System for enabling vehicle to follow pedestrian

Abstract

The invention relates to a system for enabling a vehicle to follow a pedestrian, and the system comprises a personnel recognition module which comprises a face recognition sub-module, an intelligent key recognition sub-module and a multi-modal personnel recognition and locking sub-module, the face recognition sub-module is connected with the multi-modal personnel recognition and locking sub-module, and the intelligent key recognition sub-module is connected with the multi-modal personnel recognition and locking sub-module; and the intelligent key recognition sub-module is connected with the multi-modal personnel recognition and locking sub-module, and performs face recognition on the images of the surrounding personnel by adopting a face recognition algorithm based on a Sabot algorithm optimization convolutional neural network to obtain data information of the face images of the surrounding personnel, and the data information is matched with pre-stored facial features of a vehicle owner. According to the invention, people can freely get off the vehicle to appreciate the scenery and do not need to worry about the trouble of returning to pick up the vehicle, the vehicle can automatically follow the people to move forwards, the convenience of outdoor cross-country travel and the user experience are greatly improved, the system cost is reduced, new functions are realized by using an advanced algorithm, and the resource utilization rate is improved.

Description

Technical Field

[0001] The present invention relates to the technical field of driverless vehicles, and particularly to a system for a vehicle to follow a pedestrian. Background Art

[0002] During outdoor off-road activities, there are often areas with beautiful scenery where people hope to get out of the vehicle to enjoy the view. However, traditional vehicles do not have the function of automatically following people. If people go far away from the parking position to enjoy the scenery, they need to spend time and energy to return to the original place to pick up the vehicle when they come back, which brings a lot of inconvenience to users. Although there are some intelligent driving technologies on the market currently, for the outdoor off-road environment, the system for a vehicle to automatically follow people in a scenario with good scenery and few people is not yet perfect.

[0003] In the prior art, a Chinese patent (application number: 202110640464.9, publication number: CN113370170A) discloses an automatic following utility vehicle based on in-vehicle UWB positioning, including a following vehicle, where the following vehicle includes a vehicle body, and a driving unit, a vehicle body position acquisition unit, and a trajectory positioning unit are arranged on the vehicle body. Among them, the vehicle body position acquisition unit is used to obtain the rotation angle between the vehicle body and the driving unit, and transmit the acquired rotation angle to the trajectory positioning unit; the trajectory positioning unit is used to obtain the coordinate information of a moving target, and control the driving unit to follow the moving target according to the obtained coordinate information of the moving target and the rotation angle. In this solution, the recognition accuracy and anti-interference ability in a complex environment (such as a multi-person scenario) are relatively weak.

[0004] In the prior art, a Chinese patent (application number: 202010470680.9, publication number: CN 113741506A) discloses a method and device for an unmanned aerial vehicle to follow a vehicle. The method includes: the unmanned aerial vehicle receives relevant information of the vehicle sent by an in-vehicle terminal; the relevant information of the vehicle includes: the driving speed of the vehicle, the heading angle information of the vehicle, and the first position of the vehicle; the unmanned aerial vehicle predicts a second position that the vehicle will travel to according to the relevant information of the vehicle, and the second position is a position after the first position; the unmanned aerial vehicle adjusts its flight trajectory according to the predicted second position to achieve following the vehicle. Although some vehicles or utility vehicles in this solution have an automatic following function, they mostly rely on a single sensor (such as UWB positioning) or are only applicable to simple environments such as urban roads. In an outdoor off-road environment with complex terrain and few people, the recognition accuracy and following stability are insufficient. Summary of the Invention

[0005] In view of the deficiencies of the above prior art, the present invention provides a vehicle following pedestrian system. Not only can personnel freely get off the vehicle to enjoy the scenery without worrying about the trouble of returning to pick up the vehicle, but the vehicle can automatically follow the personnel forward, greatly improving the convenience and user experience of outdoor off-road travel. Moreover, it reduces the system cost, and at the same time realizes new functions by using advanced algorithms, improving the resource utilization rate.

