CN112070942B - An intelligent management system for park gates during the epidemic - Google Patents

Abstract

The invention belongs to the field of epidemic situation monitoring equipment systems, and particularly relates to an intelligent management system for a park gate during an epidemic situation. The device comprises an identification detection component, a motion and control component and an information interconnection component; the identification detection component comprises an image database module, an acquisition module, a detection module and an identification module; the motion and control assembly comprises a mechanical arm module and a steering engine driving module; the intelligent monitoring system and the intelligent monitoring method can realize all-weather non-contact intelligent identification and detection by carrying out regional arrangement in a monitored park. The system is mainly suitable for body temperature detection and face recognition of people in a driving car, realizes real-time monitoring of population mobility, reduces the working strength of detection personnel, reduces the protection cost of the detection personnel, is suitable for being widely popularized and used in colleges and universities and living communities, and has high market value. The system needs fewer workers, has low requirements on the working strength of station workers and medical protection, has short detection time, and can conveniently record passengers.

Description

An intelligent management system for park gates during the epidemic

technical field

The invention belongs to the field of epidemic monitoring equipment systems, and in particular relates to an intelligent management system for park gates during epidemics.

Background technique

Today, although the domestic epidemic has been effectively controlled, and various regions and departments have begun to actively resume work, production and school with the support of policies, the epidemic prevention situation is still very serious. With the arrival of summer, the temperature gradually rises, and medical staff wearing heavy protective clothing often suffer from heat stroke and faint due to insufficient oxygen supply.

In response to the epidemic, robots have been widely used in various fields, such as distribution robots, warm robots and medical robots. The basic principle is to use robots to replace simple human operations to reduce the risk of infectious contact and instability in the face of high-intensity workloads. However, in dense places such as highways and closed parks or in transportation hub areas, the detection of people in vehicles is usually dominated by human detection, which will bring about large human input, slow detection efficiency, slow information update, and infection risks. High-level problems and many other adverse effects, due to the vehicle detection of the epidemic, resulting in long queues on highways, important streets, and office parks.

SUMMARY OF THE INVENTION

The purpose of the invention is to realize the all-weather non-contact intelligent identification and detection through regionalized arrangement in the park to be monitored, and is mainly used for the body temperature detection and face recognition of the people in the driving vehicle, so as to realize the real-time monitoring of population flow.

To achieve the above purpose, the present invention adopts the following technical solutions.

An intelligent management system for park gates during the epidemic, including identification and detection components, motion and control components, and information interconnection components;

The identification and detection component includes an image database module, a collection module, a detection module, and an identification module;

The database module establishes an inherent database and a temporary database for uploading for the inherent and temporary crowds. The database module is established in the following ways:

01 On the basis of the existing face data set, the deep learning model in the open source library dlib is used to train the residual network combined with the face features, and the network model is obtained;

02 There are two parts: the inherent personnel data table and the temporary personnel data table. The database module collects the information of all personnel entering the park based on the identification module. When all data detection is completed, the data is packaged and sent to the server according to a fixed data format and stored. Clear temporary personnel information regularly; the data table at least includes equipment operation status table, non-inherent personnel information table and inherent personnel information table;

The acquisition module includes a camera for image acquisition, preferably an infrared night vision camera. The acquisition module transmits the collected image data to the Raspberry Pi 3b+, and supports remote login operation and offline operation. Through the python language The program processes the image to identify the temperature measurement object;

The identification module is used to complete the following steps:

S1. Feature extraction: Calculate and count the gradient direction histogram of the local area of the image to form a feature; the gradient direction is obtained in the following way: convert the detected image into black and white, and for each pixel in the image, compare it with other surrounding Pixels are compared, find and compare the depth of the current pixel to the pixel directly surrounding it, use the arrows to point in the direction of the darkening of the image, repeat this process until every pixel is replaced by an arrow; these arrows are the gradients, they show out the flow from light to dark on the image;

S2, distance measurement, obtain the relative position through the face recognition API, and return it to the communication control board to control the operation of the slide rail, and guide the next temperature detection:

