CN106205170B - A kind of intersection accurate parking device and method for automatic driving - Google Patents

Abstract

The present invention provides a device and method for precise parking at intersections for automatic driving. The device includes an image acquisition device, an electronic control unit, a millimeter-wave radar and an odometer, wherein: a camera is used to collect zebra crossings, parking lines and Image of traffic lights above the road; millimeter-wave radar is used to detect information about vehicles ahead; the electronic control unit can detect zebra crossings, stop lines, and traffic lights, and when the stop line enters the blind spot of the camera, the number of pulses returned by the odometer can be used Calculate the distance between the vehicle and the stop line, and realize the precise parking of the autonomous vehicle at the stop line at the intersection. On the other hand, according to the data detected by the millimeter-wave radar, the electronic control unit realizes precise parking according to the position of the preceding vehicle and the status of the signal lights when there is a vehicle ahead. The present invention solves the problem of how to accurately stop the automatic driving vehicle at the stop line at the intersection.

Description

A kind of intersection accurate parking device and method for automatic driving

technical field

The present invention relates to a parking device and method in the field of traffic technology, in particular, to a precise parking device at an intersection for an automatic driving vehicle, and an implementation method corresponding to the device.

Background technique

With the rapid development of modern society and economy, automobiles have become an important means of transportation for people to travel. With the increasing number of vehicles, traffic pressure is also increasing, and traffic accidents are becoming more and more frequent. In order to reduce the human factor in traffic accidents, vehicle autonomous driving technology has become an important research hotspot. Due to its complex traffic scenes and strict traffic rules, urban traffic has become a difficult point in the field of autonomous driving research. Among them, the intersection is the place where traffic flows in all directions. It is an important guarantee to ensure the smooth traffic at intersections to stop the vehicle accurately at the stop line when instructed to stop.

At present, there are many technologies and methods in the field of intelligent driving dedicated to the identification of intersections and the detection of signal lights, such as:

The Chinese invention patent with publication number CN104751679 A discloses an "autonomous vehicle system for driverless delivery vehicles at 'cross' intersection", which can stop according to the status of red and green at the intersection, but cannot get the vehicle for the detection of traffic lights. At the location of the intersection, in the urban environment, the intersection parking needs to comply with the traffic regulations. Instead of detecting the red light signal to stop, the vehicle must be stopped at the stop line.

The Chinese invention patent with publication number CN105740827 A discloses "a stop line detection and ranging algorithm based on fast marker connectivity", and the Chinese invention patent with publication number CN103488976 A discloses "a stop line based on intelligent driving Methods of real-time detection and distance measurement".

Both of the above two invention patents can detect whether there are zebra crossings and stop lines on the road, and give the distance between the vehicle and the stop line.

However, in the automatic driving system, the camera is generally installed above the windshield of the car, and there is a blind spot where the field of view is blocked by the front of the car. How to correctly guide the automatic driving vehicle to the stop line when the camera loses the zebra crossing and stop line signals is a problem. issues that need resolving. In addition, at intersections, there are often vehicles ahead in the current lane, so it is necessary to control the stop position of the vehicle based on the position of the vehicle ahead.

SUMMARY OF THE INVENTION

In view of the defects in the prior art, the purpose of the present invention is to provide a vehicle intersection precise parking device and method for automatic driving, so as to solve the problem of how to accurately park the automatic driving vehicle at the intersection at the intersection.

According to one aspect of the present invention, there is provided an intersection precise parking device for automatic driving, comprising: an image acquisition device, an odometer, a millimeter-wave radar and an electronic control unit, wherein:

The image acquisition device is used to collect images of zebra crossings, stop lines and traffic lights on the road, and transmit the collected image information to the electronic control unit;

The odometer is used to obtain the angle turned by the vehicle through counting pulses, record the number of turns of the wheels, and transmit the obtained counting pulse signal to the electronic control unit;

The millimeter wave radar is used to detect the position and speed information of dynamic and static obstacles in front of the vehicle;

The electronic control unit is used to receive and process the image information collected by the image acquisition device, and identify the zebra crossing, lane line and traffic signal information in the image; it is used to receive the position and speed of the obstacle in front of the vehicle detected by the millimeter wave radar information; used to calculate the distance traveled by the vehicle through the counting pulse signal obtained by the odometer; and based on the above information, send control instructions to the underlying actuator of the autonomous driving vehicle.

