A collision avoidance control method and system in mobile operation

Abstract

The present invention provides a collision avoidance control method and system in mobile operation, which relates to the technical field of vehicle safe driving. The method of the present invention uses the rear camera of the main vehicle to continuously obtain the front image of the target vehicle, and inputs the front image into the trained vehicle recognition algorithm model for recognition; according to the recognition result, it is judged whether the target vehicle meets the collision warning condition and the avoidance condition with the main vehicle respectively. When the avoidance condition is met, it is detected within the calibration time whether the driver control behavior occurs. When the driver control behavior does not occur, the main vehicle avoids according to the avoidance route. When the main vehicle cannot establish communication interconnection with the rear vehicle through V2X technology, the method of the present invention uses the rear camera to obtain the image of the rear vehicle in real time and inputs it into the vehicle recognition algorithm model for recognition, and makes different warning measures according to the recognition result, which can enable the main vehicle to automatically react according to the type of the rear vehicle, thereby avoiding the occurrence of collision.

Description

Technical Field

[0001] One or more embodiments of the present specification relate to the field of vehicle safe driving technology, and more particularly, to a collision avoidance control method and system in mobile operations. Background Art

[0002] Existing intelligent autonomous driving of vehicles is mostly achieved by V2X communication technology. V2X communication can wirelessly exchange information on the speed and position of surrounding vehicles, and is the current development direction of vehicle autonomous driving technology.

[0003] However, in the current development of V2X technology, there is still a problem that only when all vehicles involved in traffic can communicate with each other can perfect information interconnection be achieved. However, for most large vehicles such as trucks, V2X technology is difficult to achieve standardization unless it is required by regulations. This puts trucks and other large vehicles outside the blind spot of V2X interconnection communication.

[0004] In the work of road accident investigation, it is necessary to ensure that there is a certain safe distance between the main vehicle and the rear vehicle. Especially when a traffic accident occurs, the road conditions are complicated. How to keep the main vehicle and the rear vehicle at a safe distance during mobile operations to prevent collisions is a technical problem that needs to be solved urgently by technical personnel in this field. Summary of the invention

[0005] In view of this, the purpose of one or more embodiments of this specification is to provide a collision avoidance control method and system in mobile operations to solve the problem that it is difficult for the main vehicle to automatically avoid being hit according to the type of the rear vehicle.

[0006] Based on the above purpose, one or more embodiments of this specification provide a collision avoidance control method in mobile operation, including:

[0007] Step 1, the system uses the rear camera of the main vehicle to continuously obtain the front image of the target vehicle, and the target vehicle refers to other vehicles located behind the main vehicle; Step 2, the system inputs the front image into the trained vehicle recognition algorithm model for recognition; Step 3, when it is recognized as a Class A vehicle, the current recognition ends; when it is recognized as a Class B vehicle, the system determines whether the Class B vehicle and the main vehicle meet the collision warning condition; Step 4, when it is determined that the Class B vehicle and the main vehicle do not meet the collision warning condition, the current judgment ends; when it is determined that the Class B vehicle and the main vehicle meet the collision warning condition but do not meet the avoidance condition, the system issues a collision warning reminder; Step 5, when it is determined that the Class B vehicle and the main vehicle meet the avoidance condition, the system issues an avoidance reminder and detects whether the driver control behavior occurs within the calibration time; Step 6, when the system detects that the driver control behavior occurs, the avoidance reminder ends; when no driver control behavior is detected, the system actively intervenes and controls the main vehicle to avoid according to the avoidance route.

[0008] Preferably, the training process of the vehicle identification algorithm model includes:

[0009] In the offline training stage, a large number of vehicle front pictures are used as samples y, and the sample y is classified for the first time. The types of the first classification include two types: yellow-plate vehicle m and non-yellow-plate vehicle n; the two types of the classified vehicle front pictures are input into the convolutional neural network, and the convolutional neural network is used to train the first 2-classification algorithm model p; the sample pictures in the non-yellow-plate vehicle n category are classified for the second time, and the types of the second classification include two types: small passenger car s and non-small passenger car t; the sample pictures of the two types of small passenger car s and non-small passenger car t are input into the convolutional neural network, and the convolutional neural network is used to train the second 2-classification algorithm model q; the trained model is the vehicle recognition algorithm model.