Specifically, while performing face recognition, the position information of the face after the first recognition is returned for the movement of the motion module, and the camera is calibrated at the same time, based on the offset of the face in the image relative to the center of the image The offset of the actual face relative to the detection terminal is linear in the fixed xoz plane. According to the formula, the fixed proportional coefficient a=θ/p _x is obtained. After the analysis of the face box in the API, the slide rail needs are obtained. The moving distance Δx to achieve the same position of the detection terminal and the detection personnel in the vehicle;

where p _x is the horizontal acquisition pixel of the camera, x ₀ , x ₁ , x ₂ and x ₃ are the position coordinates of the face, and x is the offset of the face in the image relative to the center of the image;

The motion and control components include the robotic arm module and the steering gear drive module;

The robotic arm module includes a five-degree-of-freedom robotic arm, which is equipped with a detection terminal to realize the necessary detection of people in the vehicle; the robotic arm, as an important motion module, helps the detection personnel to obtain the necessary information for entering the park; the robotic arm is forwarded by the communication control board. , the signal is converted by the steering gear drive board and driven by each steering gear. When the detection terminal judges the position between the vehicle body and the vehicle body through the distance sensor, the drive board can calculate the angle that each motor needs to rotate through inverse kinematics, and finally achieve a convenient car. The location of the personnel detection location;

The steering gear drive module uses the inverse solution of the kinematic equation to realize the motion control of the manipulator. The basic steps include:

Step 1. Establish the kinematic relationship based on joint space and Cartesian space;

变换矩阵中；The homogeneous transformation matrix is used to establish the pose expression of the end of the manipulator, describing the forward kinematics solution from joint space to Cartesian space or the inverse kinematics solution from Cartesian space to joint space, specifically; using the DH parameter The mathematical model and coordinate system determination system of the manipulator expressing the position and angle relationship between the two pairs of joint links. According to the structural parameters of the manipulator and the motion range of the steering gear, the parameters of the joints and links are obtained, and these parameters are substituted into

in the transformation matrix;

According to the formula:

Calculate each joint in turn, and finally get the positive kinematics formula of the manipulator, and substitute the DH parameters to get the rotation matrix of each joint:

After obtaining the rotation matrix of each joint, the coordinates of the end can be obtained according to the following formula:

Step 2. Inverse solution of kinematic equation

求解各转动关节的角度θ_i；机械臂的端点P的坐标(x,y)由三个部分(x₁+x₂+x₃,y₁+y₂+y₃)组成；其中上式的θ₁、θ₂和θ₃是求解的舵机的角度，a是爪子与水平面的夹角，且a＝θ₁+θ₂+θ₃，且：The coordinates returned by face recognition are used to guide the movement of the robotic arm so that the detection terminal can be positioned appropriately with the detection personnel; and based on the homogeneous transformation matrix obtained in step 1

Solve the angle θ _i of each rotating joint; the coordinates (x, y) of the end point P of the mechanical arm consist of three parts (x ₁ +x ₂ +x ₃ , y ₁ +y ₂ +y ₃ ); θ ₁ , θ ₂ and θ ₃ are the angles of the solved steering gear, a is the angle between the claw and the horizontal plane, and a=θ ₁ +θ ₂ +θ ₃ , and:

x=l ₁ cosθ ₁ +l ₂ cos(θ ₁ +θ ₂ )+l ₃ cos(θ ₁ +θ ₂ +θ ₃ )

y=l ₁ sinθ ₁ +l ₂ sin(θ ₁ +θ ₂ )+l ₃ sin(θ ₁ +θ ₂ +θ ₃ )

m=l ₃ cosa-x n=l ₃ sina-y

x and y are given by the user; l ₁ , l ₂ and l ₃ are the inherent properties of the mechanical structure of the manipulator. Substitute m and n into the existing equations, and then simplify to get: l ₂ =(l ₁ cosθ ₁ + m) ² +(l ₁ sinθ ₁ +n) ²

It can be obtained by calculation:

The above formula is the form of the root-finding formula of the quadratic equation in one variable, where a=m ² +n ²

According to this, we can obtain the angles of θ ₁ and θ ₂ ; solve the angles of the three steering gears based on the above steps, and control the steering gears according to the angles to realize the control of the coordinate position.