Preferably, the image capture device is installed on the upper part of the front windshield of the vehicle, and the installation angle ensures that the field of view can cover the zebra crossing on the ground, the stop line sign and the traffic lights above the road.

Preferably, there are two groups of odometers, and both groups of odometers include a gear encoder and a measuring head, wherein:

The gear code disc is installed on the rear axle, and the measuring head is installed on the frame; when the vehicle is running, the gear code disc rotates with the wheel, and the measuring head remains stationary. Induction signals of different intensities are generated on the sensor, and the induction signals are converted into count pulse signals, so as to obtain the angle of the gear code disc or the turning of the vehicle.

Preferably, the millimeter-wave radar is installed at the front bumper of the vehicle to detect whether there are other vehicles ahead, and the position and speed information of the other vehicles.

According to another aspect of the present invention, there is provided a method for precise parking at an intersection for automatic driving, the method comprising the following steps:

Step 1, the image acquisition device collects the image, and the electronic control unit processes the image;

The image acquisition device collects images of zebra crossings, stop lines and traffic lights on the road, and transmits the collected images to the electronic control unit;

The electronic control unit performs inverse perspective transformation on the image, and uses the checkerboard to calibrate the transformation parameters to obtain the proportional relationship between the image pixels and the world coordinates, as well as the pixel position of the front of the vehicle in the image coordinates; the electronic control unit processes the image, including zebra crossing detection, stop Line detection and traffic light detection;

Step 2. Use the odometer to record the distance traveled by the vehicle;

Step 3. Use millimeter wave radar to detect the position and speed information of obstacles in front of the vehicle;

Step 4. Intersection behavior decision;

The electronic control unit assists the autonomous vehicle at the intersection according to the perception information obtained by the image acquisition device, including the stop line and its distance, the status of the traffic light, and the position and speed information of the vehicle ahead detected by the millimeter-wave radar decision making;

When there is no other vehicle in front of the vehicle, the parking line is used to make an auxiliary decision for the autonomous vehicle at the intersection. The auxiliary decision uses two distance information: one is the parking line detected before the parking line enters the blind spot of the image acquisition device. The distance between the line and the front of the car, and the second is the distance traveled by the vehicle since the time when the stop line disappeared;

From the moment when the stop line disappears, when the distance traveled by the vehicle reaches the distance between the stop line and the front of the vehicle, a parking control command is issued, so that the vehicle can accurately stop at the stop line;

When there are other vehicles in front of the vehicle, the electronic control unit makes an auxiliary decision for the autonomous vehicle at the intersection according to the position and speed information of the preceding vehicle detected by the millimeter-wave radar: when the image acquisition device detects a signal light, And the status of the signal light is red, the motion state of the vehicle in front is detected. If the vehicle in front has not stopped, it means that the vehicle has not reached the intersection and needs to continue to drive at a distance from the vehicle in front; if the vehicle in front has stopped, it will be detected by millimeter-wave radar. When the distance to the vehicle in front is reached, the self-driving vehicle will be guided to the rear of the vehicle in front, and the vehicle will stop according to the set parking distance.

Preferably, in step 1, the zebra crossing detection method adopts an edge detection method combined with the local statistical features of the zebra crossing, so as to achieve the purpose of accurately identifying the zebra crossing.

Specifically, the image collected by the image acquisition device is converted into a grayscale image, and then the image is rotated according to the main gradient direction of the image, the edge of the rotated image in the vertical direction is extracted, and the number of edge points in each line of the image is counted. Estimate the range of the zebra crossing, and then count the number of edge points in each column within the range; the number of points in the column where the edge of the zebra stripe is located is much larger than other columns, calculate the interval between these columns with more points, and find the variance of these intervals, if The stripes are evenly distributed, and the variance value will be small, based on which to determine whether there is a zebra crossing.