[0010] Preferably, in step 2, the system inputs the vehicle head image into a trained vehicle recognition algorithm model for recognition, including:

[0011] The front image is input into the trained first 2 classification algorithm model p for classification; when the probability of the classification result of the front image in the second 2 classification algorithm model p being a yellow-plate vehicle m is the highest, the front image recognition result is set to a Class B vehicle; when the probability of the front image being a non-yellow-plate vehicle n is the highest in the classification result, the front image is further input into the second 2 classification algorithm model q for classification; when the probability of the classification result of the front image in the second 2 classification algorithm model q being a small passenger car s is the highest, the front image recognition result is set to a Class A vehicle, otherwise the front image recognition result is set to a Class B vehicle.

[0012] Preferably, the collision warning conditions include:

[0013] Set a collision threshold L1, and use the rear vehicle sensor to monitor the distance L between the host vehicle and the target vehicle in real time. When L ≤ L1, the collision warning condition is satisfied between the Class B vehicle and the host vehicle; when L > L1, the collision warning condition is not satisfied between the Class B vehicle and the host vehicle.

[0014] The avoidance conditions include:

[0015] Set an avoidance threshold L2, and L2 < L1. Use the rear vehicle sensor to monitor the distance L between the host vehicle and the target vehicle in real time. When L ≤ L2, the avoidance condition is satisfied between the Class B vehicle and the host vehicle; when L2 < L < L1, the collision warning condition is satisfied but the avoidance condition is not satisfied between the Class B vehicle and the host vehicle.

[0016] Preferably, it further includes:

[0017] Obtain the high-precision map information of the target vehicle, and obtain the target characteristics of the target vehicle according to the high-precision map information. The target characteristics include the real-time position information and the real-time vehicle speed of the target vehicle; the system dynamically adjusts the sizes of the collision threshold L1 and the avoidance threshold L2 according to the target characteristics.

[0018] Preferably, the collision warning reminder includes visual, auditory or tactile reminders; the avoidance reminder includes visual, auditory or tactile reminders.

[0019] Preferably, the calibration time is 0.5 s.

[0020] Preferably, it further includes:

[0021] In step 6, while the system detects the occurrence of a driver control behavior, monitor the driver control behavior; the monitoring of the driver control behavior includes:

[0022] Monitor whether the driver makes an action of stepping on the brake pedal. When it is detected that the driver makes an action of stepping on the brake pedal, it is determined that the driver control behavior is improper, and the system actively intervenes and controls the host vehicle to avoid according to the avoidance route.

[0023] Preferably, controlling the host vehicle to avoid according to the avoidance route includes:

[0024] The system uses the front vehicle body sensor of the host vehicle in combination with the high-precision map information to judge whether there are obstacles in the space in front of and on both sides of the host vehicle. The system generates an avoidance route according to the judgment result and controls the power system and steering system of the host vehicle to avoid according to the avoidance route.

[0025] The system generates an avoidance route according to the judgment result, including:

[0026] When there are no obstacles in front of and on the sides of the main vehicle, the planned route planned by the high-precision map information is the avoidance route; when there are obstacles in front of and on the sides of the main vehicle, the system identifies the boundary information of the obstacles on the planned route based on the front and body sensors of the vehicle, and the system generates an avoidance route for avoiding the obstacles based on the high-precision map information and boundary information.

[0027] One or more embodiments of the present specification also provide a collision avoidance control system in mobile operation, including:

[0028] Information acquisition module: used to continuously acquire the front image of the target vehicle using the rear camera of the main vehicle. The target vehicle refers to other vehicles located behind the main vehicle.

[0029] Identification module: used to input the vehicle head image into the trained vehicle identification algorithm model for identification.

[0030] The first judgment module is used to judge whether it is a Class A vehicle or a Class B vehicle.

[0031] The second judgment module is used to judge whether the collision warning condition is satisfied between the Class B vehicle and the host vehicle.

[0032] The third judgment module is used to judge whether the avoidance condition is satisfied between the Class B vehicle and the host vehicle.

[0033] Detection module: used to detect whether driver control behavior occurs within the calibration time.

