) under the selected domain value.

Figure 12(e) is the image obtained by edge erosion in (d).

Fig. 13 is a histogram of Fig. 12(c).

Fig. 14 is a schematic diagram of an 8-DOF coding template for measuring the motion direction of the center of gravity in this embodiment.

Fig. 15(a) is the track of the original center of gravity measured in this embodiment.

Fig. 15(b) is the locus of the center of gravity after filtering in this embodiment.

Fig. 15(c) is the digitized center-of-gravity track represented by the direction code in this embodiment.

Fig. 16 is the flow chart of the program for judging the turning point of the vehicle in this embodiment.

Fig. 17 is the program flow of the system adjustment part and the examination part of the present embodiment.

FIG. 18 is a frame in the 3D model diagram of the vehicle traveling in this embodiment, which is used to analyze possible problems in the driving route.

Fig. 19 is a flow chart of the examination thread program in this embodiment.

Fig. 20 is the operation screen of the system adjustment function in this embodiment.

Fig. 21 is the measurement and operation screen of the system of this embodiment, and the red curve is the vehicle motion trajectory tracked and measured in real time.

Fig. 22 (a) is the standard driving route of the present embodiment.

Fig. 22(b) is the vehicle movement track pattern tracked and recorded in this embodiment.

Detailed ways

The present invention designs a kind of embodiment based on the driver pile examination process automatic test system of image sequence analysis, in conjunction with each accompanying drawing, describe in detail as follows:

The driver pile test process automatic test system of the present embodiment is made up of following several parts:

(1) The gantry unit on the test site, which hangs the pile pole through the suspension spring, plays the role of marking the size of the garage. There are 2 pairs for a single library and 3 pairs for a set library. The gantry frame consists of a triangular beam and a support frame, with a height of 3.4 meters, a width of 9 meters (full length), and a span of 8.5 meters, welded by steel pipes; the support frame is two columns, and the column and the triangular beam are fastened by screws. connected together. The suspension brackets of the pile rods are also connected to the beams by screw fasteners.

(2) Exam poles. It is composed of rod body, connecting screw, up and down adjustment device and Hall switch. The rod body is made of aluminum alloy tube. The Hall switch is installed at the bottom of the pile, and the gap between the Hall switch and the magnetic steel buried on the ground is adjusted with an up and down adjustment device. There are 6 pairs in a single library and 11 pairs in a set library.

(3) The infrared receiving/emitting sensor unit on the test site is used to mark the boundary of the garage. There are 5 groups for a single library and 8 groups for a set library.

(4) Camera bracket and camera.

(5) The outdoor LED large-screen display and speakers are used to prompt the examination information.

(6) Computer image processing unit. It is composed of image acquisition card, computer equipment and developed program software.

(7) Wireless transmission engine flameout detection unit. Among them, the transmitting unit is adsorbed on the roof, and is composed of an electromagnetic sensor, a signal shaping circuit, and a radio transmitter. The receiving unit is installed in the computer room and is composed of a radio receiver and a signal processing circuit.

(8) Switch value acquisition unit. All the switching signals on the site are connected to the switching value acquisition unit located in the computer room through buried cables. The unit is composed of a switching value acquisition card, an adapter and a power supply.

Fig. 1 is a schematic diagram of an embodiment of a single type garage site. In order to mark the site, two secondary gantry frames 101 are used, and 6 pile poles 102 form two garages, 107 is a warehouse, and 108 is a second warehouse. Among the 5 sets of infrared sensors, emission is 103a, reception is 103b, they are respectively arranged on the front, rear, left, and right sides of the field, the front of the field is provided with a driveway 104, and the camera 105 positioned at the rear of the field is used to take pictures of vehicle movement. A machine room 106 is provided on one side. In the figure, the pole 102 is suspended on the gantry. When a student collides with the pole, the Hall sensor on the pole will notify the computer and the program will detect it. Two horizontal and three vertical pairs of infrared sensors mark the boundary and center line of the two garages. If the vehicle crosses the boundary during motion, the infrared rays will be blocked and the program will be notified for processing.

Fig. 2 is the overall structure of the system circuit of this embodiment, including: CCD camera monitoring unit, infrared sensor unit, Hall sensor unit, flameout detection unit, signal conditioning circuit, computer image processing system, large-screen display, loudspeaker, printer. Among them, the infrared sensor detects whether there is an object crossing the sideline, the Hall sensor detects whether there is an object colliding with the pile pole, the sensor in the flameout detection unit monitors whether the engine is flameout, and the computer automatically tracks the movement trajectory of the vehicle in the camera picture, and Identify compliance with exam syllabus requirements. The output terminals of the above three units are connected to the input terminals of the signal conditioning circuit, because the sensor signals of the three units are all switching values, which need to be converted into TTL level by the interface circuit, and the flameout detection signal also needs signal shaping processing. The signal conditioning circuit and the sensor incoming line are placed in a junction box.

The image signal collected by the camera is directly connected to the image acquisition card in the computer.

The computer outputs the measurement results to the outdoor large-screen display through the image processing program software, and displays the qualified and unqualified information. When the test is unqualified, 6 kinds of errors are displayed at the same time: collision with piles; driving not in accordance with the prescribed route and sequence; Enter; the body exits the line; stop twice halfway; turn off the engine.

Fig. 3 is a diagram showing the distribution positions of infrared sensors on the field and Hall sensors on piles in this embodiment. They are distributed according to the size of the site and are used to mark the boundary of the garage. The Roman numerals I, II, III, IV, V, VI, VII, and VIII marked on the site in Figure 3 are the numbers of each infrared sensor group, while the Arabic numeral 1 marked on the site is ～11 is pile code. This is the case of double-model vehicle sets. If it is a single-model small garage, sensors II, III, IV, and piles 1, 2, 4, 5, and 6 are ignored; if it is a single-model large garage, sensors VI, VII, VIII, poles 7, 8, 9, 10, 11 are ignored.

