Auxiliary docking device and method for semitrailer

Through the technology combined with binocular vision camera and inertial measurement unit, the problem of low docking efficiency between semi-trailer and tractor is solved, the precise position determination and adjustment of the traction pin is realized, and the accuracy and efficiency of docking are improved.

CN118913257AInactive Publication Date: 2024-11-08LONGYAN CHANGFENG SPECIAL VEHICLE CO LTD
View PDF 0 Cites 0 Cited by

Patent Information

Application Number
CN202410805899.8
Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
Filing Date
2024-06-21
Publication Date
2024-11-08
Estimated Expiration
Not applicable · inactive patent

AI Technical Summary

Technical Problem

In the prior art, the auxiliary docking efficiency between the semi-trailer and the tractor is low, mainly due to the subjective factors in determining the center position of the traction saddle and the traction pin, which leads to the docking process being inconsistent and efficient.

Method used

Using binocular vision camera and inertial measurement unit combined with the docking analysis module, the traction pin image and docking position data are obtained in real time, and the initial and real-time traction pin detection model is constructed to realize the precise position determination and adjustment of the traction pin, ensuring the semi-trailer and traction truck accurate docking.

Benefits of technology

The docking efficiency between semi-trailer and tractor is improved, a more accurate and efficient docking process is achieved, man-made errors are reduced, and the accuracy and stability of docking are improved.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN118913257A_ABST
    Figure CN118913257A_ABST
Patent Text Reader

Abstract

The invention discloses an auxiliary docking device and method for a semitrailer, and relates to the technical field of intelligent driving vehicle connecting pieces. The auxiliary docking device for the semitrailer comprises a binocular vision camera, an inertial measurement unit and a docking analysis module. According to the invention, the initial traction pin detection model is constructed through the obtained first traction pin image, and then the model comprehensive coefficient is obtained through the corresponding model detection coefficient and the model field coefficient so as to determine the first traction pin detection model. Inputting the obtained initial traction pin image into a first traction pin detection model to obtain initial position data of the traction pin, obtaining real-time position data of the traction pin in combination with the real-time traction pin image, finally judging whether the real-time position data of the traction pin meet requirements or not, if yes, indicating that auxiliary docking of the semitrailer is finished, and if not, indicating that auxiliary docking of the semitrailer is finished. Otherwise, the position of the specified semitrailer is adjusted until the requirement is met, the auxiliary butt joint efficiency of the semitrailer is improved, and the problem that in the prior art, auxiliary butt joint of the semitrailer is not efficient is solved.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The present invention relates to the technical field of intelligent driving vehicle connectors, and in particular to a device and method for auxiliary docking of a semitrailer. Background Art

[0002] A semi-trailer is a common truck that consists of a front part connected to a tractor and a rear part that can move partially independently. Semi-trailers are widely used, mainly to transport goods by hanging and dropping, such as delivering goods to markets, factories, warehouses and other places. They can carry a large amount of goods and travel on highways. The auxiliary docking of semi-trailers refers to some technologies and equipment used when connecting semi-trailers and tractors, such as auxiliary guidance systems and automatic auxiliary docking systems, or auxiliary personnel participate in guidance to ensure the safety and accuracy of the connection. The auxiliary guidance system can help the driver accurately align the connection point between the semi-trailer and the tractor through sensors and cameras, specifically by providing real-time image or sound guidance to help the driver dock the two together. The automatic auxiliary docking system can automatically dock the semi-trailer and the tractor according to pre-set parameters, which usually relies on sensors, cameras and motors to automatically adjust the connection position.

[0003] In the prior art, in order to achieve effective connection between the semi-trailer and the tractor, the driver mainly drives the tractor in reverse so that the center position of the traction saddle approaches the center position of the semi-trailer's traction pin, and the effective connection between the tractor and the semi-trailer is achieved through the effective docking of the traction saddle and the traction pin.