[0034] The control module is used to end the identification when the first judgment module identifies the vehicle as a Class A vehicle.

[0035] When the second judgment module determines that the collision warning condition between the Class B vehicle and the host vehicle is not satisfied, the judgment is terminated.

[0036] It is used to control the issuance of a collision warning reminder when the second judgment module determines that the collision warning condition is satisfied between the Class B vehicle and the host vehicle and the third judgment module determines that the avoidance condition is not satisfied between the Class B vehicle and the host vehicle.

[0037] It is used to control the issuance of an avoidance reminder when the third judgment module determines that the avoidance condition between the Class B vehicle and the host vehicle is met.

[0038] Used to control the end of the avoidance reminder when the detection module detects that the driver has controlled the behavior.

[0039] It is used to actively intervene and control the main vehicle to avoid according to the avoidance route when the detection module does not detect the driver's control behavior.

[0040] From the above description, it can be seen that one or more embodiments of this specification provide a collision avoidance control method in mobile operation, in which the system uses the rear camera of the main vehicle to continuously obtain the front image of the target vehicle, and inputs the front image into the trained vehicle recognition algorithm model for recognition; when it is recognized as a Class B vehicle, it is determined whether the Class B vehicle and the main vehicle meet the collision warning conditions and avoidance conditions respectively, and when the collision warning conditions are met, a collision warning reminder is issued; when the avoidance conditions are met, an avoidance reminder is issued, and whether the driver control behavior occurs within the calibration time; when the driver control behavior is not detected, the system actively intervenes to control the main vehicle to avoid according to the avoidance route. When the main vehicle cannot establish communication interconnection with the rear vehicle through V2X technology, the method of the present invention uses the rear camera to obtain the image of the rear vehicle in real time and input it into the vehicle recognition algorithm model for recognition, and makes different warning measures based on the recognition results, which can enable the main vehicle to automatically react according to the type of the rear vehicle, thereby avoiding the occurrence of collision. BRIEF DESCRIPTION OF THE DRAWINGS

[0041] In order to more clearly illustrate one or more embodiments of the present specification or the technical solutions in the prior art, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, the drawings described below are only one or more embodiments of the present specification. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.

[0042] Figure 1 The present invention is a flowchart of a collision avoidance control method in mobile operation according to one or more embodiments of the present specification. DETAILED DESCRIPTION

[0043] In order to make the objectives, technical solutions and advantages of the present disclosure more clearly understood, the present disclosure is further described in detail below in combination with specific embodiments and with reference to the accompanying drawings.

[0044] It should be noted that, unless otherwise defined, the technical terms or scientific terms used in one or more embodiments of this specification should be understood by people with ordinary skills in the field to which this disclosure belongs. The "first", "second" and similar words used in one or more embodiments of this specification do not indicate any order, quantity or importance, but are only used to distinguish different components. "Include" or "comprise" and similar words mean that the elements or objects appearing before the word cover the elements or objects listed after the word and their equivalents, without excluding other elements or objects. "Connect" or "connected" and similar words are not limited to physical or mechanical connections, but may include electrical connections, whether direct or indirect. "Up", "down", "left", "right" and the like are only used to indicate relative positional relationships. When the absolute position of the described object changes, the relative positional relationship may also change accordingly.

[0045] like Figure 1 This is a flowchart of a method for anti-collision control in mobile operation according to one or more embodiments of this specification. This specification provides an embodiment of a method for anti-collision control in mobile operation, including:

[0046] Step 1, the system uses the rear camera of the main vehicle to continuously obtain the front image of the target vehicle, and the target vehicle refers to other vehicles located behind the main vehicle; Step 2, the system inputs the front image into the trained vehicle recognition algorithm model for recognition; Step 3, when it is recognized as a Class A vehicle, the current recognition ends; when it is recognized as a Class B vehicle, the system determines whether the Class B vehicle and the main vehicle meet the collision warning condition; Step 4, when it is determined that the Class B vehicle and the main vehicle do not meet the collision warning condition, the current judgment ends; when it is determined that the Class B vehicle and the main vehicle meet the collision warning condition but do not meet the avoidance condition, the system issues a collision warning reminder; Step 5, when it is determined that the Class B vehicle and the main vehicle meet the avoidance condition, the system issues an avoidance reminder and detects whether the driver control behavior occurs within the calibration time; Step 6, when the system detects that the driver control behavior occurs, the avoidance reminder ends; when no driver control behavior is detected, the system actively intervenes and controls the main vehicle to avoid according to the avoidance route.