The two garages of the present embodiment are 2.65m wide and 12.1m long, and the driveway is 9.075m wide. The camera support is 5m away from one end of the garage and 5m high, as shown in Figure 4, 109 among the figures is the test vehicle.

The specific structure of each circuit component of this embodiment is described in detail as follows;

1. CCD camera monitoring unit

此时拍摄最远范围是27.75米。如图4a，垂直视角是：

如图4b。镜头与CCD芯片成像尺寸相同，选用1/2”摄像机则镜头也为1/2”，成象为6.5×4.8mm²。镜头焦距决定视角大小，用6毫米焦距，由透镜成像公式及相似三角形可计算出镜头水平视角为垂直视角为

均符合计算要求。选用AVENIR的SG0614CSE型6.0mm F1.4自动光圈镜头以适应野外环境下光照条件的变化，而摄像机必须有高分辨率才能检测出车辆微小运动。车辆在摄像机视野最远端纵向移动1米距离时，摄像机垂直视角变化量是

而摄像机垂直视角为43.6°，即是垂直视角的 ，用MINTRON的MTV-1881EX型600线黑白CCD摄像机，则车辆在最远端纵向移动1米时图象有4个像素的变化量。The camera shoots the site at a fixed position. Because the platform is fixed, the camera angle of view must cover the entire site. The largest model among the test vehicles 108 is Beijing 1041 (1.95m wide, 6m long). Considering the height of the tested vehicle, the height of the tripod is not included in the estimation of the horizontal viewing angle, then:

At this time, the farthest shooting range is 27.75 meters. As shown in Figure 4a, the vertical viewing angle is:

Figure 4b. The imaging size of the lens is the same as that of the CCD chip. If a 1/2” camera is used, the lens is also 1/2”, and the image size is 6.5×4.8mm ² . The focal length of the lens determines the size of the angle of view. With a focal length of 6 mm, the horizontal angle of view of the lens can be calculated from the lens imaging formula and similar triangles as The vertical viewing angle is

All meet the calculation requirements. AVENIR's SG0614CSE 6.0mm F1.4 automatic aperture lens is selected to adapt to changes in light conditions in the field environment, and the camera must have high resolution to detect the slight movement of the vehicle. When the vehicle moves longitudinally for a distance of 1 meter at the farthest end of the camera field of view, the vertical angle of view of the camera changes by

The vertical angle of view of the camera is 43.6°, that is, the vertical angle of view , with the MTV-1881EX type 600 black-and-white CCD camera of MINTRON, then the vehicle moves 1 meter longitudinally at the farthest end, and the image has a variation of 4 pixels.

The position and angle relationship of the camera in the system design of this embodiment is based on the size requirements of the pile test training standard site, which not only ensures that the entire site is within the camera screen, but also has as large an image edge resolution as possible. .

2. Flame-off detection unit

The structure of the flameout detection unit in this implementation is shown in FIG. 5 . The transmitting part is installed on the test vehicle, the engine ignition signal obtained by the electromagnetic sensor is processed by shaping and modulation, and then transmitted through the radio transmitter after modulation; the receiving and processing part is installed in the computer room, and the received radio signal is demodulated, output isolated, and transmitted to the signal conditioning circuit.

(1) Working principle of the transmitting unit

The circuit principle of the transmitting part of this embodiment is shown in FIG. 6 . The flameout measurement of the car is realized by detecting the electrical pulse of the engine. The electromagnetic sensor is clamped on any cylinder distributor cable on the high-voltage side of the ignition coil of the automobile engine. The working pulse current picked up by the probe is amplified by the LM121 amplifier and only the positive pulse is shaped and output to trigger the monostable 74221. This is a non-retriggerable monostable , the design pulse width is 330mS. Because the ignition signal period of the 4-cylinder engine is , the number of revolutions N is between 400 and 5000, so the period of the engine ignition signal is between 600mS and 48mS during operation. On the other hand, in order to prevent on-site interference, PT2262 wireless coding circuit must be used, and its coding sequence action frequency is several Hz, so the engine ignition signal cannot be directly applied, but the frequency is limited by 74221 monostable, and it follows the ignition signal frequency when the number of revolutions is low Action, when the number of revolutions is high, it is limited to around 3Hz. The launching part is adsorbed on the top of the car, no need to modify the car and it is easy to operate.

(2) The working principle of the receiving part

The logical structure of the receiving circuit of this embodiment is shown in FIG. 7 . Among them, the TWH9238 receiver is paired with the PT2262 transmitter to receive the engine ignition signal. After the signal conditioning circuit is shaped and counted, the count value is connected to the ACLD-7130 signal isolation interface inserted into the PCI slot of the computer through the signal adapter board ACLD-880 The data bus connected to the board is read into the computer by the program for processing.

3. Signal conditioning circuit

Figure 8 is the principle of the signal conditioning circuit. After the starting status line jumps low, the positive half cycle of the gate signal and the gate are released, the ignition signal enters the counter, and the lower jump edge of the gate signal closes the AND gate to stop counting. Therefore, the operating frequency of the engine ignition signal can be measured within the specified time period of the program, and if the count value is zero, it can be concluded that the engine is in a stalled state.

4. Hall sensor unit

Figure 9 is the signal connection relationship of the switch sensor. All the switching sensor signals from the field are also read into the computer by the program through the data bus connected to the ACLD-7130 signal isolation interface board inserted into the PCI slot of the computer through the signal adapter board ACLD-880.

5. Data Acquisition Card

In order to collect various signals to the computer for processing, a data acquisition card must be used. Because the library has 8 channels of infrared signals, 11 channels of magnetic induction switch signals and 1 channel of flameout detection, and flameout detection uses at least 5-bit signals (1-bit output, 4-bit input), a total of 23-bit input and 1-bit output. This embodiment selects the ACL-7130 32-channel isolated digital I/O card that can be inserted into the PCI slot of a computer, produced by Taiwan EVOC. At the same time, the ACLD-880 interface board of EVOC is used in the junction box to connect with the ACL-7130 board in the computer case for signal input and output, as shown in Figure 9.