Preferably, in step 1, the stop line detection method identifies the stop line in combination with the detection result of the zebra crossing. Specifically, after the zebra crossing is detected, take the lower edge of the zebra crossing to the last line of the image as the region of interest, use Hough straight line transformation to extract all straight lines in the area, and then extract the stop line according to the length, duty cycle, and direction constraints of the stop line. Line's line parameter in the image to calculate the distance of the vehicle to the stop line.

Preferably, in step 1, the traffic signal detection method adopts a color space-based image segmentation method and an area-to-perimeter ratio to detect and identify the traffic signal. Specifically, the image is first converted from the RGB color space to the LAB color space, and the red and green areas in the image are segmented; then the candidate areas are screened according to the area and shape of the area, and the candidate area is detected whether there is the inherent frame feature of the traffic signal. , to determine the presence or absence of traffic lights, and to determine the status of the lights based on their color.

Preferably, in step 2, the specific steps are: the odometer code disc is installed on the wheel shaft, and the odometer probe is installed on the frame; when the vehicle is running, the code disc rotates with the wheel, and each time the code disc rotates by one scale value , a pulse signal will be generated on the probe. By recording the number of pulse signals, the scale value of the code disc rotated can be obtained. The number of laps passed; then through the known circumference of the wheel, the linear distance traveled by the wheel can be calculated, and the distance between the vehicle and the stop line can also be calculated.

Preferably, in step 4, specifically:

Step 1: Determine whether the traffic light status is a red light. If it is a red light, start to decide when to stop. If no red light is detected, continue driving;

Step 2, determine whether there is a vehicle in front, if there is no vehicle in front, then go to step 3, stop according to the position of the parking line, if there is a vehicle in front, then go to step 6, stop according to the position and driving state of the vehicle in front;

Step 3: According to the distance returned by the last frame of the parking line result, determine the distance between the vehicle and the parking line at the intersection before entering the near-end blind spot of the camera, and trigger the distance calculation program;

Step 4, start the calculation of the driving distance of the vehicle. The calculation of the distance mainly relies on the pulse signal captured by the odometer. After the distance calculation program is triggered, the number of pulses is accumulated, and the formula for converting the accumulated number of pulses to the driving distance is:

Driving distance = pulse accumulation number ÷ odometer line number × π × wheel diameter;

Step 5, when the estimated driving distance reaches the stop line distance, send a stop instruction to stop the vehicle at the stop line,

Step 6: Determine whether the vehicle in front has stopped. If the vehicle in front does not stop, it means that the vehicle in front has not reached the intersection, then continue driving. If the vehicle in front stops, wait for the vehicle to arrive according to the distance from the vehicle in front detected by the millimeter-wave radar. After the parking distance is reached, a parking instruction is issued;

Step 7: Detect whether the status of the signal light is green, if it is green, continue driving, if there is a vehicle ahead, wait for the vehicle ahead to continue driving.

Compared with the prior art, the present invention has the following beneficial effects:

The invention uses the odometer to record the number of turns of the wheel, so as to calculate the distance traveled by the vehicle, and combined with the distance between the vehicle and the stop line obtained by image processing, it can effectively guide the unmanned vehicle to stop accurately at the stop line at the intersection. When there is a vehicle ahead, information such as the stop line cannot be detected, and the correct parking is realized according to the driving state of the preceding vehicle and the instructions of the traffic light, which solves the problem of only identifying the traffic light or only detecting the stop line in the traditional method without considering the camera. There are blind spots in the field of vision and there are other vehicles ahead, and it is impossible to realize the problem of precise parking of autonomous vehicles at intersections.