[0047] The method provided in this embodiment uses a rear camera to obtain real-time images of the rear vehicle and inputs them into a vehicle recognition algorithm model for recognition when the main vehicle is unable to establish communication with the rear vehicle through V2X technology. Different early warning measures are taken based on the recognition results, so that the main vehicle can automatically respond according to the type of rear vehicle, thereby avoiding collisions.

[0048] Another embodiment provided in this specification is a collision avoidance control method in mobile operation, comprising:

[0049] In step 1, the system uses the rear camera of the main vehicle to continuously obtain the front image of the target vehicle, where the target vehicle refers to other vehicles located behind the main vehicle; in step 2, the system inputs the front image into the trained vehicle recognition algorithm model for recognition.

[0050] The training process of the vehicle identification algorithm model includes:

[0051] In the offline training phase, a large number of vehicle head images are used as sample y. Before inputting sample y for training, the vehicle head images of sample y are preprocessed, including adjusting the vehicle head images to the same size, normalizing pixel values, etc. The sample y is classified for the first time, and the types of the first classification include 2 types: yellow-plate vehicle m and non-yellow-plate vehicle n; the 2 types of the classified vehicle head images are input into the convolutional neural network, and the convolutional neural network is used to train the first 2-classification algorithm model p; the sample images in the non-yellow-plate vehicle n classification are classified for the second time, and the types of the second classification include 2 types: small passenger car s and non-small passenger car t; the sample images of the two types of small passenger car s and non-small passenger car t are input into the convolutional neural network, and the convolutional neural network is used to train the second 2-classification algorithm model q; the trained model is the vehicle recognition algorithm model. After the offline training is completed, the offline trained vehicle recognition algorithm model is tested using the prepared test set to evaluate the accuracy of the trained model.

[0052] Among them, the massive amount of car front pictures refers to car front pictures that have been disassembled more than 1,000 times.

[0053] The training process of the vehicle recognition algorithm model also includes: an online training phase, using V2X to obtain the front image of the target vehicle taken by other main vehicles as a sample y', inputting it into the vehicle recognition algorithm model for training, and continuously optimizing the vehicle recognition algorithm model.

[0054] Vehicles that can cause serious damage to the main vehicle are generally yellow-plate vehicles such as large trucks, large buses, and large trucks, and blue-plate vehicles such as small trucks. Therefore, when classifying and screening the front images of the vehicles, yellow-plate vehicles and non-yellow-plate vehicles are first classified, and then small buses are found from the classification of non-yellow-plate vehicles. This method of performing two classifications not only reduces the difficulty of calculations, but also improves the accuracy of front image recognition.

[0055] In step 2, the system inputs the vehicle head image into the trained vehicle recognition algorithm model for recognition, including:

[0056] Input the front - end vehicle image into the trained first binary classification algorithm model \(p\) for classification. When the probability that the front - end vehicle image is a yellow - card vehicle \(m\) is the highest in the second binary classification algorithm model \(p\), set the recognition result of the front - end vehicle image as type - B vehicle. When the probability that the front - end vehicle image is a non - yellow - card vehicle \(n\) is the highest in the classification result, continue to input the front - end vehicle image into the second binary classification algorithm model \(q\) for classification. When the probability that the front - end vehicle image is a small passenger car \(s\) is the highest in the second binary classification algorithm model \(q\), set the recognition result of the front - end vehicle image as type - A vehicle; otherwise, set the recognition result of the front - end vehicle image as type - B vehicle.

[0057] The type - A vehicles include: small passenger cars such as minivans, sedans, off - road vehicles, SUVs, MPVs, etc., mainly for carrying people. The type - B vehicles include: medium - and large - sized vehicles such as large buses, large trucks, muck trucks, heavy trucks, medium - sized buses, etc. and small trucks, etc.