[0006] To achieve the above and other related objectives, the technical solutions provided by the present invention are as follows:

[0007] A vehicle following pedestrian system, the system includes:

[0008] A personnel recognition module, including a face recognition sub-module, an intelligent key recognition sub-module, and a multi-modal personnel recognition and locking sub-module. The face recognition sub-module is connected to the multi-modal personnel recognition and locking sub-module, and the intelligent key recognition sub-module is connected to the multi-modal personnel recognition and locking sub-module. The face recognition sub-module is used to obtain the image data information of the personnel around the vehicle in real time based on the in-vehicle camera, and use a face recognition algorithm that optimizes the convolutional neural network based on the sand cat algorithm to perform face recognition on the images of the surrounding personnel, obtain the data information of the face images of the surrounding personnel, and match them with the facial features of the pre-stored vehicle owner, and output the data information of the facial image features of the vehicle owner after matching; the intelligent key recognition sub-module is used to output the data information of the vehicle owner's identity recognition in real time; the multi-modal personnel recognition and locking sub-module is used to fuse the vehicle owner's identity and facial image features based on the data information of the vehicle owner's identity recognition and the data information of the facial image features of the vehicle owner after matching, and use a hybrid-level fusion algorithm integrating random inertia weights to output the data information of the vehicle owner's identity lock after fusion.

[0009] Further, the system further includes a vehicle motion control module, connected to the personnel recognition module, including a path planning sub-module and a speed control sub-module. The path planning sub-module is used to obtain the vehicle's own position and surrounding environment information based on the globally positioning system, inertial measurement unit, and environment perception sensor carried by itself, and generate a driving path for the vehicle to follow the personnel according to the data information of the vehicle owner's identity lock after fusion by using a path planning algorithm.

[0010] Further, the speed control sub-module is used to accurately control the driving speed of the vehicle according to the data information of the vehicle owner's identity lock after fusion, so that the vehicle can maintain an appropriate following distance from the personnel. When the walking speed of the personnel increases, the vehicle gradually increases the driving speed through an acceleration algorithm, and vice versa, it decelerates.

[0011] Further, the system further includes a safety guarantee module, which is connected to the personnel identification module. The safety guarantee module includes an obstacle detection and avoidance sub-module and an emergency braking sub-module. The emergency braking sub-module is used to immediately start when the system detects that the distance between the vehicle and the personnel is too close, the personnel suddenly stops or other emergency situations occur, so that the vehicle quickly brakes and stops to avoid collision accidents.

[0012] Further, the obstacle detection and avoidance sub-module is used to transmit the sensor data of the millimeter-wave radar and the ultrasonic radar to the central controller of the vehicle through the CAN bus. The central controller runs the obstacle detection algorithm, analyzes the sensor data, determines whether there is an obstacle, and when an obstacle is detected, the central controller sends the obstacle information to the path planning module. The path planning module re-plans the path and sends the new path information to the chassis control system to achieve vehicle avoidance.

[0013] Further, the face recognition of the images of the surrounding personnel using the face recognition algorithm that optimizes the convolutional neural network based on the sand cat algorithm includes:

[0014] M1. Based on the image data information of the personnel around the vehicle, construct an image data set of the personnel around the vehicle;

[0015] M2. Input the image data set of the personnel around the vehicle into the convolutional neural network for training and learning, identify the facial feature points of the personnel around the vehicle, and obtain the data information of the facial feature points of the personnel around the vehicle;

[0016] M3. Based on the data information of the facial feature points of the personnel around the vehicle, initialize the sand cat population, determine the population parameters and the maximum number of iterations L, and obtain the data information of the initialized sand cat population;

[0017] M4. Based on the data information of the initialized sand cat population, establish a position update function Q of the sand cat population,

[0018]

[0019] where x is the data information of the initialized sand cat population, α1 is any constant parameter between 0 and 1, α2 is the sensitivity value of the sand cat population individuals, and α3 is a random angle between 0 degrees and 360 degrees, to optimize the facial feature points of the personnel around the vehicle and obtain the data information of the optimized facial feature points of the personnel around the vehicle.

[0020] Further, the face recognition of the images of the surrounding personnel using the face recognition algorithm that optimizes the convolutional neural network based on the sand cat algorithm further includes:

[0021] M5. Based on the data information of the facial feature points of the people around the vehicle after optimization, establish a face recognition function R for the people around the vehicle,

[0022]

[0023] where y is the data information of the facial feature points of the people around the vehicle after optimization, and β1, β2, and β3 are weight coefficients, and face recognition is performed on the images of the people around to obtain the data information of the face images of the people around.