The image processing method of this embodiment is described in detail as follows:

The difficulty of using image processing and pattern recognition methods to judge vehicle driving status in the field environment is the change of field weather conditions, such as rain, snow, wind, cloud, sunlight intensity, and sunlight angle, etc. will affect the quality of the captured image. When the light intensity decreases, the overall brightness of the image will decrease, and the contrast will become poor; when the camera lens is covered with dust or in rainy and snowy weather, the image clarity will decrease, and the edge of the scene will become blurred; in winter windy weather, the camera bracket will Vibration occurs, causing the background in the sequence of images to vibrate. These factors all make image analysis difficult. In addition, the image sequence processing speed must be greater than 4 frames per second so as not to lose vehicle motion state information (vehicle speed is lower than 10km/h). Therefore, the moving target processing method must meet the requirements in both speed and precision to make the system have practical value. This system has carried out a lot of experiments on various algorithms, and the actual algorithm proposed can meet the requirements of processing speed and precision directly on the PC.

The specific image processing method of this embodiment includes: image acquisition method; field background interference removal method caused by various weather conditions in the field environment; using 8-degree-of-freedom coding to make the direction of vehicle center of gravity position change into a digital trajectory description; The algorithm of parking discrimination; the method of judging the inflection point of the driving route from forward to backward and from backward to forward; the method of judging whether the driving route meets the requirements of the examination outline through the comprehensive analysis of the change of infrared signal occlusion and the trajectory of the center of gravity, as follows Explain one by one.

6. Image acquisition

The video signal of the camera is sent to the computer through the image acquisition board. The image captured by the camera is a standard PAL video, and the signal is transmitted to the video capture card through a 75-ohm coaxial cable. According to the control of the program, the acquisition card converts the video signal into a 512*512*24bit image and stores it in the computer memory, that is, the image in the memory is 24bit true color, and each pixel is composed of three primary colors of R, G, and B. In order to facilitate processing and reduce storage space, this embodiment converts the true color image into a 256-level grayscale image. For this reason, one of the three primary colors is randomly selected as the gray value of the corresponding pixel.

Fig. 10 is the image acquisition process, and Fig. 11 is the logical structure of the acquisition card, which is a DH-VRT-CG200/1 type 24-bit true-color single-channel image acquisition card of Daheng Company. Through the connection between the camera and the image card, the program can read in the digital image of the shooting scene, which provides basic data for the next image processing.

2. Moving target recognition

The purpose of image sequence analysis is to detect motion information, identify and track moving targets, and estimate motion parameters. When the interval between image frames is short, it is assumed that the gray scale of each part of the scene on the image is basically unchanged. Suppose the image at time t ₁ is g ₁ (x, y), and the image at time t ₂ is g ₂ (x, y). If the target shifts (Δx, Δy) during this period, it is assumed that g ₂ (x, y)=g ₁ (x-Δx, y-Δy), using image difference to extract moving target information from the gray level difference of two adjacent images. For the fixed part in the image sequence, the gray level difference is zero, while the moving part shows a certain gray level difference, so the difference image only has the information of the moving part.

There are two types _of image difference: one uses the image difference g _{I (} x, y)=g _R (x-Δx _I , y-Δy _I ). However, the field environment is changing at any time (such as the instantaneous background gray level changes caused by cloud movement in summer, the instantaneous shear strong wind on the field in winter and spring makes the camera tremble, and the shadow changes of other moving objects on the field, etc.). The calculated (Δx _I , Δy _I ) has a certain deviation from the actual value. In the system of this embodiment, after experiments and comparisons, this method is adopted when extracting motion trajectories.

The other is to use the difference between two adjacent frames in the image sequence, and then binarize the gray difference image to extract the motion information. The problem with this method is that the overlapping part of the vehicle between two frames cannot be detected when the vehicle is moving slowly; secondly, the detected moving target is larger than the real object. However, this method is more sensitive than the previous method for the micro-motion response of the target. In this embodiment, it is used to detect whether the vehicle stops moving. In order to solve the problem of the difference between adjacent frames, two difference images can be calculated by using three frames of images, and then the two difference images are subjected to the second difference, and finally the image is the Laplacian image of the original three frames of images. When the shape and motion vector of the moving object do not change too much in several frames of images, the second difference (Laplacian) image can be used to better extract the moving object. This method is more sensitive to the small movement of the target than the subtraction of two frames, but its calculation amount is three times that of the subtraction of two frames, and the space occupied is twice that of the subtraction of two frames.

Figure 12(a) of the accompanying drawings is a fixed reference frame image, and the upper right corner of the field background is the starting vehicle. Figure 12(b) is the image in the motion sequence, at this time the vehicle is moving to the location of the field A warehouse. Figure 12(c) is the difference image with the fixed frame, Figure 12(d) is the binarized image of the selected domain value, and the moving object is clearly expressed, Figure 12(e) is further obtained from Figure 12( d) It is obtained by edge corrosion, which has the effect of removing some interference points.

3. Selection of fixed reference frame (background image)

Since the test site is an outdoor environment, the background changes continuously due to the constant change of the weather environment, and the main factor of the change is the shadow change caused by the change of the sun's position. In order to minimize the impact of changes in the natural environment, a new fixed reference frame must be reselected before each test. During the entire test process, within a few minutes, small shadow changes will not cause any impact. The first frame of image collected after the start of each test is used as a fixed reference frame. In the first frame image, the vehicle is in the upper right corner of the image, which is just blocked by the gantry beam, and occupies a small image area, which is not much different from the image without the vehicle. It is feasible to use it as a fixed reference frame for actual system verification.