[0004] For example, the patent application with publication number: CN115431685A discloses an automatic docking device, docking method and unmanned docking system for a semi-trailer, including a radar device, an unmanned driving controller, a safety pin controller and a semi-trailer support lifter, wherein the radar device, the safety pin controller and the semi-trailer support lifter are all electrically connected to the unmanned driving controller; wherein the radar device is arranged on a tractor, and the radar device is configured to detect point cloud data of a semi-trailer to be docked; the unmanned driving controller is configured to receive the point cloud data detected by the radar device to calculate a reversing route of the tractor; and according to the reversing route, determine the steering angle, power drive amount and braking amount of the tractor; the safety pin controller is configured to drive the safety pin to be inserted into the pin hole in response to a first instruction of the unmanned driving controller, wherein one of the safety pin and the pin hole is arranged on the tractor, and the other is arranged on the semi-trailer; the semi-trailer support lifter is configured to drive the semi-trailer support to lift in response to a second instruction of the unmanned driving controller.

[0005] For example, the patent application with publication number: CN116442692A discloses a tractor docking assistance method, device, equipment and storage medium, including: a docking assistance controller is arranged on the tractor, and at least one camera is arranged on the tractor; the tractor docking assistance method includes: if there is a trailer to be docked in the camera field of view of the camera, then the vehicle image captured by the camera is obtained; the trailer characteristic size of the trailer to be docked in the vehicle image is determined; the relative position between the tractor and the trailer to be docked is determined according to the trailer characteristic size; docking assistance information is output according to the relative position, and the docking assistance information is used to instruct the tractor to dock with the trailer to be docked.

[0006] However, in the process of implementing the technical solution of the invention in the embodiments of the present application, the present application found that the above technology has at least the following technical problems:

[0007] In the prior art, the effective connection between the semi-trailer and the tractor is to determine the center position of the traction saddle and the traction pin. Regardless of whether it depends on the driver's experience or pre-set parameters, the intervention of subjective factors has resulted in no unified rules to follow in the docking process, and there is a problem of inefficient semi-trailer assisted docking. Summary of the invention

[0008] The embodiments of the present application solve the problem of inefficient assisted docking of semi-trailers in the prior art by providing a device and method for assisted docking of semi-trailers, thereby improving the efficiency of assisted docking of semi-trailers.

[0009] An embodiment of the present application provides a device for auxiliary docking of semi-trailers, including a binocular vision camera, an inertial measurement unit and a docking analysis module, wherein the binocular vision camera is communicatively connected to the docking analysis module, and the inertial measurement unit is communicatively connected to the docking analysis module; wherein the binocular vision camera is used to obtain a real-time traction pin image during the docking process of a designated semi-trailer, and transmit the image to the docking analysis module; the inertial measurement unit is used to obtain real-time docking position data of a designated tractor and a designated semi-trailer, and transmit the data to the docking analysis module; the docking analysis module is used to obtain real-time position data of the traction pin based on the received real-time traction pin image and real-time docking position data, thereby achieving docking of the designated semi-trailer and the designated tractor.

[0010] The embodiment of the present application provides a method for auxiliary docking of semi-trailers, which is applied to the device for auxiliary docking of semi-trailers, and includes the following steps: acquiring a first traction pin image through a binocular vision camera, and constructing a first traction pin detection model based on the acquired first traction pin image; acquiring an initial traction pin image through a binocular vision camera before docking with a designated semi-trailer, and inputting it into the constructed first traction pin detection model for traction pin identification, and obtaining traction pin initial position data in combination with initial docking position data acquired by an inertial measurement unit; acquiring a real-time traction pin image through a binocular vision camera during docking with the designated semi-trailer, and inputting it into the constructed first traction pin detection model for traction pin identification, and obtaining traction pin real-time position data in combination with real-time docking position data acquired by the inertial measurement unit; judging whether the obtained traction pin real-time position data meets the requirements, and if so, it indicates that the docking of the designated tractor and the designated semi-trailer is completed, otherwise, the position of the designated semi-trailer is adjusted according to the traction pin real-time position data until the requirements are met.

