Intelligent control system and method based on unmanned vehicle

Abstract

The invention provides an intelligent control system and method based on an unmanned vehicle, and solves the problem of low driving efficiency caused by the fact that the current road traffic condition and the driving state of a front vehicle are not fully considered during lane changing operation of an existing unmanned vehicle. According to the invention, a multi-modal information fusion technology is adopted, and the driving speed, the vehicle type and other information of the front vehicle are comprehensively considered to judge whether the lane changing operation needs to be actively carried out, so that the efficiency and the safety of unmanned driving are improved. Besides, the selection of the lane changing direction is optimized, and the driving efficiency of the candidate lane is predicted by evaluating the number of vehicles on the candidate lane and the lane stability, so that invalid or unnecessary lane changing behaviors are avoided.

Description

Technical Field

[0001] The present invention relates to the field of autonomous driving technology, and in particular to an intelligent vehicle control system and method for an unmanned vehicle during driving. Background Art

[0002] With the rapid development of machine learning, artificial intelligence, sensor technology and computer vision, driverless cars have progressed from the theoretical stage to practical application. The core technologies of this type of car mainly involve environmental perception, driving route planning and automatic execution of driving operations. Especially in complex traffic environments, lane changing function has become an important part of driverless cars, which requires the vehicle to be able to autonomously make lane change decisions and effectively execute them based on the traffic conditions ahead, driving goals and safety standards.

[0003] In current driverless car technology, lane changes are mostly caused by external factors, such as encountering obstacles, road construction, vehicle failure, or following a preset navigation route. These traditional lane change strategies are mainly to avoid emergencies or comply with navigation instructions, rather than making proactive decisions based on real-time traffic flow changes.

[0004] Existing technologies have not yet effectively dealt with the need to actively change lanes due to the slow speed of the vehicle ahead in unobstructed conditions. On multi-lane roads such as highways or urban expressways, the slow speed of the vehicle ahead may slow down the speed of the vehicle behind, reducing driving efficiency. Current technology still lacks a mechanism that allows driverless cars to actively switch lanes according to the traffic flow conditions of each lane and find a smoother or more efficient driving route, thereby reducing driving time and improving fuel or battery efficiency.

[0005] Therefore, the present invention proposes a method and system for realizing intelligent control of an unmanned vehicle based on a relationship module, aiming to overcome the problem of ignoring real-time lane traffic conditions in the prior art, thereby improving the overall driving performance and user experience of the unmanned vehicle. Summary of the invention

[0006] In view of this, the present invention provides an intelligent control method based on an unmanned vehicle: the method comprises the following steps: S1: When an active lane change request is received, the candidate lane change direction is determined by the laser radar devices on the side and rear; S2: when the candidate lane-changing direction is one side, the candidate lane-changing direction determined in S1 is used as the final lane-changing direction; when the candidate lane-changing direction is both sides, the side front cameras on both side rearview mirrors are turned on to collect side front video images, and the final lane-changing direction is determined based on the side front video analysis results; Determining the final lane change direction based on the side front video analysis result specifically includes: Use the relationship module to get the association score between targets: ; is the balance coefficient used to balance the detection confidence and the associated targets in the previous and next frames. and They represent the confidence of the i-th target in the k-1-th frame and the confidence of the j-th target in the k-th frame, respectively. Propose a box for the primary ROI in the k-1th frame The i-th test result in Propose a box for the primary ROI in the kth frame The jth detection result in the video is obtained by iteratively calculating all the association matrices in the video except the first frame, and obtaining the video association matrix , is the number of video frames; Based on the video association matrix M, multiple best association connections are screened out, and the stability score of each target is calculated through the best association connection: ; in, It represents the probability that the detection result of the best associated connection ti in the kth frame belongs to the target i. The number of best associated connections n is equal to the number of targets detected in the last frame. Represents the association score between target i and target j in the kth frame in the association connection mode t.