4. Image binarization

After the image difference, the gray value of the background part should be all zero, but in the actual measured image sequence, the background cannot be kept strictly unchanged. Affected by various factors, the background of each frame of the image in the sequence will change slightly, reflecting There is a small change in the gray value of the pixel at the same position, which is obviously smaller than the gray change caused by the background being covered by the vehicle. Figure 13 is obtained after the difference image in Figure 12c is histogram equalized, from which it can be seen that there are bright areas and a large number of stripes around the vehicle on the background, which are caused by interference. In order to eliminate this small grayscale change, the image needs to be binarized to separate the vehicle from the background. Analyzing the histogram of the difference image shows that there is no clear boundary between the target and the background. We use the fast fixed threshold method for segmentation, and the selected threshold T is obtained through continuous testing and adjustment.

Figure 12(c) of the attached drawing is histogram equalized to get Figure 13. The horizontal axis is the pixel gray value, and the vertical axis is the number of pixels with a certain gray value. Obviously, there are fewer pixels with a gray value near 128 , is the pixel of the moving target; the gray value of most of the pixels is below 70, which should be the pixels of the background. Therefore, we can choose a threshold value for binarization, and all pixels with a grayscale value exceeding the threshold value are set to the 255 grayscale value of the target area, and those below the threshold value are set to all zeros in the background area.

5. Edge corrosion

During the measurement process, the person on the edge of the field will be photographed on the screen, and his movement will affect the calculation of the center of gravity of the vehicle and the identification of the trajectory. At the same time, the shadow of the benchmark pole in the field will move with time. Although the angle of change is not large, it will cause errors if it is not dealt with. The common feature of these objects is that the area occupied by the image is much smaller than the area occupied by the vehicle, so it can be eliminated by edge erosion. The general edge erosion method can only reduce the width by 2 pixels at a time, and it takes multiple erosions to eliminate these small objects. When the scale of an object in a certain direction such as the vertical axis is relatively large, and the scale on the horizontal axis is small, as long as one-dimensional erosion is performed in the horizontal direction, small objects can be eliminated, and the object can be etched away by any number of pixels at a time. width, which can drastically reduce processing time.

Figure 12(e) of the accompanying drawings is the effect of etching the binarized Figure 12(d), which removes edge burr noise.

6. Center of gravity calculation and motion trajectory smoothing filter

一阶矩

二值化图象其背景灰度值都为零而目标子图灰度值都为1，则零阶矩由像素灰度值的和变为像素和，即区域面积。一阶矩由横、纵坐标分别相加得到。每帧图像重心坐标所构成的序列是车辆运动轨迹。由于现场各种干扰造成计算出的运动轨迹有细节突变，不经处理直接做模式匹配会得出错误的结果，故对运动轨迹进行如下的平滑滤波： ${\{}_{\mathrm{\&Delta;}{\stackrel{-}{y}}_n=\frac{(y_n-y_{n-1})+(y_n-y_{n-2})+(y_n-y_{n-3})}{6}}^{\mathrm{\&Delta;}{\stackrel{-}{x}}_n=\frac{(x_n-x_{n-1})+(x_n-x_{n-2})+(x_n-x_{n-3})}{6}}------{\{}_{{\stackrel{-}{y}}_n={\stackrel{-}{y}}_{n-1}+{\mathrm{\&Delta;}\stackrel{-}{y}}_n=\underset{i=1}{\overset{n}{\mathrm{\&Sigma;}}}\mathrm{\&Delta;}{\stackrel{-}{y}}_i}^{{\stackrel{-}{x}}_n={\stackrel{-}{x}}_{n-1}+{\mathrm{\&Delta;}\stackrel{-}{x}}_n=\underset{i=1}{\overset{n}{\mathrm{\&Sigma;}}}\mathrm{\&Delta;}{\stackrel{-}{x}}_i}$ 用( x， y)代替( i， j)得平滑后轨迹，原轨迹形状及变化趋势不变，但消除了突变扰动。Center of gravity coordinates zero moment

first moment

The background gray value of the binarized image is zero and the gray value of the target sub-image is 1, then the zero-order moment changes from the sum of pixel gray values to the sum of pixels, that is, the area of the region. The first-order moment is obtained by adding the horizontal and vertical coordinates respectively. The sequence formed by the coordinates of the center of gravity of each frame of image is the trajectory of the vehicle. Due to various disturbances on the scene, the calculated movement trajectory has sudden changes in details, and direct pattern matching without processing will give wrong results. Therefore, the following smoothing filter is performed on the movement trajectory: ${\{}_{\mathrm{\&Delta;}{\stackrel{-}{\mathrm{the\; y}}}_{\mathrm{no}}=\frac{({\mathrm{the\; y}}_{\mathrm{no}}-{\mathrm{the\; y}}_{\mathrm{no}-1})+({\mathrm{the\; y}}_{\mathrm{no}}-{\mathrm{the\; y}}_{\mathrm{no}-2})+({\mathrm{the\; y}}_{\mathrm{no}}-{\mathrm{the\; y}}_{\mathrm{no}-3})}{6}}^{\mathrm{\&Delta;}{\stackrel{-}{x}}_{\mathrm{no}}=\frac{(x_{\mathrm{no}}-x_{\mathrm{no}-1})+(x_{\mathrm{no}}-x_{\mathrm{no}-2})+(x_{\mathrm{no}}-x_{\mathrm{no}-3})}{6}}------{\{}_{{\stackrel{-}{\mathrm{the\; y}}}_{\mathrm{no}}={\stackrel{-}{\mathrm{the\; y}}}_{\mathrm{no}-1}+{\mathrm{\&Delta;}\stackrel{-}{\mathrm{the\; y}}}_{\mathrm{no}}=\underset{i=1}{\overset{\mathrm{no}}{\mathrm{\&Sigma;}}}\mathrm{\&Delta;}{\stackrel{-}{\mathrm{the\; y}}}_i}^{{\stackrel{-}{x}}_{\mathrm{no}}={\stackrel{-}{x}}_{\mathrm{no}-1}+{\mathrm{\&Delta;}\stackrel{-}{x}}_{\mathrm{no}}=\underset{i=1}{\overset{\mathrm{no}}{\mathrm{\&Sigma;}}}\mathrm{\&Delta;}{\stackrel{-}{x}}_i}$ Replace (i, j) with (x, y) to get the smoothed track, the shape and trend of the original track remain unchanged, but the mutation disturbance is eliminated.