Description of drawings

Other features, objects and advantages of the present invention will become more apparent by reading the detailed description of non-limiting embodiments with reference to the following drawings:

1 is a schematic structural diagram of a preferred embodiment of the present invention;

FIG. 2 is a flowchart of a zebra detection part of a preferred embodiment of the present invention;

3 is a flowchart of a stop line detection part of a preferred embodiment of the present invention;

FIG. 4 is a flowchart of a traffic signal detection part of a preferred embodiment of the present invention;

FIG. 5 is a flow chart of realizing precise parking of an automatic driving vehicle at an intersection according to a preferred embodiment of the present invention.

Detailed ways

The present invention will be described in detail below with reference to specific embodiments. The following examples will help those skilled in the art to further understand the present invention, but do not limit the present invention in any form. It should be noted that, for those skilled in the art, several modifications and improvements can be made without departing from the concept of the present invention. These all belong to the protection scope of the present invention.

As shown in Figure 1, a device for precise parking at intersections of autonomous vehicles includes a camera 1, an odometer 2, a millimeter-wave radar 3, and an electronic control unit 4, wherein:

The camera 1 is used to collect images of zebra crossings, stop lines and traffic lights on the road, and transmit the collected image information to the electronic control unit 4;

The odometer 2 is used to obtain the angle turned by the vehicle through counting pulses, record the number of turns of the wheels, and transmit the obtained counting pulse signal to the electronic control unit 4;

The millimeter wave radar 3 is used to detect the position and speed information of the vehicle ahead, and transmit the detection result to the electronic control unit 4 .

The electronic control unit 4 receives and processes the road image collected by the camera 1, and identifies the zebra crossing, lane line and traffic signal information in the image; it is used to calculate the distance traveled by the vehicle through the counting pulse signal obtained by the odometer 3; Receive the position and speed information of the obstacle in front of the vehicle detected by the millimeter-wave radar 3; and send a control command to the underlying executive mechanism 5 of the autonomous driving vehicle according to the above information.

As a preferred embodiment, the camera 1 selects a wide-angle lens with a larger field of view, the installation position is the upper part of the front windshield of the vehicle, and the height to the ground is about 1.6 meters. The installation angle should ensure that the field of view of the camera 1 can cover the ground. Zebra crossings, stop line signs, and traffic lights above the road.

As a preferred embodiment, the odometer 2 and the gear code disc are installed on the rear axle, and the measuring head is installed on the frame; when the vehicle is running, the gear code disc rotates with the wheel, the measuring head remains stationary, and the gear wheel When the teeth and tooth slots pass the position of the probe, different intensities of induction signals are generated on the probe, and the induction signals are converted into count pulses, and the angle that the code disc or the vehicle has turned can be obtained.

As shown in FIG. 2-FIG. 5, the auxiliary method flow of using the above device includes the following steps:

Step 1: The camera 1 collects an image, and the electronic control unit 4 processes the image;

The camera 1 collects images of zebra crossings, stop lines and traffic lights on the road, and transmits the collected images to the electronic control unit 4;

The electronic control unit 4 performs inverse perspective transformation on the image, and uses the checkerboard to calibrate the transformation parameters to obtain the proportional relationship between the image pixels and the world coordinates, and the pixel position of the head of the vehicle in the image coordinates; the electronic control unit 4 processes the image, including zebra crossing detection. , stop line detection and traffic light detection;

Step 2: Calculate the driving distance of the vehicle

In order to park the autonomous vehicle at the parking line when the camera 1 has a blind spot and cannot use the parking line information to achieve longitudinal positioning, two distance information needs to be used. The distance between the line and the front of the car, and the second is the distance traveled by the vehicle from the moment when the stop line disappears; with these two distance information, from the moment when the stop line disappears, when the distance traveled by the vehicle reaches the distance between the stop line and the front of the car, the The parking control command can realize the accurate parking of the vehicle at the parking line; therefore, the odometer 2 is used to record the distance traveled by the vehicle;

Step 3: Use the millimeter wave radar 3 to detect whether there is a vehicle or other obstacle ahead.