[0007] The present invention provides an intelligent control system based on an unmanned vehicle, the system comprising: a candidate lane change direction determination module, wherein upon receiving an active lane change request, the candidate lane change direction determination module determines a candidate lane change direction through a laser radar device at the side and rear; Candidate lane change direction determination module: when receiving an active lane change request, the candidate lane change direction determination module determines the candidate lane change direction through the laser radar devices on the side and the rear side; Final lane-changing direction determination module: The final lane-changing direction determination module is used to, when the candidate lane-changing direction is one side, use the candidate lane-changing direction determined in the candidate lane-changing direction determination module as the final lane-changing direction; when the candidate lane-changing direction is two sides, turn on the side front cameras on the rearview mirrors on both sides to collect side front video images, and determine the final lane-changing direction based on the side front video analysis results; Determining the final lane change direction based on the side front video analysis result specifically includes: Use the relationship module to get the association score between targets: ; is the balance coefficient used to balance the detection confidence and the associated targets in the previous and next frames. and They represent the confidence of the i-th target in the k-1-th frame and the confidence of the j-th target in the k-th frame, respectively. is the primary ROI proposal box in the k-1th frame The i-th test result, Propose a box for the primary ROI in the kth frame The jth detection result in the video is obtained by iteratively calculating all the association matrices in the video except the first frame, and obtaining the video association matrix , is the number of video frames; Based on the video association matrix M, multiple best association connections are screened out, and the stability score of each target is calculated through the best association connection: ;in, It represents the probability that the detection result of the best associated connection ti in the kth frame belongs to target i. The number of best associated connections n is equal to the number of targets detected in the last frame. Represents the association score between target i and target j in the kth frame in the association connection mode t.

[0008] The present invention also provides an electronic device, comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor implements the above-mentioned intelligent control method based on an unmanned vehicle when executing the computer program.

[0009] The present invention also provides a computer-readable storage medium, wherein the computer-readable storage medium stores a computer program, and when the computer program is executed by a processor, the above-mentioned intelligent control method based on an unmanned vehicle is implemented.

[0010] Compared with the existing technology, the present invention solves the problem that the existing unmanned vehicles do not fully consider the current road traffic conditions and the driving status of the preceding vehicle when performing lane change operations, resulting in reduced driving efficiency. The present invention adopts multimodal information fusion technology to comprehensively consider the driving speed, vehicle model and other information of the preceding vehicle to determine whether it is necessary to actively change lanes, thereby improving the efficiency and safety of unmanned driving. In addition, the present invention also optimizes the selection of lane change direction, and predicts the driving efficiency of the candidate lane by evaluating the number of vehicles on the candidate lane and the lane stability, thereby avoiding invalid or unnecessary lane change behaviors. BRIEF DESCRIPTION OF THE DRAWINGS

[0011] In order to more clearly illustrate the technical solutions of the embodiments of the present application, the drawings required for use in the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present application. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.

[0012] Figure 1 It is a specific target recognition network structure diagram in this application; Figure 2 It is a schematic diagram of the settings of the camera device and the lidar device in this application. DETAILED DESCRIPTION

[0013] The embodiments of the present application are described in detail below with reference to the accompanying drawings.

[0014] The following describes the implementation methods of the present application through specific examples, and those skilled in the art can easily understand other advantages and effects of the present application from the contents disclosed in this specification. Obviously, the described embodiments are only a part of the embodiments of the present application, rather than all the embodiments. The present application can also be implemented or applied through other different specific implementation methods, and the details in this specification can also be modified or changed in various ways based on different viewpoints and applications without departing from the spirit of the present application. It should be noted that, in the absence of conflict, the following embodiments and the features in the embodiments can be combined with each other. Based on the embodiments in the present application, all other embodiments obtained by ordinary technicians in the field without making creative work belong to the scope of protection of the present application.

[0015] It should be noted that various aspects of the embodiments within the scope of the appended claims are described below. It should be apparent that the aspects described herein can be embodied in a wide variety of forms, and any specific structure and / or function described herein is merely illustrative. Based on the present application, it should be understood by those skilled in the art that an aspect described herein can be implemented independently of any other aspect, and two or more of these aspects can be combined in various ways. For example, any number and aspect described herein can be used to implement the device and / or practice the method. In addition, other structures and / or functionalities other than one or more of the aspects described herein can be used to implement this device and / or practice this method.

[0016] Additionally, in the following description, specific details are provided to facilitate a thorough understanding of the examples. However, it will be understood by those skilled in the art that the examples can be practiced without these specific details.