[0024] Further, the sensitivity value α2 of the individual of the sand cat population is,

[0025]

[0026] where x is the data information of the initialized sand cat population.

[0027] Further, the fusion of the identity and facial image features of the vehicle owner by using the hybrid-level fusion algorithm with integrated random inertia weight includes:

[0028] U1. Input the data information of the identity recognition of the vehicle owner into the linear regression layer of the model to characterize the feature matrix of the identity recognition of the vehicle owner, and obtain the data information of the feature matrix of the identity recognition of the vehicle owner;

[0029] U2. Based on the data information of the feature matrix of the vehicle owner's identity recognition and the data information of the matched facial image features of the vehicle owner, establish a fusion function G based on integrated random weights,

[0030]

[0031] where z1 is the data information of the feature matrix of the vehicle owner's identity recognition, z2 is the data information of the matched facial image features of the vehicle owner, and δ1, δ2, and δ3 are integrated random weight factors;

[0032] U3. Based on the fusion function G based on integrated random weights, fuse the identity and facial image features of the vehicle owner, and output the data information of the locked identity of the vehicle owner after fusion.

[0033] Further, the integrated random weight factors δ1, δ2, and δ3 are,

[0034]

[0035] where z1 is the data information of the feature matrix of the vehicle owner's identity recognition, z2 is the data information of the matched facial image features of the vehicle owner.

[0036] The present invention has the following positive effects:

[0037] 1. Through the coordinated operation among the personnel recognition module, the vehicle motion control module, and the safety guarantee module, the present invention not only enables personnel to freely get off the vehicle to enjoy the scenery without worrying about the trouble of returning to pick up the vehicle, and the vehicle can automatically follow the personnel forward, greatly improving the convenience of outdoor off-road travel and the user experience, but also reuses the hardware devices such as cameras of the intelligent driving system, reducing the system cost. At the same time, new functions are realized by using advanced algorithms, improving the resource utilization rate.

[0038] 2. The present invention performs face recognition on the images of surrounding personnel through a face recognition algorithm that optimizes the convolutional neural network based on the Sand Cat algorithm, and combines and uses a hybrid-level fusion algorithm that integrates random inertia weights to fuse the identity and facial image features of the vehicle owner, outputting the data information of the identity lock of the fused vehicle owner. It can not only accurately lock the information of the vehicle owner, but also ensure the safety of the vehicle during the process of following the personnel, effectively avoiding the occurrence of accidents such as collisions. BRIEF DESCRIPTION OF THE DRAWINGS

[0039] Figure 1 It is a schematic diagram of the system framework of the present invention;

[0040] Figure 2 It is a schematic flow diagram of the face recognition algorithm that optimizes the convolutional neural network based on the Sand Cat algorithm of the present invention;

[0041] Figure 3 It is a schematic flow diagram of the hybrid-level fusion algorithm that integrates random inertia weights of the present invention. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0042] The following describes exemplary embodiments of the present disclosure, including various details of the embodiments of the present disclosure to facilitate understanding. It should be considered that they are merely exemplary. Therefore, those of ordinary skill in the art should recognize that various changes and modifications can be made to the embodiments described herein without departing from the scope and spirit of the present disclosure. Similarly, for clarity and conciseness, the description below omits the description of well-known functions and structures.

[0043] Embodiment 1: As Figure 1 shown, a system for a vehicle to follow a pedestrian, the method system includes:

[0044] The personnel identification module includes a face recognition sub-module, an intelligent key recognition sub-module, and a multi-modal personnel recognition and locking sub-module. The face recognition sub-module is connected to the multi-modal personnel recognition and locking sub-module, and the intelligent key recognition sub-module is connected to the multi-modal personnel recognition and locking sub-module. The face recognition sub-module is used to obtain the image data information of the personnel around the vehicle in real time based on the in-vehicle camera, perform face recognition on the images of the surrounding personnel using a face recognition algorithm that optimizes the convolutional neural network based on the Sand Cat algorithm, obtain the data information of the face images of the surrounding personnel, and match them with the pre-stored facial features of the vehicle owner, and output the data information of the facial image features of the vehicle owner after matching; The intelligent key recognition sub-module is used to output the data information of the vehicle owner's identity recognition in real time; The multi-modal personnel recognition and locking sub-module is used to fuse the vehicle owner's identity and facial image features based on the data information of the vehicle owner's identity recognition and the data information of the facial image features of the vehicle owner after matching, and output the data information of the vehicle owner's identity locking after fusion.