[0058] Step 3, when it is recognized as a type - A vehicle, the current recognition ends. When it is recognized as a type - B vehicle, the system determines whether the type - B vehicle and the host vehicle meet the collision warning conditions.

[0059] The collision warning conditions include:

[0060] Set a collision threshold \(L1\), and use the rear - end sensor to monitor the distance \(L\) between the host vehicle and the target vehicle in real - time. When \(L\leq L1\), the type - B vehicle and the host vehicle meet the collision warning conditions. When \(L > L1\), the type - B vehicle and the host vehicle do not meet the collision warning conditions.

[0061] Step 4, when it is determined that the type - B vehicle and the host vehicle do not meet the collision warning conditions, the current judgment ends. When it is determined that the type - B vehicle and the host vehicle meet the collision warning conditions but do not meet the avoidance conditions, the system issues a collision warning reminder. The collision warning reminder includes visual, auditory, or tactile reminders.

[0062] The avoidance conditions include:

[0063] Set an avoidance threshold \(L2\), and \(L2<L1\). Use the rear - end sensor to monitor the distance \(L\) between the host vehicle and the target vehicle in real - time. When \(L\leq L2\), the type - B vehicle and the host vehicle meet the avoidance conditions. When \(L2 < L < L1\), the type - B vehicle and the host vehicle meet the collision warning conditions but do not meet the avoidance conditions.

[0064] The system can obtain high-precision map information of the target vehicle, and obtain target features of the target vehicle based on the high-precision map information, wherein the target features include data such as the real-time position information and real-time vehicle speed of the target vehicle; the system dynamically adjusts the size of the collision threshold L1 and the avoidance threshold L2 based on the target features.

[0065] For example, the collision threshold L1 is set to 10m, and the avoidance threshold L2 is set to 1m. Of course, both can be set manually by the driver.

[0066] The system can also set the collision threshold L1 to be divided into a1, a2, and a3 levels, for example, a1 is 20m, a2 is 10m, a3 is 5m, and the system default level is a2; the system can also set the avoidance threshold L2 to be divided into b1, b2, and b3 levels, for example, b1 is 5m, b2 is 1m, and b3 is 0.5m, and the system default level is b2; it can also be adjusted manually. Generally speaking, the faster the speed of the rear vehicle, the larger the corresponding threshold setting. Generally, on highways, the speed of large trucks can reach about 100km / h. At this time, even considering the relative driving speed and the braking deceleration of large trucks before stopping, the collision threshold between the main vehicle and it should be greater than 30m; on urban roads, the speed of large trucks is generally about 60km / h. At this time, considering the relative driving speed and the braking deceleration of large trucks before stopping, the collision threshold setting between the main vehicle and it can be reduced accordingly, for example, 2m, and the avoidance threshold is the same.

[0067] The rear camera can be Mobileye, and the rear sensor can be a millimeter-wave radar. The system can obtain the target features of the target vehicle based on the high-precision map information combined with the rear camera and the rear sensor. Both Mobileye and millimeter-wave radar have ID tracking functions, and the Kalman weighted fusion algorithm can be used to perform weighted summation on the target features to obtain more accurate target features.

[0068] In actual vehicle driving, often only the vehicle behind the main vehicle in the same lane will pose a danger to the main vehicle, so the rear sensor only detects vehicles behind the main vehicle in the same lane.

[0069] Step 5: When it is determined that the avoidance conditions between the Class B vehicle and the host vehicle are met, the system issues an avoidance reminder, which includes visual, auditory or tactile reminders; and detects whether the driver's control behavior occurs within the calibration time. The driver's minimum reaction time before starting the operation is 0.4 seconds, so the calibration time can be 0.5s. Of course, the calibration time can also be increased or decreased as needed. Step 6: When the system detects that the driver's control behavior occurs, the avoidance reminder ends; when no driver control behavior is detected, the system actively intervenes and controls the host vehicle to avoid according to the avoidance route.

[0070] In step 6, the system detects the occurrence of driver control behavior and monitors the driver control behavior; the monitoring of the driver control behavior includes:

[0071] Monitor whether the driver steps on the brake pedal. When it is detected that the driver steps on the brake pedal, it is judged that the driver's control behavior is inappropriate. The system actively intervenes and controls the main vehicle to avoid the vehicle according to the avoidance route.