[0017] The embodiment of this specification proposes an intelligent control method based on an unmanned vehicle: the method comprises the following steps: S1: When an active lane change request is received, the candidate lane change direction is determined by the laser radar devices on the side and rear; S2: when the candidate lane-changing direction is one side, the candidate lane-changing direction determined in S1 is used as the final lane-changing direction; when the candidate lane-changing direction is both sides, the side front cameras on both side rearview mirrors are turned on to collect side front video images, and the final lane-changing direction is determined based on the side front video analysis results; Determining the final lane change direction based on the side front video analysis result specifically includes: Use the relationship module to get the association score between targets: ; is the balance coefficient used to balance the detection confidence and the associated targets in the previous and next frames. and They represent the confidence of the i-th target in the k-1-th frame and the confidence of the j-th target in the k-th frame, respectively. Propose a box for the primary ROI in the k-1th frame The i-th test result in Propose a box for the primary ROI in the kth frame The jth detection result in the video is obtained by iteratively calculating all the association matrices in the video except the first frame, and obtaining the video association matrix , is the number of video frames; Based on the video association matrix M, multiple best association connections are screened out, and the stability score of each target is calculated through the best association connection: ; in, It represents the probability that the detection result of the best associated connection ti in the kth frame belongs to the target i. The number of best associated connections n is equal to the number of targets detected in the last frame. Represents the association score between target i and target j in the kth frame in the association connection mode t.

[0018] Among them, the front and rear nodes of the best associated connection have the largest relationship score. After each best associated connection is screened out, the elements corresponding to the nodes on the best associated connection in the video associated matrix M are set to zero, and the best associated connection is continued to be screened using the updated video associated matrix to obtain the best associated connection set T.

[0019] After calculating the stability score of each target through the best association connection, the method further includes: obtaining the number of rearview mirrors in the left front image and the right front image according to the detection. , and their respective target stability scores , Determine the final lane change direction; ; ; ; in, represents the initial judgment score, and It is the attenuation fraction of ... number of rearview mirrors and the attenuation fraction of the stability fraction.

[0020] Before determining the candidate lane change direction by the side and rear laser radar devices, it also includes whether to perform active lane change judgment, and the active lane change judgment includes: S3: Obtain image information collected by the front camera of the unmanned vehicle; S4: performing specific target recognition on the image information to obtain the preceding vehicle category information, performing preceding vehicle driving status analysis based on the position information and the speed information, and determining whether to perform active lane change according to the preceding vehicle category information and the preceding vehicle driving status analysis results.

[0021] like Figure 2 As shown, the present application sets a plurality of cameras and laser radar devices in the unmanned vehicle, wherein the front camera is set at the front of the unmanned vehicle, and its specific position can be set above or below the logo of the unmanned vehicle. The front camera is mainly used to shoot the picture directly in front of the vehicle, and the shooting angle can be adjusted; The side front camera is set on the rearview mirrors on both sides of the unmanned vehicle. The side front camera is used to shoot the side and front scenes of the vehicle, and the shooting angle can be adjusted; The front laser radar is set at the front of the unmanned vehicle, and its specific position can be set above or below the unmanned vehicle logo. The front laser radar is mainly used to measure the real-time distance between the current unmanned vehicle and the vehicle directly in front.

[0022] The side laser radars are arranged on both sides of the unmanned vehicle, and their specific positions can be set above the front wheels of the vehicle. The side laser radars are mainly used to determine whether there are any moving vehicles on the side of the current unmanned vehicle and the real-time distance.

[0023] The side rear laser radars are installed on both sides of the unmanned vehicle, and their specific positions can be set above the rear wheels of the vehicle or near the taillights on both sides. The side rear laser radars are mainly used to determine whether there are any moving vehicles behind the current unmanned vehicle and the real-time distance.

[0024] The unmanned vehicle of the present application is equipped with a GNSS positioning system, which is used to obtain the current location information of the unmanned vehicle.

[0025] This application uses a front camera to collect image information directly in front of the unmanned vehicle. The front camera collection frequency is pre-set, and the front camera is used to capture vehicle information directly in front of the current driving direction of the unmanned vehicle.

[0026] This application uses a front-mounted laser radar to collect distance information between the current unmanned vehicle and the vehicle directly in front.