[0045] In this embodiment, the system further includes a vehicle motion control module, which is connected to the personnel identification module and includes a path planning sub-module and a speed control sub-module. The path planning sub-module is used to obtain the vehicle's own position and surrounding environment information based on the globally positioning system, inertial measurement unit, and environment perception sensor carried by itself, and generate a driving path for the vehicle to follow the personnel according to the data information of the vehicle owner's identity locking after fusion using a path planning algorithm.

[0046] In this embodiment, the speed control sub-module is used to accurately control the driving speed of the vehicle according to the data information of the vehicle owner's identity locking after fusion, so that the vehicle can maintain an appropriate following distance from the personnel. When the walking speed of the personnel increases, the vehicle gradually increases the driving speed through an acceleration algorithm, and vice versa, it decelerates.

[0047] In this embodiment, the system further includes a safety guarantee module, which is connected to the personnel identification module. The safety guarantee module includes an obstacle detection and avoidance sub-module and an emergency braking sub-module. The emergency braking sub-module is used to immediately start when the system detects that the distance between the vehicle and the personnel is too close, the personnel suddenly stops, or other emergency situations occur, so that the vehicle quickly brakes and stops to avoid collision accidents.

[0048] In this embodiment, the obstacle detection and avoidance sub-module is used to transmit the sensor data of the millimeter-wave radar and the ultrasonic radar to the central controller of the vehicle through the CAN bus. The central controller runs the obstacle detection algorithm, analyzes the sensor data, and determines whether there is an obstacle. When an obstacle is detected, the central controller sends the obstacle information to the path planning module. The path planning module re-plans the path and sends the new path information to the chassis control system to achieve vehicle avoidance.

[0049] Implementation of the personnel recognition module

[0050] Face recognition sub-module: Integrate the face recognition algorithm library into the camera driver program of the vehicle intelligent driving system. Upload the facial photos of the vehicle owner and accompanying personnel to the vehicle system through a dedicated APP. After the system pre-processes the photos (such as grayscale conversion, noise reduction, etc.), it extracts facial features and stores them in the local database of the vehicle. When the vehicle starts the automatic following function, the camera captures the surrounding images at a certain frequency (such as 30 frames per second), and uses the face recognition algorithm to identify and compare the personnel in the images.

[0051] Smart key recognition sub-module: The smart key uses Bluetooth 5.0 technology to communicate with the vehicle. Send information such as the signal strength, orientation, and distance of the smart key to the vehicle. On the vehicle side, set up a Bluetooth communication module and write a communication program to achieve pairing, connection, and data transmission with the smart key. The smart key is equipped with a built-in battery that can be charged through a USB interface. Its internal chip continuously generates encrypted data containing identity recognition information and sends it to the vehicle through a Bluetooth signal. After receiving the signal, the vehicle decrypts and authenticates it to confirm the target following personnel.

[0052] Implementation of the vehicle motion control module

[0053] Path planning sub-module: This software module integrates a path planning algorithm library. The vehicle's GPS module real-time obtains the vehicle's position information, the IMU measures the vehicle's attitude information, and the millimeter-wave radar and ultrasonic radar obtain surrounding obstacle information. The path planning software plans a following path based on the personnel position (through face recognition and smart key positioning) and environmental information, and sends the path information to the vehicle's chassis control system.

[0054] Speed control sub-module: Through the camera image analysis algorithm, extract the limb movement characteristics of the personnel, and combine with a deep learning model to predict the walking speed of the personnel. At the same time, according to the change in the Bluetooth signal strength between the smart key and the vehicle, use the signal propagation model to calculate the change in the distance between the personnel and the vehicle, and then obtain the moving speed of the personnel. The vehicle's motor controller adjusts the output power of the motor according to the control instruction to achieve precise control of the vehicle's driving speed.

[0055] The present invention adopts an adaptive dynamic following algorithm, enabling the vehicle to adjust its own driving speed and direction in real time according to the moving speed and direction of the person, ensuring a stable relative position with the person. The system not only tracks the current position of the person in real time but also calculates the future position of the person in advance through a trajectory prediction algorithm, thereby adjusting the driving path of the vehicle in advance. This algorithm can handle situations such as sudden acceleration, deceleration, or change of direction of the person, effectively reducing the lag phenomenon during the following process and improving the following accuracy.