[0072] Wherein controlling the main vehicle to avoid according to the avoidance route comprises:

[0073] The system uses the main vehicle's front body sensors combined with high-precision map information to determine whether there are obstacles in the space in front of and on both sides of the main vehicle. The system generates an avoidance route based on the judgment results, controls the main vehicle's power system and steering system, controls the front wheel angle of the car, and avoids according to the avoidance route.

[0074] The system generates an avoidance route based on the judgment result, including:

[0075] When there are no obstacles in front of and on the sides of the main vehicle, the planned route planned by the high-precision map information is the avoidance route; when there are obstacles in front of and on the sides of the main vehicle, the system identifies the boundary information of the obstacles on the planned route based on the front and body sensors of the vehicle, and the system generates an avoidance route for avoiding the obstacles based on the high-precision map information and boundary information.

[0076] There are many ways for the system to identify obstacles and generate avoidance routes. For example, one of them is to use tools such as OpenCV to generate boundary information of obstacles on the planned route, and generate a minimum bounding rectangle of the obstacle based on the boundary information; generate an equation for an elliptical obstacle boundary based on the minimum bounding rectangle and the unit safety distance; obtain the entry and exit intersection points formed when the planned route passes through the elliptical obstacle boundary, and determine the local curve segments corresponding to the entry and exit intersection points on the elliptical obstacle boundary; calculate the cut-out angles that match the entry and exit intersection points respectively based on the entry and exit intersection points, the planned route and the elliptical obstacle boundary; generate an avoidance route for avoiding the target obstacle based on the entry and exit intersection points, the cut-out angles that match the entry and exit intersection points respectively, and the local curve segments.

[0077] The method provided in this embodiment uses a rear camera to obtain images of the rear vehicle in real time and inputs them into a vehicle recognition algorithm model for recognition when the main vehicle cannot establish communication with the rear vehicle through V2X technology. Secondary classification is performed through a convolutional neural network image classification algorithm, with high recognition accuracy and fast recognition speed. This allows small vehicles to recognize possible collision hazards when the rear vehicle is a large vehicle during road driving or parking, thereby quickly taking different warning measures and responding, thereby avoiding collisions and protecting the personal safety of drivers and passengers of small vehicles.

[0078] One or more embodiments of the present specification also provide a collision avoidance control system in mobile operation, including:

[0079] Information acquisition module: used to continuously acquire the front image of the target vehicle using the rear camera of the main vehicle. The target vehicle refers to other vehicles located behind the main vehicle.

[0080] Identification module: used to input the vehicle head image into the trained vehicle identification algorithm model for identification.

[0081] The first judgment module is used to judge whether it is a Class A vehicle or a Class B vehicle.

[0082] The second judgment module is used to judge whether the collision warning condition is satisfied between the Class B vehicle and the host vehicle.

[0083] The third judgment module is used to judge whether the avoidance condition is satisfied between the Class B vehicle and the host vehicle.

[0084] Detection module: used to detect whether driver control behavior occurs within the calibration time.

[0085] The control module is used to end the identification when the first judgment module identifies the vehicle as a Class A vehicle.

[0086] When the second judgment module determines that the collision warning condition between the Class B vehicle and the host vehicle is not satisfied, the judgment is terminated.

[0087] It is used to control the issuance of a collision warning reminder when the second judgment module determines that the collision warning condition is satisfied between the Class B vehicle and the host vehicle and the third judgment module determines that the avoidance condition is not satisfied between the Class B vehicle and the host vehicle.

[0088] It is used to control the issuance of an avoidance reminder when the third judgment module determines that the avoidance condition between the Class B vehicle and the host vehicle is met.

[0089] Used to control the end of the avoidance reminder when the detection module detects that the driver has controlled the behavior.

[0090] It is used to actively intervene and control the main vehicle to avoid according to the avoidance route when the detection module does not detect the driver's control behavior.