[0027] Performing specific target recognition on the image information to obtain the preceding vehicle category information specifically includes: S41: building a specific target recognition network; S42: pre-collecting road vehicle data, where the road vehicle data is a rear image of a vehicle while it is traveling on the road; S43: labeling the collected vehicle data, wherein the labeling information includes position labeling of the vehicle, license plate, and vehicle identification in the image and category labeling of the license plate and vehicle identification, wherein the position labeling is to select the position of the entire rear of the vehicle, the license plate range, and the vehicle identification range; The license plate categories are marked as: blue license plate and non-blue license plate, and the vehicle identification categories are marked as: novice identification and non-novice identification; S44: Divide the labeled vehicle data to obtain a training set, and train a specific target recognition network; S45: Input the image captured by the front camera of the current unmanned vehicle into the specific target recognition network, and make an active lane change judgment based on the front vehicle category information output by the specific target recognition network.

[0028] The analysis of the driving status of the preceding vehicle based on the position information and speed information specifically includes: S46: When receiving a request for analyzing the driving status of the preceding vehicle, the preceding laser radar detector is started, and the distance S_1 between the current unmanned vehicle and the preceding vehicle at the first moment is calculated based on the time delay difference between sending the laser and receiving the laser; S47: Repeat step S46 to calculate the distance S_2 between the current unmanned vehicle and the preceding vehicle at the second moment, and obtain the preceding vehicle speed information V_f according to the distance S_1 at the first moment, the distance S_2 at the second moment, the time interval t, and the speed information V_c of the current unmanned vehicle; S48: starting the GNSS system to obtain the current position information of the unmanned vehicle, performing position matching between the current position information of the unmanned vehicle and the electronic map, and obtaining the road information on which the current unmanned vehicle is traveling, wherein the road information includes the maximum speed limit information V_max of the current road and the congestion condition of the current road; S49: pre-set the standard low speed percentage threshold, and dynamically lower the standard low speed percentage threshold according to the current road congestion to obtain the low speed percentage threshold N% of the current road at the current moment. , request active lane change, otherwise return to step S3 to perform image acquisition at the next moment.

[0029] Front vehicle speed information V_f=(S_2-S_1) / t+V_c; The current road congestion conditions include normal traffic, light congestion, moderate congestion, and severe congestion.

[0030] In addition to the type of vehicle in front, the present application also analyzes the driving status of the vehicle in front. The present application measures the speed of the vehicle in front through radar laser, and judges the driving status of the vehicle in front according to the speed of the vehicle in front and the speed limit information of the current road section. If the speed of the vehicle in front is too low, the driver of the vehicle in front may have bad driving behaviors such as driving fatigue, making phone calls, playing with mobile phones, etc. At the same time, following the vehicle in front driving at a low speed is also not conducive to the traffic efficiency of the unmanned vehicle. Therefore, when the speed of the vehicle in front is too low, the present invention will also request active lane change to ensure the driving safety and efficiency of the unmanned vehicle.

[0031] When an active lane change request is received, determining a candidate lane change direction by using the laser radar devices on the side and the rear side specifically includes: S11: presetting a dangerous lane-changing area of ​​the current unmanned vehicle, wherein the dangerous lane-changing area is a fan-shaped area on both sides constructed with the geometric center of the current unmanned vehicle as the vertex; S12: Turn on the laser radar devices on the side and rear, and calculate whether there are any vehicles on the side and rear of the current unmanned vehicle and the distance between the vehicles based on the time delay difference between the laser sending and receiving of the laser radar devices; S13: If there are no vehicles on both sides of the current unmanned vehicle, determining that the candidate lane change direction is both sides; If there are vehicles on both sides of the current driverless vehicle, determine whether the vehicles on both sides are in the dangerous lane change area. If the vehicles on both sides are in the dangerous lane change area, cancel the lane change request; if the vehicles on both sides are not in the dangerous lane change area, determine the candidate lane change direction as both sides; if only one side of the vehicle is in the dangerous lane change area, determine the candidate lane change direction as the other side; If there is a vehicle driving on only one side of the current unmanned vehicle, it is determined whether the vehicle driving on that side is in a dangerous lane change area. If not in the dangerous area, the candidate lane change direction is determined to be both sides. If in the dangerous area, the other side is used as a candidate lane change direction.