The data returned by the millimeter-wave radar 3 includes the distance of the obstacle in front and the relative speed to the vehicle. By calculating the driving speed of the vehicle, the movement speed of the obstacle in front can be calculated.

Step 4. Intersection Behavior Decision

According to the perception information obtained by the camera 1, including the stop line and its distance, the status of traffic lights, and the position and speed of the preceding vehicle, the autonomous driving vehicle can make an auxiliary decision at the intersection;

First, considering the situation that there is no vehicle ahead, the auxiliary decision-making uses two distance information: one is the distance between the parking line and the front of the vehicle detected before the parking line enters the blind spot of the camera 1, and the other is the vehicle from the moment when the parking line disappears. Distance traveled; from the moment when the stop line disappears, when the distance traveled by the vehicle reaches the distance between the stop line and the front of the vehicle, a parking control command is issued, so that the vehicle can accurately stop at the stop line.

Next, considering the presence of other vehicles ahead, it is necessary to stop based on the status of the traffic lights and the distance from the vehicle in front. When a traffic light is detected, the vehicle is approaching the intersection. At this time, if the status of the signal light is green, keep a distance from the preceding vehicle. If the status of the signal light is red, the motion status of the vehicle in front is detected. If the vehicle in front has not stopped, it means that the vehicle has not reached the intersection and needs to continue to drive at a distance from the vehicle in front. If the vehicle in front has stopped, according to the distance of the vehicle in front detected by the millimeter-wave radar 3, the autonomous driving vehicle will be guided to the rear of the vehicle in front, and the vehicle will stop at the set parking distance.

As shown in FIG. 2 , it is a flow chart of the zebra detection part of a preferred embodiment. The electronic control unit 4 adopts the edge detection method combined with the local statistical features of the zebra crossing to achieve the purpose of accurately identifying the zebra crossing. Specifically, it is divided into several steps.

Step 1. Perform inverse perspective transformation on the grayscale image of the obtained road;

Step 2: Count the gradient histogram of the inverse perspective transformed image to obtain the maximum gradient direction of the image, and rotate the image according to the maximum gradient direction of the image. In the rotated image, the zebra stripes are distributed along the vertical direction of the image;

Step 3. Extract the edge of the rotated image. Since the main edge of the zebra crossing is along the vertical direction, only the edge in the vertical direction is extracted to reduce the influence of noise edges;

Step 4. For each line of the edge image, count the number of edge points in the line. For the zebra crossing area, the number of points obtained is much more than other non-zebra crossing areas, and accordingly, the width range of a zebra crossing can be obtained; For each column of , count the number of edge points in the column, and get a one-dimensional array. The values in the array represent the number of edge points in this column. If the column is the column where the edge of the zebra line is located, the number of points will be much larger than that of other non-zebra line edges. List;

Step 5. Perform a binarization operation on the values in the one-dimensional array, and the result is a one-dimensional binary sequence. There may be a column value of 1 for a zebra crossing, a value of 0 for other positions, and a value of 1 in the calculation sequence. The variance of the interval between two adjacent columns, since the edges of the zebra crossing are evenly distributed, if there is a zebra crossing, the calculated variance is very small. If there is interference such as arrow marks, the calculated variance will be much larger than the case of only the zebra crossing, so it can be A threshold is used to discriminate.

As shown in FIG. 3 , it is a flow chart of the stop line detection part of a preferred embodiment. Generally speaking, the image features of the zebra crossing are stronger than the stop line, and the obtained detection result is more stable and reliable. Therefore, the center detects the stop line on the basis of detecting the zebra crossing. It is divided into the following processes:

Step 1. Determine whether the zebra crossing has been detected, if so, start to detect the stop line;

Step 2. Extract the horizontal edge of the image:

The stop line is different from the zebra crossing, its edge direction is the horizontal direction, and the horizontal direction edge is extracted from the rotated image in the process of detecting the zebra crossing;

Step 3, select the region of interest for detection:

In the zebra crossing detection process, the longitudinal distribution orientation of the zebra crossing is obtained, and in the actual scene, the parking line must be near the lower edge of the zebra crossing, so the area from the lower edge of the zebra crossing to the bottom of the image is selected as the area of interest;

Step 4. Use Hough line transformation to detect straight lines in the region of interest, and use length constraints, duty cycle constraints and parallel constraints to detect the stop line;

Step 5. After the parking line is detected, the distance between the front of the car and the parking line is obtained according to the pixel coordinates of the front of the car in the image and according to the formula for calculating the distance from the point to the straight line.

As shown in FIG. 4 , it is a flow chart of the traffic signal detection part of a preferred embodiment. The behavior of vehicles at intersections depends to a large extent on the status of traffic lights, so it is critical to accurately detect traffic lights and identify their indicated states for decision-making on vehicle behavior at intersections. The electronic control unit 4 adopts the image segmentation method based on the color space and the area-perimeter ratio to detect and identify the traffic lights, which is divided into the following processes:

Step 1. Convert the color space of the image captured by the camera 1 from the RGB color space to the LAB color space. The A channel of the LAB color space represents the color change from green to red, and it is easy to use a threshold to segment the red and green areas from the image;

Step 2. Screening of candidate regions:

Since there are many other objects in the road whose colors satisfy the characteristics of red and green, some rules must be used to filter the areas that may be signal lights; the method adopted by the electronic control unit 4 is to use the area and shape to obtain red and green using the seed growth method The number of pixels contained in the seed growth area is the area of the area; the shape of the signal light is circular, and the circularity value of the candidate area is used to determine whether the area is circular. The calculation method of circularity is the area of the area. The square of the perimeter is divided by the area of the area. For a standard circle, the value of the roundness should be 4π; combine these two indicators to select a candidate area that meets the requirements;

Step 3. Background frame verification:

After the previous screening steps, there may still be billboards, car tail lights, etc. in the candidate area; another important feature of the signal light is that it has a black background frame. By selecting a rectangular area of a certain size in the candidate area, it is detected whether there is a satisfactory The pair of straight lines required for parallelism and width to determine whether there is a background box, thereby verifying whether the candidate area is a signal light.

As shown in FIG. 5 , it is a flowchart of a preferred embodiment for realizing precise parking of an automatic driving vehicle at an intersection, that is, an intersection behavior decision-making part. The detection results obtained by the image processing of the electronic control unit 4 include traffic lights, zebra crossings and stop lines, and the millimeter wave radar 3 detects the position and speed of the vehicle ahead. This information is used to determine whether the autonomous vehicle is approaching the intersection and whether it should stop, and if it is necessary to stop, start the parking position guidance. The specific process is divided into the following steps:

Step 1: Determine whether the traffic light status is a red light. If it is a red light, start to decide when to stop. If no red light is detected, continue driving.

Step 2, determine whether there is a vehicle ahead, if there is no vehicle ahead, go to step 3, and stop according to the position of the parking line. If there is a preceding vehicle, go to step 6, and stop according to the position and driving state of the preceding vehicle.

Step 3: According to the distance returned in the last frame of the parking line result, determine the distance between the vehicle and the parking line at the intersection before entering the near-end blind spot of the camera, and trigger the distance calculation program.

Step 4, start the calculation of the driving distance of the vehicle. The calculation of the distance mainly relies on the pulse signal captured by the odometer 2. When the distance calculation program is triggered, the number of pulses is accumulated, and the formula for converting the accumulated number of pulses to the driving distance is:

Driving distance = pulse accumulation number ÷ odometer 2 lines × π × wheel diameter;

Step 5: When the estimated travel distance reaches the stop line distance, a stop instruction is sent to stop the vehicle at the stop line.

Step 6: Determine whether the vehicle ahead is stopped. If the vehicle in front does not stop, it means that the vehicle in front has not reached the intersection, then continue driving. If the vehicle in front stops, according to the distance from the vehicle in front detected by the millimeter wave radar 3, after the vehicle reaches the required parking distance, a parking instruction is issued.