[0091] The system provided in this embodiment can achieve the same effect as the method of the present invention. When the main vehicle cannot establish communication with the rear vehicle through V2X technology, the information acquisition module is used to obtain the image of the rear vehicle in real time and input it into the recognition module for recognition. The first to third judgment modules make judgments based on the recognition results, and the detection module makes detections based on the judgment results. Different early warning measures are then implemented through the control module, which can enable the main vehicle to automatically respond according to the type of the rear vehicle, thereby avoiding collisions.

[0092] It can be understood that the present invention can be executed by any device, equipment, platform, or device cluster with computing and processing capabilities, and produce beneficial effects corresponding to the method and system.

[0093] It should also be noted that, as the number of new energy green-license plate vehicles gradually increases, it is also achievable to add the result of green license plate recognition to the vehicle recognition algorithm model using the method and system described in the present invention.

[0094] Those skilled in the art should understand that the discussion of any of the above embodiments is merely illustrative and is not intended to imply that the scope of the present disclosure (including the claims) is limited to these examples. In line with the present disclosure, the technical features in the above embodiments or different embodiments may be combined, the steps may be implemented in any order, and there are many other variations of different aspects of one or more embodiments of the present specification as described above, which are not provided in detail for the sake of simplicity.

[0095] One or more embodiments of this specification are intended to cover all such substitutions, modifications and variations that fall within the broad scope of the appended claims. Therefore, any omissions, modifications, equivalent substitutions, improvements, etc. made within the spirit and principles of one or more embodiments of this specification should be included in the scope of protection of this disclosure.

Claims

1. A collision avoidance control method in mobile operation, characterized in that: include: Step 1: The system uses the rear camera of the main vehicle to continuously obtain the front image of the target vehicle, where the target vehicle refers to other vehicles located behind the main vehicle; Step 2, the system inputs the vehicle head image into the trained vehicle recognition algorithm model for recognition; Step 3: If the vehicle is identified as a Class A vehicle, the identification is completed; When it is identified as a Class B vehicle, the system determines whether the Class B vehicle and the host vehicle meet the collision warning conditions; Step 4: when it is determined that the Class B vehicle and the host vehicle do not meet the collision warning condition, this determination ends; When it is determined that the Class B vehicle and the host vehicle meet the collision warning condition but do not meet the avoidance condition, the system issues a collision warning reminder; Step 5, when it is determined that the avoidance condition between the Class B vehicle and the host vehicle is met, the system issues an avoidance reminder and detects whether the driver's control behavior occurs within a calibrated time; Step 6: When the system detects that the driver has taken control action, the avoidance reminder is terminated; When no driver control behavior is detected, the system actively intervenes to control the main vehicle to avoid the obstacle according to the avoidance route.

2. The anti-collision control method in mobile operation according to claim 1, characterized in that: The training process of the vehicle identification algorithm model includes: In the offline training phase, a large number of vehicle head images are used as samples y, and the samples y are classified for the first time. The types of the first classification include two types: yellow-card vehicle m and non-yellow-card vehicle n; The two types in the classified car head pictures are input into the convolutional neural network, and the convolutional neural network is used to train the first two-classification algorithm model p; Performing a second classification on the sample images in the non-yellow plate vehicle n classification, wherein the second classification includes two types: small passenger car s and non-small passenger car t; Input sample images of two types, small passenger car s and non-small passenger car t, into the convolutional neural network, and use the convolutional neural network to train the second 2-classification algorithm model q; The trained model is the vehicle recognition algorithm model.

3. The anti-collision control method in mobile operation according to claim 2, characterized in that: In step 2, the system inputs the vehicle head image into the trained vehicle recognition algorithm model for recognition, including: Input the vehicle front image into the trained first 2 classification algorithm model p for classification; when the classification result of the vehicle front image in the second 2 classification algorithm model p is the highest probability of yellow plate vehicle m, the vehicle front image recognition result is set as a Class B vehicle; When the probability that the front image of the vehicle is a non-yellow plate vehicle n is the highest in the classification result, the front image of the vehicle is further input into the second 2-classification algorithm model q for classification; When the probability that the classification result of the front image in the second classification algorithm model q is a small passenger car s is the largest, the recognition result of the front image is set as a Class A vehicle, otherwise the recognition result of the front image is set as a Class B vehicle.