Vehicle Movement Pattern Recognition

The method of identification and pattern matching of vehicle motion characteristic parameters based on pile test training standard driving route is as follows:

Direction template and driving trajectory analysis

By convolving image pixels with 8 convolution kernels of the Kirsch edge operator, the maximum response of each kernel to a specific edge direction can be obtained, and the maximum value of all 8 directions is used as the output of the edge amplitude image, and the maximum response The ordinal number of the core constitutes the edge direction code. Similarly, the trajectory of the center of gravity is coded with 8 degrees of freedom to obtain the direction change trajectory, and the coding template is shown in Figure 14. When the vehicle sticks to the warehouse, the trajectory is to reverse to the left, and then gradually turn downward. When the car enters the second warehouse, it changes to reverse downward, and finally stops at the bottom of the second warehouse. The description is 5→6→7→0; when moving the warehouse The trajectory is to stop at the top of the warehouse, stop at the bottom of the warehouse at the first time, stop at the top of the warehouse at the second time, and stop at the bottom of the warehouse at the second time. The description is 3→0→7→0→3→0→7→0; out of the warehouse and When reversing the warehouse, the trajectory is to drive upwards first, because it needs to pass through the second warehouse to leave the warehouse, so the trajectory should turn to the left and up. For going down to the car until going out of the warehouse at the bottom of the first warehouse, the description is 3→4→5→0 (out of the warehouse)→1→8→7. Since the camera is not directly above the target, and the center of gravity is biased by the shadow of the vehicle, there is a small deviation in the calculated trajectory direction, but the trend remains the same.

Figure 14 is the 8-degree-of-freedom encoding template of the center of gravity movement direction, Figure 15(a) is the original measured center-of-gravity trajectory, Figure 15(b) is the center-of-gravity trajectory after filtering, and Figure 15(c) is the center-of-gravity trajectory represented by the direction code . It can be clearly seen that the trajectory of the center of gravity encoded by the direction exactly matches the actual motion pattern of the vehicle.

Parking mode discrimination

The difference between parking when moving from backward to forward or from forward to backward and halfway parking during driving is that the former has the opposite direction of movement before and after parking, and the latter has the same direction of movement before and after parking. The position of the center of gravity remains unchanged when the vehicle stops, but when the speed is extremely slow or turning, the position of the center of gravity obtained by image processing may also remain unchanged, so the position of the center of gravity remains unchanged only for the vehicle to stop. To this end, two adjacent frames of images are differentiated and then binarized to see if there are pixels with grayscale changes. If there are no pixels, the difference is made after 4 frames. If there are still no changed pixels, it can be determined that the vehicle is stopped. In order to avoid the impact of body vibration caused by braking (the pixel change caused by vibration may be large, exceeding the pixel change when the vehicle is driving far away), a threshold value should be limited to distinguish between vibration and inflection point. Discrimination of turning point of driving route

The process of the vehicle turning from forward to backward and from backward to forward is defined as the turning back of the driving route. We cannot simply rely on the difference in driving direction before and after parking to judge the turning point of the driving route. This is because during the test, students usually use the brakes to stop, then shift gears and start. When the brakes are relatively sharp, it will cause the body to vibrate. When the vehicle is relatively close to the camera, the pixel changes caused by this vibration may be very large, which will cause misjudgment. Such as missing a return and misjudging the direction. To solve this problem, we use a new method to judge the steering of the vehicle.

After analysis, it can be seen that when the vehicle body is trembling, its center of gravity changes back and forth in a small range, and its value ranges from 2 to 3, and finally it will stabilize at one point; If you change in the opposite direction, you will soon cross the range of 2 to 3. According to this feature, the following methods can be adopted to judge the turn-back of the vehicle. Since the main concern is the change of the center of gravity of the vehicle in the garage, the change of the center of gravity of the vehicle is basically in the y direction at this time, and the change in the x direction is very small, which will not affect the judgment result.

When the vehicle enters the bottom of the garage and moves upward from the bottom of the garage (the vehicle moves forward), record the maximum value y _max of the center of gravity y value, and compare the value of y with the value of y _max after calculating the center of gravity of the vehicle for each frame of image , if y≥y _max , then let y _max =y, if y<y _max , check whether the difference Δy between y _max minus y is greater than the preset threshold T _h , if Δy≤T _h , consider the vehicle is still If there is no turning back, if Δy>T _h , it is considered that the vehicle has turned back and starts to move toward the bottom of the warehouse (vehicle reverse); The smallest y _min , if y≤y _min , set y _min =y, if y>y _min , check whether the difference Δy between y minus y _min is greater than the preset threshold T _h , if Δy≤T _h , It is considered that the vehicle has not turned back from the bottom of the warehouse, and if Δy>T _h , it is considered that the vehicle has turned back from the bottom of the warehouse and starts to move toward the top of the warehouse (the vehicle is moving forward).

The flow chart of the program for judging the turning point of driving is shown in Figure 16.

Discrimination of driving route

In this system, by comprehensively analyzing the changes of the infrared signal occlusion and the track of the center of gravity obtained through image processing, it can be judged whether the driving route of the vehicle meets the requirements of the examination outline. According to the change of the vehicle's occlusion of the five infrared signals, most of the vehicle's driving route can be obtained, and the center of gravity trajectory is mainly used to judge the vehicle's midway stop and double up and down during the transfer process. This is because the vehicle stops halfway It will not change the infrared rays, and when the vehicle enters the garage (it can be judged by the change of infrared rays), it will only block the infrared rays in the middle during the process of moving the warehouse. At this time, the specific position and location of the vehicle cannot be obtained only based on the infrared rays. mobile information.