The further improvement of the above-mentioned intelligent management system for park gates during the epidemic also includes information interconnection components; including a communication control board based on STM32F103C8T6 as the core, KEIL programming to read sensor information and upload 485 bus, stepper motor control, Expand GPIO peripherals; include two TTL serial ports for connecting sensors; one 485 serial port interface, connecting 485 control bus; 5 GPIOs, GPIO includes general timer TIMER3 channels 1 and 2; 74LS244 is used for 3.3V-5V boost , for four-wire stepper motor control.

To further improve the above-mentioned intelligent management system for gates in the park during the epidemic, the information interconnection component uses KEIL to program and control the STM32F103, UART2 communicates with the upper computer, receives interrupts, verifies and parses the issued instructions, and controls the UART1 and 3 pairs. Corresponding peripheral communication control. UART1 and its interrupt communicate with the temperature sensor, and UART3 sends coordinate information to the robotic arm according to the instructions of the host computer. The external interrupt is triggered by the falling edge of the PB0 pin to start the distance sensor.

A further improvement of the above-mentioned intelligent management system for gates in the park during the epidemic, the identification module includes an identification module and a license plate identification module; the identification module adopts the TY-801T embedded module, and uses RFID technology to identify the ID card information. Read and transmit the read information to the database. The identity recognition module needs to record the identities of people entering and leaving the park and not in the warehouse, and upload the identification information to the database in real time; the license plate recognition module uses Http Post , so that the camera is connected to the Http server through ssl, when there is a vehicle, the recognition system will automatically recognize and display the vehicle information.

Its beneficial effects are:

This application can realize all-weather non-contact intelligent identification and detection by regionalized arrangement in the park to be monitored. The system is mainly suitable for body temperature detection and face recognition of people in driving vehicles, and realizes real-time monitoring of population flow. In addition, the functions of license plate, face and identity recognition are also added to conduct all-weather non-contact detection of all personnel entering and leaving the park to ensure the safety of staff and the interior of the park. It is suitable for large-scale promotion and use in colleges and universities and living quarters, and has high market value. This application requires less staff, less work intensity for site staff, low requirements for medical protection, less testing time, and can easily record passengers.

Description of drawings

Fig. 1 is the structural block diagram of the identity recognition module;

Figure 2 is a schematic diagram of the communication flow between the radio frequency card and the reader;

3 is a schematic diagram of camera calibration;

Figure 4 is a schematic diagram of the accuracy of the measured object covering the field of view;

FIG. 5 is a schematic diagram of the experimental data of the temperature compensation curve of the thermometer and the infrared temperature sensor.

Detailed ways

The present invention will be described in detail below with reference to specific embodiments.

An intelligent management system for park gates in the present application mainly includes identification and detection components including an image database module, a collection module, an identification module, and a detection module;

The database module establishes an inherent database and a temporary database for uploading for the inherent crowd and the temporary crowd. The database is established in the following manner: on the basis of the existing face data set, the deep learning model in the open source library dlib is used to train the residual network combined with the face features to obtain a network model; there are inherent personnel data tables and There are two parts of the temporary personnel data table. The database module collects the information of all personnel entering the park based on the identification module. When all data detection is completed, the data is packaged and sent to the server according to a fixed data format and stored, and the temporary personnel information is cleared regularly; The data sheet consists of at least three parts: equipment operating status table, non-inherent personnel information table and inherent personnel information table, see Tables 1, 2 and 3.