[0056] Implementation of the safety guarantee module

[0057] Obstacle detection and avoidance sub-module: The sensor data of the millimeter-wave radar and ultrasonic radar are transmitted to the vehicle's central controller through the CAN bus. The central controller runs an obstacle detection algorithm to analyze the sensor data and determine whether there are obstacles. When an obstacle is detected, the central controller sends the obstacle information to the path planning software, which re-plans the path and sends the new path information to the chassis control system to achieve vehicle avoidance.

[0058] Emergency braking sub-module: An emergency braking actuator is added to the vehicle's braking system, and this actuator is connected to the central controller. When the central controller determines that emergency braking is required based on camera, radar data, or an emergency stop signal from the smart key, it immediately sends a braking instruction to the emergency braking actuator, and the actuator quickly brakes the vehicle to a stop hydraulically or electronically.

[0059] Embodiment 2: Based on the system of a vehicle following a pedestrian in Embodiment 1, the present invention will be further described and explained below.

[0060] As Figure 1 shown, a system for a vehicle to follow a pedestrian, the method system includes:

[0061] The personnel recognition module includes a face recognition sub-module, an intelligent key recognition sub-module, and a multi-modal personnel recognition and locking sub-module. The face recognition sub-module is connected to the multi-modal personnel recognition and locking sub-module, and the intelligent key recognition sub-module is connected to the multi-modal personnel recognition and locking sub-module. The face recognition sub-module is used to obtain the image data information of the personnel around the vehicle in real time based on the in-vehicle camera, perform face recognition on the images of the surrounding personnel using a face recognition algorithm that optimizes the convolutional neural network based on the sand cat algorithm, obtain the data information of the face images of the surrounding personnel, and match them with the pre-stored facial features of the vehicle owner, and output the data information of the facial image features of the vehicle owner after matching; the intelligent key recognition sub-module is used to output the data information of the vehicle owner's identity recognition in real time; the multi-modal personnel recognition and locking sub-module is used to fuse the vehicle owner's identity and facial image features based on the data information of the vehicle owner's identity recognition and the data information of the facial image features of the vehicle owner after matching, and output the data information of the vehicle owner's identity locking after fusion.

[0062] In this embodiment, as Figure 2 shown, the face recognition of the images of the surrounding personnel using the face recognition algorithm that optimizes the convolutional neural network based on the sand cat algorithm includes:

[0063] M1. Based on the image data information of the personnel around the vehicle, construct an image data set of the personnel around the vehicle;

[0064] M2. Input the image data set of the personnel around the vehicle into the convolutional neural network for training and learning, identify the facial feature points of the personnel around the vehicle, and obtain the data information of the facial feature points of the personnel around the vehicle;

[0065] M3. Based on the data information of the facial feature points of the personnel around the vehicle, initialize the sand cat population, determine the population parameters and the maximum number of iterations L, and obtain the data information of the initialized sand cat population;

[0066] M4. Based on the data information of the initialized sand cat population, establish a position update function Q of the sand cat population,

[0067]

[0068] where x is the data information of the initialized sand cat population, α1 is any constant parameter between 0 and 1, α2 is the sensitivity value of the sand cat population individuals, and α3 is a random angle between 0 degrees and 360 degrees, optimize the facial feature points of the personnel around the vehicle, and obtain the data information of the optimized facial feature points of the personnel around the vehicle.

[0069] In this embodiment, the face recognition algorithm for the images of surrounding people by using the convolutional neural network optimized based on the sand cat algorithm further includes:

[0070] M5. Based on the data information of the facial feature points of the people around the vehicle after optimization, establish a face recognition function R for the people around the vehicle,

[0071]

[0072] where y is the data information of the facial feature points of the people around the vehicle after optimization, β1, β2, and β3 are weight coefficients, and face recognition is performed on the images of the surrounding people to obtain the data information of the face images of the surrounding people.

[0073] In this embodiment, the sensitivity value α2 of the individuals in the sand cat population is

[0074]

[0075] where x is the data information of the sand cat population after initialization.