Use 3DS MAX software to build a 3D model of the vehicle, the test site and the movement route for simulation analysis, and take pictures of the site from the perspective of the camera. During the test, the vehicle travels along the prescribed route to block the infrared rays as follows:

(1) When the test starts, the vehicle does not cover the infrared rays;

(2) When the vehicle is attached to the garage, first cover the infrared rays on the right; continue to reverse, the body blocks the infrared rays on the right and the middle at the same time; then only blocks the infrared rays in the middle; the rear of the car enters the garage, and blocks the infrared rays in the middle and in front of the garage at the same time; then It only blocks the infrared rays in front;

(3) After the vehicle enters the garage completely, no infrared rays are blocked;

(4) During the process of moving the vehicle, it has to go up and down in the garage, which will block the infrared rays in the middle;

(5) After the storage transfer is completed, no infrared rays are blocked.

(6) When the vehicle leaves the warehouse, first block the infrared rays in the middle; when the front of the car leaves the warehouse, block the infrared rays in the middle and in front of the garage at the same time; then block the infrared rays in the left, middle and front of the garage; then block the infrared rays in the left and in front of the garage; After the tail is out of the warehouse, only the infrared rays on the left are blocked;

(7) After leaving the warehouse and getting on the driveway and stopping, no infrared rays are blocked.

(8) During the process of reversing the vehicle from the driveway to the garage (B), the situation of blocking infrared rays is symmetrical to that of pasting the warehouse, except that it changes from the right to the left;

(9) After the test is over, the rear of the car will cover the infrared rays behind the garage when it exits the second garage, and the whole process is over.

According to the above analysis, the changes in the infrared signal shielding of the vehicles traveling according to the standard driving route stipulated in the outline during the whole process are obtained, that is, the logic truth table of the infrared sensor. According to this, it is possible to judge the route stage of the vehicle and some route errors in the test procedure.

Table 1 is the standard change state of the infrared signal of the large garage. Roman numerals are infrared sensor numbering (referring to Fig. 3 of accompanying drawing), when traveling according to outline regulation route, in stage one blocking sequence is No. II, No. V, and No. II recovers and No. I is blocked after the rear of the car enters storage. The second stage is that the vehicles gradually enter the warehouse A, and the blocking sequence is No. II and No. V. After the rear of the car enters the warehouse, No. V recovers and No. I is blocked. By analogy, the truth table of the state change of the infrared sensor signal at each stage can be obtained. If you do not drive according to the regulations during the test, various types of errors can be judged according to the current trajectory position of the center of gravity, combined with the signal status of the infrared sensor.

Based on the route inversion point combined with the current garage location of the vehicle, it can be analyzed whether the current vehicle sticking operation meets the standard of two up and two down.

Whether the vehicle stops halfway can be judged by the combination of parking and inversion point. The specific software process is described in the next section. Figure 18 of the accompanying drawings is a 3D model diagram, which we use to analyze possible problems in the driving route:

(1) This is the angle of view seen from the installation position of the camera at the rear of the venue.

(2) Initially, the vehicle enters from the upper right corner of the screen, and travels to the left and then down according to the lane in reverse mode.

(3) The track on the field is the standard driving route.

(4) The thin lines on the field are infrared sensor rays.

(5) The four corners and the middle of the field are piles.

(6) Because the simulation is a dynamic process, Figure 18 is just one of the snapshots. The image sequence analysis flow chart and program used by the system.

The software system is divided into two parts: system adjustment and examination process. System adjustment is the auxiliary function part, Table 1 Infrared sensor logic state truth table serial number I II III IV V stage description error description

In the table: set 1 to cover the infrared rays, and 0 otherwise; the cells with gray shading in the table represent the infrared rays whose occlusion situation changes. It can be used to detect the on and off of infrared rays, the clutch of magnetic induction switches, the angle of view of cameras, and the radio signal of flameout detection. This part of the function is used in the following two aspects: First, when the equipment is installed, it is necessary to adjust the camera and straighten the infrared rays, and observe the signal conditions according to the software to help adjust; All aligned, all the piles return to their positions, and the flameout detection signal is received normally.

The examination process is the main functional part. It detects and judges the entire pile examination process, gives error prompts and examination results, and prints score cards.

flow chart

The software system structure is shown in Fig. 17 of the accompanying drawings, and it is divided into two action program flows of a system adjustment part and an examination part. The examination thread program flow is shown in Figure 19, and it is completely programmed according to the aforementioned image processing steps. The image processing process expressed in Fig. 19(a) is as follows:

After starting the program (double-click the icon on the computer screen), first collect an image into the computer memory to become a grayscale image, and the program performs digital difference to form a differential image. The area with gray value is the moving target object, and the other Fixed background objects are subtracted to zero gray value. Because only the center of gravity of the moving object needs to be calculated, and the area of the moving object is represented by 1, the product operation of the gray value of a certain point pixel and its coordinate can be simplified to an addition operation to improve the processing speed, so the program does image processing. Binary processing. At this time, a fixed gray scale value is used as the boundary between the target region and the background region, and those smaller than this threshold value are considered as background pixels, and all are set to zero; those greater than this threshold value are considered to be target pixels, and all are set to 1. However, due to the influence of ambient light changes and noise, there are some pixels with strong gray values around the moving vehicle, resulting in the existence of some sporadic pixels with a value of 1 around the object after sorting by threshold value (shot noise). , the program removes these noises by edge erosion.