Table 1 Equipment operating status table

name type length Equipment total system number varchar 20 Overall system location of equipment varchar 20 Raspberry Pi CPU temperature varchar 20 Raspberry Pi CPU usage varchar 20 Raspberry Pi RAM_total varchar 20 Raspberry Pi RAM_used varchar 20 Raspberry Pi RAM_free varchar 20 Hard disk capacity total varchar 20 Hard disk capacity used varchar 20 free hard disk capacity varchar 20 temperature sensor status varchar 20 camera sensor status varchar 20 ID sensor status varchar 20 Robot arm status varchar 20 Rail status varchar 20

Table 2 Non-inherent personnel information table

name type length Temporary second-generation certificate id varchar 20 body temperature varchar 10 facial feature point data varchar 1600 number plate varchar 20 time varchar 100 gate location varchar 20

Table 3 Inherent Personnel Information Table

name type length Second generation certificate id varchar 20 Name varchar 20 address varchar 200 identification number varchar 20 body temperature varchar 10 facial feature point data varchar 1600 number plate varchar 20 time varchar 100 gate location varchar 20

The acquisition module includes a camera for image acquisition, preferably an infrared night vision camera. The acquisition module transmits the collected image data to the Raspberry Pi 3b+, and supports remote login operation and offline operation. Through the python language The program processes the image to identify the temperature measurement object,

The recognition module is used to complete: feature extraction: calculate and count the gradient direction histogram of the local area of the image to form a feature; the gradient direction is obtained in the following way: convert the detected image into black and white, and for each pixel in the picture, Compare it to other pixels around it, find and compare the depth of the current pixel to the pixel directly surrounding it, use arrows to point in the direction the image is darkened, repeat this process until every pixel is replaced by an arrow; these arrows That is, gradients, they show the flow process from bright to dark on the image; distance measurement, this application obtains the relative position through the API of face recognition, and returns it to the communication control board to control the operation of the slide rail, effectively guiding the next step The temperature detection is convenient for personnel to carry out temperature detection.

Specifically, while performing face recognition, the position information of the face after the first recognition is returned for the movement of the motion module, and the camera is to be calibrated at the same time. The offset and the actual offset of the face relative to the detection terminal are linear in the fixed xoz plane. According to the formula, the fixed proportional coefficient a=θ/p _x is obtained, which is obtained through the analysis of the face box in the API. The distance Δx that the slide rail needs to move to achieve the same position of the detection terminal and the detection personnel in the vehicle; and

Where p _x is the horizontal acquisition pixel of the camera, x ₀ , x ₁ , x ₂ and x ₃ are the coordinates of the face position, and x is the offset of the face in the image relative to the center of the picture.

The distance measurement obtains the relative position through the face recognition API, and returns it to the communication control board to control the operation of the slide rail, which effectively guides the next temperature detection and facilitates the personnel to perform body temperature detection.

At the same time of face recognition, the position information of the face after the first recognition is returned for the movement of the motion module. The camera needs to be calibrated here. The method is as follows: Assume that the offset of the face in the image relative to the center of the picture and the actual offset of the face relative to the detection terminal are linear in the fixed xoz plane. And a fixed proportional coefficient a=θ/p _x is obtained according to the formula. Finally, through the analysis of the face box in the API, the distance Δx that the slide rail needs to move is obtained to achieve the same position of the detection terminal and the detection personnel in the vehicle. The principle is shown in Figure 3;

Where p _x is the horizontal acquisition pixel of the camera, x ₀ , x ₁ , x ₂ and x ₃ are the coordinates of the face position, and x is the offset of the face in the image relative to the center of the picture.

The identity recognition module uses the TY-801T embedded module to read the ID card information through RFID, and transmit the read information to the database. The identity recognition module needs to record the identities of people entering and leaving the park and not in the warehouse, and upload the identification information to the database in real time. The overall structure diagram is shown in Figure 1; when the ID card is recognized, the reader sends a set of For electromagnetic waves with a fixed frequency, there is an LC series resonant circuit in the card, and its frequency is the same as the frequency emitted by the reader. Under the excitation of electromagnetic waves, the LC resonant circuit resonates, so that there is charge in the capacitor. One end is connected with a unidirectional electronic pump, which sends the charge in the capacitor to another capacitor for storage. When the accumulated charge reaches 2V, this capacitor can be used as a power supply to provide working voltage for other circuits, and transmit the data in the card. To send or receive data from the reader, see Figure 2.