4. The anti-collision control method in mobile operation according to claim 1, characterized in that: The collision warning conditions include: A collision threshold L1 is set, and the distance L between the main vehicle and the target vehicle is monitored in real time using the rear sensor. When L≤L1, the distance between the Class B vehicle and the main vehicle meets the collision warning condition; when L>L1, the distance between the Class B vehicle and the main vehicle does not meet the collision warning condition; The avoidance conditions include: Set an avoidance threshold L2, and L2 < L1. Use the rear vehicle sensor to monitor the distance L between the host vehicle and the target vehicle in real time. When L ≤ L2, the avoidance condition is satisfied between the Class B vehicle and the host vehicle; when L2 < L < L1, the collision warning condition is satisfied between the Class B vehicle and the host vehicle, but the avoidance condition is not satisfied.

5. The anti-collision control method in mobile operation according to claim 4, characterized in that: It further includes: Obtain the high-precision map information of the target vehicle, and obtain the target features of the target vehicle according to the high-precision map information. The target features include the real-time position information and the real-time vehicle speed of the target vehicle. The system dynamically adjusts the sizes of the collision threshold L1 and the avoidance threshold L2 according to the target features.

6. The anti-collision control method in mobile operation according to claim 1, characterized in that: The collision warning reminder includes visual, auditory or tactile reminders; the avoidance reminder includes visual, auditory or tactile reminders.

7. The anti-collision control method in mobile operation according to claim 1, characterized in that: The calibration time is 0.5s.

8. The anti-collision control method in mobile operation according to claim 1, characterized in that: It further includes: In step 6, while the system detects the occurrence of the driver's control behavior, monitor the driver's control behavior; The monitoring of the driver's control behavior includes: Monitor whether the driver makes an action of stepping on the brake pedal. When it is detected that the driver makes an action of stepping on the brake pedal, it is determined that the driver's control behavior is improper, and the system actively intervenes to control the host vehicle to avoid according to the avoidance route.

9. The anti-collision control method in mobile operation according to claim 1, characterized in that: Controlling the host vehicle to avoid according to the avoidance route includes: The system uses the front vehicle body sensor of the host vehicle combined with the high-precision map information to judge whether there are obstacles in the space in front of and on both sides of the host vehicle. The system generates an avoidance route according to the judgment result, and controls the power system and the steering system of the host vehicle to avoid according to the avoidance route; The system generating an avoidance route according to the judgment result includes: When there are no obstacles in front of and on the side of the host vehicle, the planned route planned by the high-precision map information is the avoidance route; when there are obstacles in front of and on the side of the host vehicle, the system identifies the boundary information of the obstacles on the planned route according to the front vehicle body sensor, and the system generates an avoidance route for avoiding obstacles according to the high-precision map information and the boundary information.

10. An anti-collision control system in mobile operation, characterized in that: It includes: Information acquisition module: used to continuously acquire the front vehicle pictures of the target vehicle by using the rear camera of the host vehicle. The target vehicle refers to other vehicles located behind the host vehicle; Recognition module: used to input the front vehicle pictures into the trained vehicle recognition algorithm model for recognition; First judgment module: used to judge whether it is a Class A vehicle or a Class B vehicle; Second judgment module, used to judge whether the collision warning condition is satisfied between the Class B vehicle and the host vehicle; Third judgment module, used to judge whether the avoidance condition is satisfied between the Class B vehicle and the host vehicle; Detection module: used to detect whether the driver's control behavior occurs within the calibration time; Control module, used to end the recognition when the first judgment module recognizes a Class A vehicle; Used to end the judgment when the second judgment module judges that the collision warning condition is not satisfied between the Class B vehicle and the host vehicle; When the second judgment module determines that the collision warning condition between the Class B vehicle and the host vehicle is satisfied and the third judgment module determines that the avoidance condition between the Class B vehicle and the host vehicle is not satisfied, a collision warning reminder is issued; When the third judgment module determines that the avoidance condition between the Class B vehicle and the host vehicle is met, control the issuance of an avoidance reminder; Used to control the end of the avoidance reminder when the detection module detects that the driver has controlled the behavior; It is used to actively intervene and control the main vehicle to avoid according to the avoidance route when the detection module does not detect the driver's control behavior.

Images