After the image is processed cleanly, the program can use the center of gravity formula to calculate the center of gravity coordinates of the current moving object, or because of the interference of the environment, the center of gravity coordinates calculated each time will be disturbed, and smoothing filtering is required to obtain the moving trend trajectory of the object Coordinates, compare the (x, y) value of the current center of gravity coordinates with the (x, y) value calculated last time to obtain the motion direction of the object, and use the aforementioned 8-degree-of-freedom direction template to perform motion direction matching, and the most The direction of a degree of freedom close to the current object is used as the direction of motion of the current object, and the code of the current object’s motion direction is obtained from the digital code of this direction, such as moving to the right is 1; moving to the right is 2; moving forward is 3; moving to the left is 4; moving to the left is 5; left rear is 6; backward movement is 7; right rear is 8; zero means stop.

The object movement direction code obtained by each calculation forms the digital trajectory of the object movement, which is very convenient for the next step of the program to process the movement mode.

Sports mode processing is broken down into the following points:

(1) You cannot stop twice in the whole journey.

(2) Vehicles are allowed to turn backwards only in the garage.

(3) Vehicles are not allowed to cross the center line before there is no turning point.

(4) The whole process cannot exceed the boundary (according to the infrared signal indication).

(5) Do not collide with the pile during the whole process (according to the Hall sensor signal).

(6) The flame cannot be turned off during the whole process (according to the flameout detection signal).

(7) After 4 turns of two ups and two downs, the vehicle must move to the B warehouse and cannot stay in the A warehouse or hold down the center line.

(8) After the vehicle is moved to the second warehouse, it crosses the center line and enters the A warehouse, and then goes out of the A warehouse to the apex of the lane.

(9) From the apex of the lane, turn the library into the second library and end the exam.

program

According to the aforementioned image processing principle and motion pattern recognition method, a set of automatic measurement system software for vehicle pile testing based on image sequence analysis is developed with VC++ programming language under WIN95/98 environment, and the program is directly installed under WIN95/98 environment .

Figure 20 of the accompanying drawings is the running screen of the system adjustment function, and Figure 21 is the running screen of the system measurement, and the red curve is the vehicle motion trajectory tracked and measured in real time.

Invention effect

For the purpose of designing and developing a computer image processing and recognition system for vehicle pile examination subjects, the vehicle movement pattern recognition method in the field environment based on image sequence analysis proposed by the present invention realizes automatic tracking and recognition of vehicles in various weather conditions in the field environment The pile test motion trajectory mode focuses on solving the differential image processing and center of gravity extraction of moving vehicles under background interference such as strong wind, sunlight, clouds, rain and snow, and then removes noise through spatial smoothing filtering, and then matches the target motion according to the set 8-degree-of-freedom template Trajectory, to realize the automatic recognition of vehicle movement mode. Through nearly 2,000 experiments on three models of Beijing 1041, 121 and Jetta in three months in winter, satisfactory results have been achieved.

Figure 22 is a comparison between the actual measurement results and the standard route of the Ministry of Public Security's post examination subject. It can be seen that compared with the standard driving route in Figure 22(a), Figure 22(b) shows that the system accurately tracks and records the vehicle movement trajectory mode.

System Test Method

Sensor System Test

(1) start the computer system;

(2) Start the system software under the WIN95/98 interface, and choose to enter the system adjustment function interface;

(3) Select the infrared and pole detection system for the test object, and at this time, you can see the working status of each infrared sensor and the signal state of the pole according to the software system prompts;

(4) Select the CCD camera monitoring system in the test object selection box, and the test site screen will appear at this time, and you can check whether the CCD camera is working normally;

System performance test

(1) start the computer system;

(2) Start the system software under the WIN95/98 interface, and choose to enter the system measurement function interface;

(3) Enter the student's name, test admission ticket number, and select the type of garage;

(4) Click the "Start" button, the system starts to work, and the running status of the vehicles on the site can be monitored from the camera screen;

(5) If the vehicle touches the bar at any time, the system will immediately prompt an error, stop the test at the same time, and give a message of unqualified test results;

(6) If the line (infrared) is pressed at any time, the system will prompt an error immediately, stop the test at the same time, and give a message of unqualified test results;

(7) If the vehicle is still in the A warehouse (including a part of the vehicle in the A warehouse) after two ups and downs when moving the warehouse, the system will immediately prompt an error, stop the test at the same time, and give a message of unqualified test results;

(8) If there are 2 stops in the middle, the system will immediately prompt an error, stop the test at the same time, and give a message of unqualified test results;

(9) If the vehicle is turned off at any time, the system will immediately prompt an error, stop the test at the same time, and give a message of unqualified test results;

(10) If none of the above errors occurs, after reversing into the second warehouse, the system immediately prompts that the test is completed, and gives the test result passing information, and the system automatically resets simultaneously (the driver drives the car out of the garage to the starting point);

(11) Repeat at step 3.

Claims (9)