The license plate recognition module uses Http Post to connect the camera with the Http server through ssl. Triggered when there is a vehicle, the identification system will automatically identify and display vehicle information. After cooperating with other modules to detect that there is no abnormal situation, the camera will raise the barrier through IO output, so as to control the entry and exit of vehicles.

The recognition module has two different data verification methods for the inherent population and the non-inherent population, and is used to complete deep learning, image acquisition, face image preprocessing, face image feature extraction, and matching and recognition; face recognition refers to, Obtain the face information obtained by the image acquisition module, generate a multi-dimensional vector of the face through ResNet, and calculate the Euclidean distance from the vector in the database after the generated vector to determine the similarity of the face. In the second stage, through the collection of video streams, face detection is performed on each picture first, and then face recognition is performed on it, and the recognized information is marked on the picture;

The body temperature detection module adopts a non-contact infrared temperature sensor, the model is MLX90614ESF-BCC. The sensor can detect the surface temperature of the object according to the infrared radiation energy size and wavelength distribution of the measured object. For non-contact infrared temperature measurement modules, a very important concept is "field of view (FOV)." The field of view is determined by the thermopile receiving 50% of the radiation signal and is related to the main axis of the sensor. The measured temperature is a weighted average of the temperature of the measured object within the field of view, so the accuracy is highest when the measured object completely covers the FOV field of view, see Figure 4.

In order to obtain an accurate detection of body temperature, a high-precision thermometer and an infrared temperature sensor are used to measure separately, and a temperature compensation curve is drawn in combination with the distance sensor. The detected temperature value obtained through the compensation curve is closer to the actual temperature. Test the MLX90614ESF-BCC, use the MLX90614ESF-BCC to detect objects with different distances and a constant body temperature of 35.5 ℃, and draw the experimental data as Figure 5

The module samples the temperature information through the single-chip microcomputer, and then transmits the data through the serial port conversion IC through TTL level communication. The technical parameters are as follows in Table 4:

Table 4 Data Parameters

name parameter target temperature range -70°～330° Sensor environment range -40°～125° measurement accuracy 0.5°C (0-50°C) Resolution 0.02℃ response frequency 2Hz Operating Voltage 3～5V Working current 15mA size 21.5mm×23mm

The motion and control components include a robotic arm module, a steering gear drive module, and a communication control module;

The robotic arm module includes a five-degree-of-freedom robotic arm equipped with a detection terminal to realize the necessary detection of people in the vehicle. As an important motion module, the robotic arm helps inspectors take necessary information to enter the park. The mechanical arm is forwarded by the communication control board, and the signal is converted by the steering gear drive board, which is driven by each steering gear. When the detection terminal judges the position between the vehicle body and the vehicle body through the distance sensor, the driving board can calculate each motor through inverse kinematics. The angle that needs to be rotated will eventually reach a position that is convenient for people in the car to detect.

The servo drive module uses the inverse solution of the kinematic equation to realize the motion control of the manipulator. The basic steps include:

Step 1. Establish the kinematic relationship based on joint space and Cartesian space

变换矩阵中。The homogeneous transformation matrix is used to establish the pose expression of the end of the manipulator, describing the forward kinematics solution from joint space to Cartesian space or the inverse kinematics solution from Cartesian space to joint space, specifically; using the DH parameter The mathematical model and coordinate system determination system of the manipulator expressing the position and angle relationship between the two pairs of joint links. According to the structural parameters of the manipulator and the motion range of the steering gear, the parameters of the joints and links are obtained, and these parameters are substituted into

in the transformation matrix.