[0032] The present invention fully considers the basic element of the type of the preceding vehicle in road traffic. First, a specific target recognition network is established and trained, and the specific recognition network can be used to identify specific targets. In an actual road environment, if the preceding vehicle type is a large truck, a large bus, a school bus, etc., due to its large size and serious obstruction of vision, the driving risk of the following vehicle will be increased. Similarly, if the preceding vehicle type is a training vehicle or the preceding vehicle is affixed with a novice logo, since the driver of the preceding vehicle is in the training practice period or internship period, the driving skills are not yet mature, and he may make dangerous driving behaviors such as sudden stops and sharp turns. Therefore, the following vehicle following the training vehicle or the vehicle marked with a novice also has a greater driving risk. The risk of an unmanned vehicle driving on the road following the above-mentioned type of vehicle is higher than that of other types of vehicles. Therefore, the present invention identifies the preceding vehicle type based on the common characteristic yellow license plate of large vehicles and training vehicles and the circular label of the vehicle marking the novice. When it is identified that the preceding vehicle type belongs to the above-mentioned vehicle that may increase the driving risk of the unmanned vehicle, an active lane change request is started. The lane change request of the present invention is not a passive lane change due to a real traffic accident ahead or due to driving planning, but an active lane change based on the estimated driving risk according to the type of the preceding vehicle. This active lane change method based on the type of preceding vehicle can reduce the danger of unmanned vehicle driving and improve the safety of passengers.

[0033] The specific target recognition network includes a feature extraction module, a multi-scale feature fusion module, a multi-scale suggestion box generation module, and a target output module; The feature extraction layer includes four convolution modules, an attention module is added between two adjacent convolution modules, and the outputs of the second, third and fourth convolution modules are connected to the multi-scale feature fusion module and the multi-scale suggestion box generation module respectively; The multi-scale suggestion box generation module includes convolution kernels of multiple scales, and the convolution kernels of multiple scales are applied to the feature maps output by the second, third, and fourth convolution modules respectively to obtain position information of suggestion boxes of multiple scales; The multi-scale fusion module inputs the feature map upsampled by the second convolution module, output by the third convolution module, and outputs the fused feature map; The position information of the multiple scale suggestion boxes is used to guide the fusion feature map to obtain multiple suggestion boxes, and the fusion feature map with the multiple suggestion boxes is input into the target output module to obtain the position and category information of the vehicle, license plate and vehicle logo.

[0034] The present invention requires the overall recognition of the rear of the preceding vehicle in the captured image to determine whether there is a preceding vehicle in front of the current unmanned vehicle. At the same time, the license plate and vehicle logo must also be recognized in the captured image. Obviously, the size of the entire vehicle and the license plate or logo is very different. If a single convolutional recognition network is used, it is impossible to achieve a high recognition accuracy for targets of all scales. Therefore, the present application uses a multi-scale convolution kernel to perform convolution processing on feature maps of different levels in the feature extraction network, and uses a multi-scale convolution kernel to adapt to the target features of different scales to be recognized by the present application. Using feature maps of different levels to perform multi-scale convolution operations separately can identify the same target by fusing the feature details of multiple feature maps, thereby improving the accuracy of target recognition and avoiding the problem of information loss caused by too many convolution operations on small target features.

[0035] The present invention arranges side front cameras at the positions of the left and right rearview mirrors of the vehicle. The front cameras in the prior art are mostly arranged at the vehicle logo or above the windshield. Although the above cameras can shoot and observe a certain range in front of the vehicle, due to the limited height of the vehicle and the line of sight blocked by the vehicle in front, in fact, the front cameras can only shoot a very limited range in front and side front. In the prior art, when judging the driving condition of the vehicle in front, it is often necessary to move the vehicle laterally, that is, to drive the vehicle close to one side of the lane line to obtain a wider front field of view. However, the driving behavior of the above vehicle driving close to one side of the lane line (driving on the line) increases the driving risk and is likely to cause panic to passengers. Therefore, the present invention arranges side front cameras at the left and right rearview mirrors of the vehicle to obtain a wide field of view in the side and front, avoiding the need to move the vehicle laterally and incurring risks.