Step 7: Detect whether the status of the signal light is green, and if it is green, continue driving. If there is a vehicle ahead, wait for the vehicle ahead to continue driving.

The invention can effectively guide the unmanned vehicle to stop accurately at the intersection stop line, and solves the problem that the traditional method only recognizes the traffic signal or only detects the stop line, and does not consider the blind spot in the field of view of the camera 1 and the existence of other vehicles ahead. Realize the problem of accurate parking of unmanned vehicles at intersections.

Specific embodiments of the present invention have been described above. It should be understood that the present invention is not limited to the above-mentioned specific embodiments, and those skilled in the art can make various variations or modifications within the scope of the claims, which do not affect the essential content of the present invention.

Claims (7)

1. a method for precise parking at intersection for automatic driving, it is characterized in that, adopts the precise parking device at intersection for automatic driving, and described precise parking device at intersection comprises: image acquisition device, odometer, millimeter-wave radar and electronic control unit, where: The image acquisition device is used to collect images of zebra crossings, stop lines and traffic lights on the road, and transmit the collected image information to the electronic control unit; The odometer is used to obtain the angle turned by the vehicle through counting pulses, record the number of turns of the wheels, and transmit the obtained counting pulse signal to the electronic control unit; The millimeter wave radar is used to detect the position and speed information of dynamic and static obstacles in front of the vehicle; The electronic control unit is used to receive and process the image information collected by the image acquisition device, and identify the zebra crossing, lane line and traffic signal information in the image; it is used to receive the position and speed of the obstacle in front of the vehicle detected by the millimeter wave radar information; used to calculate the distance traveled by the vehicle through the counting pulse signal obtained by the odometer; and send control instructions to the underlying actuator of the autonomous vehicle according to the above information; The method for precise parking at intersections includes the following steps: Step 1, the image acquisition device collects the image, and the electronic control unit processes the image; The image acquisition device collects images of zebra crossings, stop lines and traffic lights on the road, and transmits the collected images to the electronic control unit; The electronic control unit performs inverse perspective transformation on the image, and uses the checkerboard to calibrate the transformation parameters to obtain the proportional relationship between the image pixels and the world coordinates, as well as the pixel position of the head of the vehicle in the image coordinates; the electronic control unit processes the image, including zebra crossing detection, stop Line detection and traffic light detection; Step 2. Use the odometer to record the distance traveled by the vehicle; Step 3. Use millimeter wave radar to detect the position and speed information of obstacles in front of the vehicle; Step 4. Intersection behavior decision; The electronic control unit assists the autonomous vehicle at the intersection according to the perception information obtained by the image acquisition device, including the stop line and its distance, the status of the traffic light, and the position and speed information of the vehicle ahead detected by the millimeter-wave radar decision making; When there is no other vehicle in front of the vehicle, the parking line is used to make an auxiliary decision for the autonomous vehicle at the intersection. The auxiliary decision uses two distance information: one is the parking line detected before the parking line enters the blind spot of the image acquisition device. The distance between the line and the front of the car, and the second is the distance traveled by the vehicle since the time when the stop line disappeared; From the moment when the stop line disappears, when the distance traveled by the vehicle reaches the distance between the stop line and the front of the vehicle, a parking control command is issued, so that the vehicle can accurately stop at the stop line; When there are other vehicles in front of the vehicle, the electronic control unit makes an auxiliary decision for the autonomous vehicle at the intersection according to the position and speed information of the preceding vehicle detected by the millimeter-wave radar: when the image acquisition device detects a signal light, And the status of the signal light is red, the motion state of the vehicle in front is detected. If the vehicle in front has not stopped, it means that the vehicle has not reached the intersection and needs to continue to drive at a distance from the vehicle in front; if the vehicle in front has stopped, it will be detected by millimeter-wave radar. When the distance to the vehicle in front is reached, the self-driving vehicle will be guided to the rear of the vehicle in front, and the vehicle will stop according to the set parking distance. 