[0076] In this embodiment, as Figure 3 shown, the fusion of the identity and facial image features of the vehicle owner by using the hybrid-level fusion algorithm with integrated random inertia weights includes:

[0077] U1. Input the data information of the identity recognition of the vehicle owner into the linear regression layer of the model to characterize the feature matrix of the identity recognition of the vehicle owner, and obtain the data information of the feature matrix of the identity recognition of the vehicle owner;

[0078] U2. Based on the data information of the feature matrix of the vehicle owner's identity recognition and the data information of the matched facial image features of the vehicle owner, establish a fusion function G based on integrated random weights,

[0079]

[0080] where z1 is the data information of the feature matrix of the vehicle owner's identity recognition, z2 is the data information of the matched facial image features of the vehicle owner, and δ1, δ2, and δ3 are integrated random weight factors;

[0081] U3. Based on the fusion function G based on integrated random weights, fuse the identity and facial image features of the vehicle owner, and output the data information of the locked identity of the vehicle owner after fusion.

[0082] In this embodiment, the integrated random weight factors δ1, δ2, and δ3 are

[0083]

[0084]

[0085] Among them, z1 is the data information of the feature matrix for vehicle owner identity recognition, and z2 is the data information of the facial image features of the matched vehicle owner.

[0086] Any reference to a memory, storage, database, or other medium used in the embodiments provided by this application may include non-volatile and / or volatile memories. Non-volatile memory may include read-only memory (ROM), programmable ROM (PROM), electrically programmable ROM (EPROM), electrically erasable programmable ROM (EEPROM), or flash memory. Volatile memory may include random access memory (RAM) or external cache memory. By way of illustration and not limitation, RAM is available in various forms, such as static RAM (SRAM), dynamic RAM (DRAM), synchronous DRAM (SDRAM), double data rate SDRAM (DDR SDRAM), enhanced SDRAM (ESDRAM), synchronous link DRAM (SLDRAM), Rambus direct RAM (RDRAM), direct memory bus dynamic RAM (DRDRAM), and Rambus dynamic RAM (RDRAM), etc.

[0087] In summary, in the present invention, not only can people get off the vehicle freely to enjoy the scenery without worrying about the trouble of returning to pick up the vehicle, but also the vehicle can automatically follow the people forward, greatly improving the convenience and user experience of outdoor off-road travel. Moreover, the system cost is reduced, and at the same time, new functions are achieved by using advanced algorithms, improving the resource utilization rate.

[0088] The above specific embodiments do not constitute a limitation on the protection scope of the present disclosure. Those skilled in the art should understand that various modifications, combinations, sub-combinations, and substitutions can be made according to design requirements and other factors. Any modifications, equivalent replacements, and improvements made within the spirit and principles of the present disclosure shall be included within the protection scope of the present disclosure.

Claims

1. A system for a vehicle to follow a pedestrian, characterized in that, The system includes: A personnel identification module, including a face recognition sub-module, an intelligent key recognition sub-module, and a multi-modal personnel recognition and locking sub-module. The face recognition sub-module is connected to the multi-modal personnel recognition and locking sub-module, and the intelligent key recognition sub-module is connected to the multi-modal personnel recognition and locking sub-module. The face recognition sub-module is used to obtain the image data information of the personnel around the vehicle in real time based on the in-vehicle camera, perform face recognition on the images of the surrounding personnel using a face recognition algorithm that optimizes the convolutional neural network based on the sand cat algorithm, obtain the data information of the face images of the surrounding personnel, and match them with the pre-stored facial features of the vehicle owner, and output the data information of the facial image features of the vehicle owner after matching; The intelligent key recognition sub-module is used to output the data information of the vehicle owner's identity recognition in real time; The multi-modal personnel recognition and locking sub-module is used to fuse the vehicle owner's identity and facial image features based on the data information of the vehicle owner's identity recognition and the data information of the facial image features of the vehicle owner after matching, and output the data information of the vehicle owner's identity locking after fusion using a hybrid-level fusion algorithm that integrates random inertia weights.

2. The system for a vehicle to follow a pedestrian according to claim 1, wherein, The system further includes a vehicle motion control module, which is connected to the personnel identification module and includes a path planning sub-module and a speed control sub-module. The path planning sub-module is used to obtain the vehicle's own position and surrounding environment information based on the globally positioning system, inertial measurement unit, and environment perception sensor carried by itself, and generate the driving path of the vehicle following the personnel using a path planning algorithm according to the data information of the vehicle owner's identity locking after fusion.