[0036] The present invention also optimizes lane change selection. When lanes can be changed to the left and the right at the same time, the present invention uses a side front camera to capture a side front image, and determines the number of vehicles in front of the lanes on both sides by analyzing the number of rearview mirrors in the image. At the same time, the stability scores of the vehicles in front of the lanes on both sides are used to predict the safety and necessity after the lane change. When vehicles in a lane change frequently, actively changing lanes to this lane may be unstable, increasing the driving risk. The present invention uses the number of vehicles in front of the lanes on both sides and the stability scores of the lanes on both sides to select lanes with lower traffic density and better stability, thereby ensuring driving safety and efficiency and avoiding invalid lane changes.

[0037] The present invention provides an intelligent control system based on an unmanned vehicle, the system comprising: a candidate lane change direction determination module, wherein upon receiving an active lane change request, the candidate lane change direction determination module determines a candidate lane change direction through a laser radar device at the side and rear; Final lane-changing direction determination module: The final lane-changing direction determination module is used to, when the candidate lane-changing direction is one side, use the candidate lane-changing direction determined in the candidate lane-changing direction determination module as the final lane-changing direction; when the candidate lane-changing direction is two sides, turn on the side front cameras on the rearview mirrors on both sides to collect side front video images, and determine the final lane-changing direction based on the side front video analysis results; Determining the final lane change direction based on the side front video analysis result specifically includes: Use the relationship module to get the association score between targets: ; is the balance coefficient used to balance the detection confidence and the associated targets in the previous and next frames. and They represent the confidence of the i-th target in the k-1-th frame and the confidence of the j-th target in the k-th frame, respectively. is the primary ROI proposal box in the k-1th frame The i-th test result, Propose a box for the primary ROI in the kth frame The jth detection result in the video is obtained by iteratively calculating all the association matrices in the video except the first frame, and obtaining the video association matrix , is the number of video frames; Based on the video association matrix M, multiple best association connections are screened out, and the stability score of each target is calculated through the best association connection: ; in, It represents the probability that the detection result of the best associated connection ti in the kth frame belongs to target i. The number of best associated connections n is equal to the number of targets detected in the last frame. Represents the association score between target i and target j in the kth frame in the association connection mode t.

[0038] The present invention also provides an electronic device, comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor implements the above-mentioned intelligent control method based on an unmanned vehicle when executing the computer program.

[0039] The present invention also provides a computer-readable storage medium, wherein the computer-readable storage medium stores a computer program, and when the computer program is executed by a processor, the above-mentioned intelligent control method based on an unmanned vehicle is implemented.

[0040] Those skilled in the art can understand that all or part of the processes in the above-mentioned embodiment methods can be completed by instructing the relevant hardware through a computer program, and the computer program can be stored in a non-volatile computer-readable storage medium. When the computer program is executed, it can include the processes of the embodiments of the above-mentioned methods. Among them, any reference to memory, storage, database or other media used in the embodiments provided in this application can include non-volatile and / or volatile memory. 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. As an illustration and not limitation, RAM is available in many forms, such as static RAM (SRAM), dynamic RAM (DRAM), synchronous DRAM (SDRAM), double data rate SDRAM (DDRSDRAM), enhanced SDRAM (ESDRAM), synchronous link (Synchlink) DRAM (SLDRAM), memory bus (Rambus) direct RAM (RDRAM), direct memory bus dynamic RAM (DRDRAM), and memory bus dynamic RAM (RDRAM).

[0041] In this specification, the same or similar parts between the various embodiments can be referred to each other, and each embodiment focuses on the differences from other embodiments. In particular, for the embodiments described later, the description is relatively simple, and the relevant parts can be referred to the partial description of the previous embodiments.

[0042] The above is only a specific implementation of the present application, but the protection scope of the present application is not limited thereto. Any changes or substitutions that can be easily thought of by a person skilled in the art within the technical scope disclosed in the present application should be included in the protection scope of the present application. Therefore, the protection scope of the present application shall be based on the protection scope of the claims.