According to the formula:

Calculate each joint in turn, and finally get the positive kinematics formula of the manipulator, and substitute the DH parameters to get the rotation matrix of each joint:

After obtaining the rotation matrix of each joint, the coordinates of the end can be obtained according to the following formula:

Step 2. Inverse solution of kinematic equation

The coordinates returned by face recognition are used to guide the movement of the robotic arm, so that the detection terminal can be properly positioned with the detection personnel;

求解各转动关节的角度θ_i；机械臂的端点P的坐标(x,y)由三个部分(x₁+x₂+x₃,y₁+y₂+y₃)组成。其中上式的θ₁、θ₂和θ₃是求解的舵机的角度，a是爪子与水平面的夹角，且a＝θ₁+θ₂+θ₃，且：and based on the homogeneous transformation matrix obtained in step one

Solve the angle θ _i of each rotating joint; the coordinate (x, y) of the end point P of the manipulator consists of three parts (x ₁ +x ₂ +x ₃ , y ₁ +y ₂ +y ₃ ). where θ ₁ , θ ₂ and θ ₃ in the above formula are the angles of the steering gear to be solved, a is the angle between the claw and the horizontal plane, and a=θ ₁ +θ ₂ +θ ₃ , and:

x=l ₁ cosθ ₁ +l ₂ cos(θ ₁ +θ ₂ )+l ₃ cos(θ ₁ +θ ₂ +θ ₃ )

y=l ₁ sinθ ₁ +l ₂ sin(θ ₁ +θ ₂ )+l ₃ sin(θ ₁ +θ ₂ +θ ₃ )

m=l ₃ cosa-x n=l ₃ sina-y

x and y are given by the user. l ₁ , l ₂ and l ₃ are the inherent properties of the mechanical structure of the manipulator. Substitute m and n into the existing equations, and then simplify to get: l ₂ =(l ₁ cosθ ₁ +m) ² +(l ₁ sinθ ₁ +n) ²

It can be obtained by calculation:

The above formula is the form of the root-finding formula of the quadratic equation in one variable, where a=m ² +n ²

According to this, we can obtain the angles of θ ₁ and θ ₂ ; solve the angles of the three steering gears based on the above steps, and control the steering gears according to the angles to realize the control of the coordinate position.

It also includes information interconnection components, including a design communication control board based on STM32F103C8T6, KEIL programming to read sensor information and upload 485 bus, stepper motor control, and expand GPIO peripherals; including two TTL serial ports for connecting sensors; one 485 serial port interface, connected to 485 control bus; 5-way GPIO, GPIO includes general timer TIMER3 channel 1, 2; 74LS244 is used for 3.3V-5V boost, and four-wire stepper motor control; UART2 communicates with the host computer, receiving interruption Verifies and parses the issued instructions, and controls UART1 and 3 to control the communication of corresponding peripherals. UART1 and its interrupt communicate with the temperature sensor, and UART3 sends coordinate information to the robotic arm according to the instructions of the host computer. The external interrupt is triggered by the falling edge of the PB0 pin to start the distance sensor.

Its specific implementation is:

When the vehicle is approaching, the ground-sensing coil triggers the high-definition camera to first identify the model and license plate, and at the same time, the remote control voice in the control room prompts the occupants to lower the windows. According to the camera feedback model, first move the robotic arm to the corresponding height and position the slide rail. The Raspberry Pi carried by the terminal judges whether there is a passenger in the seat through face capture, and if so, calculates the position deviation between the terminal and the person to be tested, and sends the coordinates to the communication control board connected to the terminal, so as to precisely control the position of the slide rail, and calculate through inverse kinematics. Adjust the posture of the robotic arm, and compare the facial features of the tested person with the database synchronously in the process to determine whether additional registration information is required. After the detection terminal recognizes and collects the passenger information to be tested according to the process, it will be aggregated through the communication control panel, and the information will be sent to the upper computer in the control room, which will then upload it to the cloud. Each node can call the information in the database at any time to realize the interconnection of park management.

Compare the three detection methods of hand-held thermometer, site-based thermal imaging detector and this work. Compare the functions of 3 different ways. According to the general distribution of passengers in the car, it is time-consuming to compare the detection methods. Through comparative analysis, it is found that this work requires less staff, less work intensity for site staff, low medical protection requirements, and less detection time, which can easily record the passengers. The specific parameters are shown in Table 5. and Table 6.