2 . The method for precise parking at an intersection for automatic driving according to claim 1 , wherein the image acquisition device is installed on the upper part of the front windshield of the vehicle, and the installation angle ensures that its field of view can cover the zebra crossing on the ground. 3 . , stop line signs, and traffic lights above the road. 3. a kind of intersection accurate parking method for automatic driving according to claim 1, is characterized in that, in step 1, described zebra crossing detection method, adopts edge detection method and combines the local statistical feature of zebra crossing, to achieve accurate The purpose of identifying zebra crossings. 4 . The method for precise parking at an intersection for automatic driving according to claim 1 , wherein the electronic control unit converts the image collected by the image acquisition device into a grayscale image, and then according to the main gradient of the image Rotate the image in the direction, extract the edge of the rotated image in the vertical direction, count the number of edge points in each row of the image, estimate the range of the zebra crossing, and then count the number of edge points in each column within the range; the column where the edge of the zebra stripe is located The number of points is much larger than other columns, calculate the intervals between these columns with more points, and find the variance of these intervals. If the stripes are evenly distributed, the variance value will be very small, based on which to determine whether there is a zebra crossing. 5. A kind of accurate parking method for automatic driving according to claim 1, it is characterized in that, in step 1, described stop line detection method, combines the detection result of zebra crossing to identify stop line, after detecting zebra crossing, Take the lower edge of the zebra line to the last line of the image as the region of interest, use the Hough line transform to extract all the lines in the region, and then extract the line parameters of the stop line in the image according to the length, duty cycle, and direction constraints of the stop line , thereby calculating the distance from the vehicle to the stop line. 6. a kind of intersection accurate parking method for automatic driving according to claim 1, is characterized in that, described step 2, is specifically: Install the odometer code disc on the wheel shaft, and the odometer probe is installed on the frame; when the vehicle is running, the code disc rotates with the wheel, and every time the code disc rotates a scale value, a pulse signal will be generated on the probe, By recording the number of pulse signals, the scale value that the code disc has rotated can be known. Divide the number of pulses by the range of the code disc rotated one circle to obtain the code disc, that is, the number of turns the wheel has rotated; The circumference of the wheel can be used to calculate the linear distance traveled by the wheel, which can also calculate the distance between the vehicle and the stop line. 7. The method for precise parking at intersections for automatic driving according to any one of claims 1-6, wherein the step 4 is specifically: Step 1: Determine whether the traffic light status is a red light. If it is a red light, start to decide when to stop. If no red light is detected, continue driving; Step 2, determine whether there is a vehicle in front, if there is no vehicle in front, then go to step 3, stop according to the position of the parking line, if there is a vehicle in front, then go to step 6, stop according to the position and driving state of the vehicle in front; Step 3: According to the distance returned by the last frame of the parking line result, determine the distance between the vehicle and the parking line at the intersection before entering the near-end blind spot of the camera, and trigger the distance calculation program; Step 4, start the calculation of the vehicle's driving distance. The calculation of the distance relies on the pulse signal captured by the odometer. From the trigger of the distance calculation program, the number of pulses is accumulated, and the formula for converting the accumulated number of pulses to the driving distance is: Driving distance = pulse accumulation number ÷ odometer line number × π × wheel diameter; Step 5, when the estimated travel distance reaches the stop line distance, send a stop instruction to stop the vehicle at the stop line; Step 6: Determine whether the vehicle in front has stopped. If the vehicle in front does not stop, it means that the vehicle in front has not reached the intersection, then continue driving. If the vehicle in front stops, wait for the vehicle to arrive according to the distance from the vehicle in front detected by the millimeter-wave radar. After the parking distance is reached, a parking instruction is issued; Step 7: Detect whether the status of the signal light is green, if it is green, continue driving, if there is a vehicle ahead, wait for the vehicle ahead to continue driving.

Images