3. The system for a vehicle to follow a pedestrian according to claim 2, characterized in that: The speed control sub-module is used to accurately control the driving speed of the vehicle according to the data information of the vehicle owner's identity locking after fusion, so that the vehicle can maintain an appropriate following distance from the personnel. When the walking speed of the personnel increases, the vehicle gradually increases the driving speed through an acceleration algorithm, and vice versa, it decelerates.

4. The system for a vehicle to follow a pedestrian according to claim 1, wherein The system further includes a safety guarantee module, which is connected to the personnel identification module. The safety guarantee module includes an obstacle detection and avoidance sub-module and an emergency braking sub-module. The emergency braking sub-module is used to immediately start when the system detects that the distance between the vehicle and the personnel is too close, the personnel suddenly stops, or other emergency situations occur, so that the vehicle quickly brakes and stops to avoid collision accidents.

5. The system for a vehicle to follow a pedestrian according to claim 4, wherein: The obstacle detection and avoidance sub-module is used to transmit the sensor data of the millimeter-wave radar and ultrasonic radar to the vehicle's central controller through the CAN bus. The central controller runs an obstacle detection algorithm to analyze the sensor data to determine whether there are obstacles. When an obstacle is detected, the central controller sends the obstacle information to the path planning module, and the path planning module re-plans the path and sends the new path information to the chassis control system to achieve vehicle avoidance.

6. The vehicle following pedestrian system according to claim 1, characterized in that, The face recognition of the images of the surrounding personnel using a face recognition algorithm that optimizes the convolutional neural network based on the sand cat algorithm includes: M1. Based on the image data information of the personnel around the vehicle, construct an image data set of the personnel around the vehicle; M2. Input the image dataset of the people around the vehicle into a convolutional neural network for training and learning, identify the facial feature points of the people around the vehicle, and obtain the data information of the facial feature points of the people around the vehicle; M3. Based on the data information of the facial feature points of the people around the vehicle, initialize the sand cat population, determine the population parameters and the maximum number of iterations L, and obtain the data information of the initialized sand cat population; M4. Based on the data information of the initialized sand cat population, establish the position update function Q of the sand cat population, where x is the data information of the initialized sand cat population, α1 is any constant parameter between 0 and 1, α2 is the sensitivity value of the individual of the sand cat population, and α3 is a random angle between 0 degrees and 360 degrees, optimize the facial feature points of the people around the vehicle, and obtain the data information of the optimized facial feature points of the people around the vehicle.

7. The system for a vehicle to follow a pedestrian according to claim 6, wherein, The face recognition algorithm for the images of the people around based on the sand cat algorithm to optimize the convolutional neural network further includes: M5. Based on the data information of the optimized facial feature points of the people around the vehicle, establish the face recognition function R of the people around the vehicle, where y is the data information of the optimized facial feature points of the people around the vehicle, and β1, β2, and β3 are weight coefficients, perform face recognition on the images of the people around, and obtain the data information of the face images of the people around.

8. The vehicle following pedestrian system according to claim 6, characterized in that, The sensitivity value α2 of the individual of the sand cat population is where x is the data information of the initialized sand cat population.

9. The system for a vehicle to follow a pedestrian according to claim 1, wherein The hybrid-level fusion algorithm using the integrated random inertia weight to fuse the identity and facial image features of the vehicle owner includes: U1. Input the data information of the identity recognition of the vehicle owner into the linear regression layer of the model to characterize the feature matrix of the identity recognition of the vehicle owner, and obtain the data information of the feature matrix of the identity recognition of the vehicle owner; U2. Based on the data information of the feature matrix of the identity recognition of the vehicle owner and the data information of the matched facial image features of the vehicle owner, establish the fusion function G based on the integrated random weight, where z1 is the data information of the feature matrix of the identity recognition of the vehicle owner, z2 is the data information of the matched facial image features of the vehicle owner, and δ1, δ2, and δ3 are integrated random weight factors; U3. Based on the fusion function G based on the integrated random weight, fuse the identity and facial image features of the vehicle owner, and output the data information of the locked identity of the vehicle owner after fusion.

10. The system for a vehicle to follow a pedestrian according to claim 9, characterized in that: The integrated random weight factors δ1, δ2, and δ3 are where z1 is the data information of the feature matrix of the identity recognition of the vehicle owner, z2 is the data information of the matched facial image features of the vehicle owner.

Images