Table 5 Function comparison of each detection method

Table 6 Comparison of the average time consumption of each detection method

In the process of resuming work, production and school, it is particularly important to control the second outbreak of the new crown epidemic. Especially in crowded places such as colleges and living quarters, prevention and timely detection of people infected with the new coronavirus plays an extremely important role in epidemic prevention and control. Looking at the reasons, it is the lack of a strong identification, detection and monitoring system. The market is a high-risk intensive field with a relatively large population density. Once a case occurs, a large outbreak of clusters is likely to occur. As the temperature rises, the temperature measurement mode of the whole body protection handheld thermometer also seems out of date. The development of artificial intelligence provides a prerequisite for changing this temperature measurement method and establishing an intelligent detection and monitoring system.

In order to solve the problems of non-contact temperature measurement and intelligent prevention and control under the situation of epidemic prevention and control, the functions of license plate, face and identity recognition are also added, and all personnel entering and leaving the park are subject to all-weather non-contact detection to ensure that staff and the inside of the park security. The system can effectively ensure the safety of the inspectors, reduce the work intensity of the inspectors, and at the same time reduce the protection costs of the inspectors.

Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, but not to limit the protection scope of the present invention. Persons should understand that the technical solutions of the present invention may be modified or equivalently replaced without departing from the spirit and scope of the technical solutions of the present invention.

Claims (4)

1. an intelligent management system for park gates during an epidemic, characterized in that it comprises identification and detection components, motion and control components, and information interconnection components; The identification and detection component includes an image database module, a collection module, a detection module, and an identification module; The database module establishes an inherent database and a temporary database for uploading for the inherent and temporary crowds. The database module is established in the following ways: S01. On the basis of the existing face data set, use the deep learning model in the open source library dlib to combine the face features to train the residual network to obtain a network model; S02. There are two parts: the inherent personnel data table and the temporary personnel data table. The database module collects the information of all personnel entering the park based on the identification module. When all data detection is completed, the data is packaged and sent to the server according to a fixed data format and stored. And periodically clear temporary personnel information; the data table at least includes equipment operation status table, non-inherent personnel information table and inherent personnel information table; The acquisition module includes a camera for image acquisition, preferably an infrared night vision camera. The acquisition module transmits the collected image data to the Raspberry Pi 3b+, and supports remote login operation and offline operation. Through the python language The program processes the image to identify the temperature measurement object, The identification module is used to complete the following steps: S1. Feature extraction: Calculate and count the gradient direction histogram of the local area of the image to form a feature; the gradient direction is obtained in the following way: convert the detected image into black and white, and for each pixel in the image, compare it with other surrounding Pixels are compared, find and compare the depth of the current pixel to the pixel directly surrounding it, use the arrows to point in the direction of the darkening of the image, repeat this process until every pixel is replaced by an arrow; these arrows are the gradients, they show out the flow from light to dark on the image; S2, distance measurement, obtain the relative position through the face recognition API, and return it to the communication control board to control the operation of the slide rail, and guide the next temperature detection: Specifically, while performing face recognition, the position information of the face after the first recognition is returned for the movement of the motion module, and the camera is calibrated at the same time, based on the offset of the face in the image relative to the center of the image The offset of the actual face relative to the detection terminal is linear in the fixed xoz plane. According to the formula, the fixed proportional coefficient α=θ/p _x is obtained. After the analysis of the face box in the API, the slide rail needs are obtained. The moving distance Δx to achieve the same position of the detection terminal and the detection personnel in the vehicle;

where p _x is the horizontal acquisition pixel of the camera, x ₀ , x ₁ , x ₂ and x ₃ are the position coordinates of the face, and x is the offset of the face in the image relative to the center of the image; The motion and control components include the robotic arm module and the steering gear drive module; The robotic arm module includes a five-degree-of-freedom robotic arm, which is equipped with a detection terminal to realize the necessary detection of people in the vehicle; the robotic arm, as an important motion module, helps the detection personnel to obtain the necessary information for entering the park; the robotic arm is forwarded by the communication control board. , the signal is converted by the steering gear drive board and driven by each steering gear. When the detection terminal judges the position between the vehicle body and the vehicle body through the distance sensor, the drive board can calculate the angle that each motor needs to rotate through inverse kinematics, and finally achieve a convenient car. The location of the personnel detection location; The servo drive module uses the inverse solution of the kinematic equation to realize the motion control of the manipulator. The basic steps include: Step 1. Establish the kinematic relationship based on joint space and Cartesian space The homogeneous transformation matrix is used to establish the pose expression of the end of the manipulator, describing the forward kinematics solution from joint space to Cartesian space or the inverse kinematics solution from Cartesian space to joint space, specifically; using the DH parameter The mathematical model and coordinate system determination system of the manipulator expressing the position and angle relationship between the two pairs of joint links. According to the structural parameters of the manipulator and the motion range of the steering gear, the parameters of the joints and links are obtained, and these parameters are substituted into