1, a kind of driver pile examination process automatic test system is characterized in that comprising the following several parts: (1) The gantry unit used to mark the size of the garage on the test site is composed of two or more gantry units composed of triangular beams and support frames; (2) A plurality of pile poles used for examination on the beam of the gantry frame suspended by suspension springs, which consist of a pole body, a connecting screw, an up and down adjustment device and a Hall switch installed on the pole body; (3) The camera bracket installed on the examination site and the camera fixed on the bracket; (4) LED large-screen displays and speakers installed on the test site to prompt test information; (5) Multiple groups of infrared receiving/emitting sensor units set on the test site to mark the garage boundary; (6) be connected with this video camera and be arranged on the computer image processing unit in the computer room, it is made up of the image acquisition card installed in the computer equipment and the image processing program software stored in the computer storage unit; (7) The wireless transmission engine flameout measurement unit consisting of two parts: transmitting and receiving, wherein the transmitting part is composed of an electromagnetic sensor and a signal shaping circuit connected to it, and a radio transmitter, which is adsorbed on the roof of the test vehicle; the receiving unit consists of Composed of radio receivers and signal processing circuits connected to them, installed in the machine room; (8) The switching value acquisition unit is placed in the computer room and consists of a switching value acquisition card, an adapter, and a power supply. All the switching value signals collected on the site are connected to the acquisition unit by buried cables. 2. an automatic test method that adopts the driver's pile test process automatic test system as claimed in claim 1, is characterized in that, combines with digital image processing and pattern recognition method, automatically discriminates driving with computer without mechanical contact Whether the driver is driving and operating according to the route stipulated in the test outline; Said digital image processing method specifically comprises the following steps: 1) Obtain the image information of the vehicle by using a camera and an image acquisition card; 2) Then use image difference, image binarization, edge erosion and smoothing filter to eliminate the noise interference of image information; 3) Finally, use the coordinates of the center of gravity to describe the trajectory of the vehicle, and then obtain the digital motion parameters through direction coding; Said pattern recognition method specifically comprises: adopt said digital image processing method to carry out parking discrimination; Reversed point discrimination in motion; The method of judging the entry into the B library after entering the second round; Said using the computer to automatically judge whether the driver is driving and operating according to the route specified in the test outline specifically includes: using the vehicle pile test automatic measurement system software based on image sequence analysis stored in the computer to measure each pile test that occurs during the pile test. This situation is measured in real time and the test information is prompted through the LED large-screen display and the speaker in a timely and accurate manner. 3. The automatic test method as claimed in claim 2, characterized in that, said image difference method is to extract moving target information with the gray scale difference of two frames of images; for its gray scale difference of fixed part in the image sequence The value is zero, and the moving part shows a certain grayscale difference, so the difference image only has the information of the moving part; specifically, it includes the following three methods: (1) Use each frame in the image sequence to make image difference with a static reference frame; (2) Differentiate two adjacent frames in the image sequence, and then binarize the grayscale difference image to extract motion information; (3) Two difference images are calculated by using three frames of images, and then a second difference is performed on the two difference images. 4. The automatic testing method as claimed in claim 2, characterized in that, said image binarization adopts the fast fixed threshold method to segment, selects the binarization threshold T according to experiments, and the gray value exceeds the threshold value above the threshold. All pixels are set to the 255 gray value of the target area, and the gray values of all pixels below the threshold are set to zero in the background area. 5. The automatic test method as claimed in claim 2, characterized in that, said edge erosion method is: the background of each frame of image produces shot noise due to the variation of light direction and illuminance, and utilizes it to account for in the image. The area occupied by the vehicle is much smaller than the area occupied by the vehicle, and these shot noise pixels are eliminated through edge erosion; according to the size and distribution of other object images, one-dimensional or two-dimensional erosion is performed to eliminate the width of any pixel; The smoothing filtering method is as follows: the sequence formed by the coordinates of the center of gravity of each frame of image is the trajectory of the vehicle. Due to various disturbances on the scene, the calculated trajectory has a sudden change in detail, so the trajectory must be smoothed, that is, according to the four sample points The weighted mean is recursively obtained to obtain the current coordinates. 6. The automatic testing method as claimed in claim 2, characterized in that, the direction encoding of said vehicle trajectory adopts 8 convolution kernels of image pixels and Kirsch edge operator to do convolution, and each can be obtained Check the maximum response of a specific edge direction, the maximum value of all 8 directions is used as the output of the edge amplitude image, the serial number of the maximum response kernel constitutes the edge direction code, and the trajectory of the center of gravity is coded with 8 degrees of freedom to obtain the direction change trajectory . 7. The automatic testing method as claimed in claim 2, characterized in that, said parking mode discrimination method is: within a limited threshold, two adjacent frames of images are differentiated and then binarized to see if there is a grayscale change If there are no pixels, then make a difference at an interval of 4 frames. If there are still no changing pixels, it can be determined that the vehicle has stopped. 8. The automatic testing method as claimed in claim 2, characterized in that, the turning point discrimination method of said driving route comprises: 1) When the vehicle enters the bottom of the garage and moves upward from the bottom of the garage, that is, the vehicle moves forward, record the maximum value y _max of the center of gravity y value; 2) After calculating the vehicle center of gravity of each frame image, compare the y value with the y _max value, if y≥y _max , then set y _max =y, if y<y _max , then look at the difference between y _max minus y Whether Δy is greater than the preset threshold T _h , if Δy≤T _h , it is considered that the vehicle has not turned back, if Δy>T _h , it is considered that the vehicle has turned back, and it starts to move towards the bottom of the reservoir, that is, the vehicle reverses; and vice versa: 3) When the vehicle reaches the roof of the warehouse and moves downward from the roof of the warehouse, that is, the vehicle reverses, by recording the minimum y _min , if y≤y _min , then let y _min =y, if y>y _min , then look at y Whether the difference Δy minus y _min is greater than the preset threshold T _h , if Δy≤T _h , it is considered that the vehicle has not turned back from the bottom of the storage, if Δy>T _h , it is considered that the vehicle has turned back from the bottom of the storage Library top movement. 9. The automatic testing method as claimed in claim 2, characterized in that, the said method of judging entering into the second library after the first library completes the second entry and the second entry is that the route inversion point is combined with the garage where the vehicle is currently located Location, you can analyze whether the current vehicle sticker warehouse operation meets the standard of two-in-two-down: (1) The vehicle has entered the warehouse; (2) The vehicle moves toward the top of the warehouse, and a turning point is detected; (3) The vehicle reverses toward the bottom of the warehouse, and a turning point is detected; (4) The vehicle moves toward the top of the warehouse, and a turning point is detected; (5) The vehicle reverses toward the bottom of the warehouse, and a turning point is detected; (6) At this time, if the vehicle is already in the second warehouse, the posting to the warehouse is completed. If the vehicle presses the middle line of the A and B warehouses, the paste warehouse will fail.

Images