Claims

1. An intelligent control method based on an unmanned vehicle, characterized in that: The control method comprises the following steps: S1: When an active lane change request is received, the candidate lane change direction is determined by the laser radar devices on the side and rear; S2: when the candidate lane-changing direction is one side, the candidate lane-changing direction determined in S1 is used as the final lane-changing direction; when the candidate lane-changing direction is both sides, the side front cameras on both side rearview mirrors are turned on to collect side front video images, and the final lane-changing direction is determined based on the side front video analysis results; Determining the final lane change direction based on the side front video analysis result specifically includes: Use the relationship module to get the association score between targets: ； is the balance coefficient used to balance the detection confidence and the associated targets in the previous and next frames. and They represent the confidence of the i-th target in the k-1-th frame and the confidence of the j-th target in the k-th frame, respectively. Propose a box for the primary ROI in the k-1th frame The i-th test result in Propose a box for the primary ROI in the kth frame The jth detection result in the video is obtained by iteratively calculating all the association matrices in the video except the first frame, and obtaining the video association matrix , is the number of video frames; Based on the video association matrix M, multiple best association connections are screened out, and the stability score of each target is calculated through the best association connection: ; in, It represents the probability that the detection result of the best associated connection ti in the kth frame belongs to the target i. The number of best associated connections n is equal to the number of targets detected in the last frame. Represents the association score between target i and target j in the kth frame in the association connection mode t.

2. The method for intelligent control of unmanned vehicles according to claim 1, characterized in that: After calculating the stability score of each target through the best association connection, the method further includes: obtaining the number of rearview mirrors in the left front image and the right front image according to the detection. , and their respective target stability scores , Determine the final lane change direction; ； ； ； in, represents the initial judgment score, and It is the attenuation fraction of ... number of rearview mirrors and the attenuation fraction of the stability fraction.

3. The method for intelligent control of unmanned vehicles according to claim 1, characterized in that: Before determining the candidate lane change direction by the side and rear laser radar devices, it also includes whether to perform active lane change judgment, and the active lane change judgment includes: S3: Obtain image information collected by the front camera of the unmanned vehicle; S4: performing specific target recognition on the image information to obtain the preceding vehicle category information, performing preceding vehicle driving status analysis based on the position information and the speed information, and determining whether to perform active lane change according to the preceding vehicle category information and the preceding vehicle driving status analysis results.

4. The method for intelligent control of unmanned vehicles according to claim 3 is characterized in that: Performing specific target recognition on the image information to obtain the preceding vehicle category information specifically includes: S41: building a specific target recognition network; S42: pre-collecting road vehicle data, where the road vehicle data is a rear image of a vehicle while it is traveling on the road; S43: labeling the collected vehicle data, wherein the labeling information includes position labeling of the vehicle, license plate, and vehicle identification in the image and category labeling of the license plate and vehicle identification, wherein the position labeling is to select the position of the entire rear of the vehicle, the license plate range, and the vehicle identification range; The license plate categories are marked as: blue license plate and non-blue license plate, and the vehicle identification categories are marked as: novice identification and non-novice identification; S44: Divide the labeled vehicle data to obtain a training set, and train a specific target recognition network; S45: Input the image captured by the front camera of the current unmanned vehicle into the specific target recognition network, and make an active lane change judgment based on the front vehicle category information output by the specific target recognition network.

5. The method for intelligent control of unmanned vehicles according to claim 3 is characterized in that: The analysis of the driving status of the preceding vehicle based on the position information and speed information specifically includes: S46: When receiving a request for analyzing the driving status of the preceding vehicle, the preceding laser radar detector is started, and the distance S_1 between the current unmanned vehicle and the preceding vehicle at the first moment is calculated based on the time delay difference between sending the laser and receiving the laser; S47: Repeat step S46 to calculate the distance S_2 between the current unmanned vehicle and the preceding vehicle at the second moment, and obtain the preceding vehicle speed information V_f according to the distance S_1 at the first moment, the distance S_2 at the second moment, the time interval t, and the speed information V_c of the current unmanned vehicle; S48: starting the GNSS system to obtain the current position information of the unmanned vehicle, performing position matching between the current position information of the unmanned vehicle and the electronic map, and obtaining the road information on which the current unmanned vehicle is traveling, wherein the road information includes the maximum speed limit information V_max of the current road and the congestion condition of the current road; S49: pre-set the standard low speed percentage threshold, and dynamically lower the standard low speed percentage threshold according to the current road congestion to obtain the low speed percentage threshold N% of the current road at the current moment. , request active lane change, otherwise return to step S3 to perform image acquisition at the next moment.