in the transformation matrix; According to the formula:

Calculate each joint in turn, and finally get the positive kinematics formula of the manipulator, and substitute the DH parameters to get the rotation matrix of each joint:

After obtaining the rotation matrix of each joint, the coordinates of the end can be obtained according to the following formula:

Step 2. Inverse solution of kinematic equation The coordinates returned by face recognition are used to guide the movement of the robotic arm, so that the detection terminal can be properly positioned with the detection personnel; and based on the homogeneous transformation matrix obtained in step one

Solve the angle θ _i of each rotating joint; the coordinates (x, y) of the endpoint P of the robotic arm consist of three parts (x ₁ +x ₂ +x ₃ , y ₁ +y ₂ +y ₃ ); θ ₁ , θ ₂ and θ ₃ are the angles of the solved steering gear, α is the angle between the claw and the horizontal plane, and α=θ ₁ +θ ₂ +θ ₃ , and: x=l ₁ cosθ ₁ +l ₂ cos(θ ₁ +θ ₂ )+l ₃ cos(θ ₁ +θ ₂ +θ ₃ ) y=l ₁ sinθ ₁ +l ₂ sin(θ ₁ +θ ₂ )+l ₃ sin(θ ₁ +θ ₂ +θ ₃ ) m=l ₃ cosα-x n=l ₃ sinα-y x and y are given by the user; l ₁ , l ₂ and l ₃ are the inherent properties of the mechanical structure of the manipulator. Substitute m and n into the existing equations, and then simplify to get:

It can be obtained by calculation:

The above formula is the form of the root-finding formula of the quadratic equation in one variable, where a=m ² +n ²

According to this, the angles of θ ₁ and θ ₂ can be obtained; the angles of the three steering gears can be obtained based on the above steps, and the control of the coordinate positions can be realized by controlling the steering gears according to the angles. 2. A kind of intelligent management system for park gates during the epidemic period according to claim 1, it is characterized in that, also comprises information interconnection components; Comprises the communication control board based on STM32F103C8T6 as the core, KEIL programming realizes reading sensor information Upload 485 bus, stepper motor control, expand GPIO peripherals; include two TTL serial ports for connecting sensors; one 485 serial port interface for connecting 485 control bus; 5 GPIOs, GPIO includes general timer TIMER3 channel 1, 2; The 74LS244 is used for 3.3V-5V boost and four-wire stepper motor control. 3. A kind of intelligent management system for park gates during the epidemic period according to claim 1, it is characterized in that, the information interconnection component uses KEIL to carry out programming control for STM32F103, UART2 communicates with the host computer, and the receiving interrupt corrects the issued instruction. Check and analyze, control the communication control of UART1 and 3 to the corresponding peripherals; UART1 and its interrupt communicate with the temperature sensor, and UART3 sends coordinate information to the robotic arm according to the command of the host computer; the external interrupt is triggered by the falling edge of the PB0 pin to start the distance sensor . 4. a kind of intelligent management system for park gates during epidemic situation according to claim 1, is characterized in that, identification module comprises identification module, license plate identification module; Described identification module adopts TY-801T embedded module , Read the ID card information through RFID technology, and transmit the read information to the database. The identity recognition module needs to record the identities of people entering and leaving the park and not in the warehouse, and upload the identification information to the database in real time. ;The license plate recognition module uses Http Post, so that the camera is connected to the Http server through ssl. When there is a vehicle, the recognition system will automatically recognize and display the vehicle information.

Images