6. The method for intelligent control of unmanned vehicles according to claim 1, characterized in that: When an active lane change request is received, determining a candidate lane change direction by using the laser radar devices on the side and the rear side specifically includes: S11: presetting a dangerous lane-changing area of ​​the current unmanned vehicle, wherein the dangerous lane-changing area is a fan-shaped area on both sides constructed with the geometric center of the current unmanned vehicle as the vertex; S12: Turn on the laser radar devices on the side and rear, and calculate whether there are any vehicles on the side and rear of the current unmanned vehicle and the distance between the vehicles based on the time delay difference between the laser sending and receiving of the laser radar devices; S13: If there are no vehicles on both sides of the current unmanned vehicle, determining that the candidate lane change direction is both sides; If there are vehicles on both sides of the current driverless vehicle, determine whether the vehicles on both sides are in the dangerous lane change area. If the vehicles on both sides are in the dangerous lane change area, cancel the lane change request; if the vehicles on both sides are not in the dangerous lane change area, determine the candidate lane change direction as both sides; if only one side of the vehicle is in the dangerous lane change area, determine the candidate lane change direction as the other side; If there is a vehicle driving on only one side of the current unmanned vehicle, it is determined whether the vehicle driving on that side is in a dangerous lane change area. If not in the dangerous area, the candidate lane change direction is determined to be both sides. If in the dangerous area, the other side is used as a candidate lane change direction.

7. The method for intelligent control of unmanned vehicles according to claim 1, characterized in that: The specific target recognition network includes a feature extraction module, a multi-scale feature fusion module, a multi-scale suggestion box generation module, and a target output module; The feature extraction layer includes four convolution modules, an attention module is added between two adjacent convolution modules, and the outputs of the second, third and fourth convolution modules are connected to the multi-scale feature fusion module and the multi-scale suggestion box generation module respectively; The multi-scale suggestion box generation module includes convolution kernels of multiple scales, and the convolution kernels of multiple scales are applied to the feature maps output by the second, third, and fourth convolution modules respectively to obtain position information of suggestion boxes of multiple scales; The multi-scale fusion module inputs the feature map upsampled by the second convolution module, output by the third convolution module, and outputs the fused feature map; The position information of the multiple scale suggestion boxes is used to guide the fusion feature map to obtain multiple suggestion boxes, and the fusion feature map with the multiple suggestion boxes is input into the target output module to obtain the position and category information of the vehicle, license plate and vehicle logo.

8. An intelligent control system based on unmanned vehicles, characterized in that The system includes: a candidate lane change direction determination module, wherein upon receiving an active lane change request, the candidate lane change direction determination module determines a candidate lane change direction through a laser radar device at the side and rear; Candidate lane change direction determination module: when receiving an active lane change request, the candidate lane change direction determination module determines the candidate lane change direction through the laser radar devices on the side and the rear side; Final lane-changing direction determination module: The final lane-changing direction determination module is used to, when the candidate lane-changing direction is one side, use the candidate lane-changing direction determined in the candidate lane-changing direction determination module as the final lane-changing direction; when the candidate lane-changing direction is two sides, turn on the side front cameras on the rearview mirrors on both sides to collect side front video images, and determine the final lane-changing direction based on the side front video analysis results; Determining the final lane change direction based on the side front video analysis result specifically includes: Use the relationship module to get the association score between targets: ； is the balance coefficient used to balance the detection confidence and the associated targets in the previous and next frames. and They represent the confidence of the i-th target in the k-1-th frame and the confidence of the j-th target in the k-th frame, respectively. is the primary ROI proposal box in the k-1th frame The i-th test result, Propose a box for the primary ROI in the kth frame The jth detection result in the video is obtained by iteratively calculating all the association matrices in the video except the first frame, and obtaining the video association matrix , is the number of video frames; Based on the video association matrix M, multiple best association connections are screened out, and the stability score of each target is calculated through the best association connection: ； in, It represents the probability that the detection result of the best associated connection ti in the kth frame belongs to the target i. The number of best associated connections n is equal to the number of targets detected in the last frame. Represents the association score between target i and target j in the kth frame in the association connection mode t.

9. An electronic device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein when the processor executes the computer program, the method for intelligent control of an unmanned vehicle as described in any one of claims 1 to 7 is implemented.

10. A computer-readable storage medium storing a computer program, wherein the computer program, when executed by a processor, implements an intelligent control method for an unmanned vehicle according to any one of claims 1 to 7.

Images