[0011] Furthermore, the specific method for constructing the first traction pin detection model is as follows: constructing an identification sample data set based on the collected first traction pin image, and dividing the identification sample data set into a training set, a validation set and a test set according to a preset ratio; inputting the manually annotated training set into the to-be-selected detection model for model training, and verifying and optimizing the model training result through the validation set to obtain an initial traction pin detection model; performing a first test on the initial traction pin detection model through the test set to obtain a test time and a test result, and thereby obtaining a model detection coefficient of the initial traction pin detection model, and the model detection coefficient is used to numerically describe the performance of the initial traction pin detection model in the test set; performing a second test on the initial traction pin detection model through a field docking experiment to obtain a field docking time, and thereby obtaining a model field coefficient of the initial traction pin detection model, and the model field coefficient is used to numerically describe the performance of the initial traction pin detection model in the field docking experiment; combining the model detection coefficient and the model field coefficient of the initial traction pin detection model to obtain a corresponding model comprehensive coefficient, and screening the initial traction pin detection model according to the model comprehensive coefficient to obtain a first traction pin detection model, and the model comprehensive coefficient is used to comprehensively describe the performance of the initial traction pin detection model.

[0012] Furthermore, the comprehensive coefficient of the model is calculated using the following formula: ;

[0013] Where n is the number of the initial traction pin detection model, , N is the total number of initial traction pin detection models, is the comprehensive model coefficient of the nth initial traction pin detection model, is the model detection coefficient of the nth initial traction pin detection model, is the left model detection coefficient of the initial traction pin detection model, is the right model detection coefficient of the initial traction pin detection model, is the model field coefficient of the nth initial traction pin detection model, is the left model field coefficient of the initial traction pin detection model, is the right model field coefficient of the initial traction pin detection model, is the detection weight, is the field weight, is a logarithmic factor, and .

[0014] Furthermore, the specific method for obtaining the model field coefficient is as follows: obtaining the field docking time by performing a field docking experiment in the experimental area by a preset driver within a preset time period; obtaining the total docking time of the preset driver by performing statistics on the field docking time through an integral operation; sequentially sorting the field docking time of the preset driver to obtain an empirical docking sequence and a model docking sequence, thereby obtaining the empirical docking extreme value and the model docking extreme value, and combining the obtained total docking time to obtain the model field coefficient of the initial traction pin detection model, and the model field coefficient is calculated using the following formula: ;

[0015] Where n is the number of the initial traction pin detection model, , N is the total number of initial traction pin detection models, is the model field coefficient of the nth initial traction pin detection model, q is the preset driver number, , Q is the total number of preset drivers, x is the number of times of experience docking experiment, , A is the total number of empirical docking experiments, y is the number of field docking experiments, , B is the total number of field docking experiments, represents the experience docking time function of the qth preset driver in the xth experience docking experiment, represents the model docking time function of the qth preset driver in the yth model docking experiment through the nth initial tow pin detection model, is the minimum value of the experience docking of the qth preset driver, is the maximum value of the experience connection of the qth preset driver, is the model docking minimum value corresponding to the model docking experiment performed by the qth preset driver through the nth initial traction pin detection model, It is the model docking maximum value corresponding to the model docking experiment performed by the qth preset driver through the nth initial traction pin detection model.

[0016] One or more technical solutions provided in the embodiments of the present application have at least the following technical effects or advantages:

[0017] 1. An initial tow pin detection model is constructed by acquiring the first tow pin image, and then the model comprehensive coefficient is obtained by the corresponding model detection coefficient and the model field coefficient to determine the first tow pin detection model. Subsequently, the acquired initial tow pin image is input into the constructed first tow pin detection model to obtain the tow pin initial position data. Finally, the tow pin real-time position data is obtained in combination with the real-time tow pin image to achieve the docking of the designated semi-trailer with the designated tractor, thereby achieving the precise docking of the designated semi-trailer, and then achieving the improvement of the efficiency of the semi-trailer auxiliary docking, which effectively solves the problem of inefficient semi-trailer auxiliary docking in the prior art.

[0018] 2. An initial tow pin detection model is constructed through the collected first tow pin image, and then the corresponding model detection coefficient and model field coefficient are obtained through the test set and the field docking experiment respectively, and then the model comprehensive coefficient is obtained. Finally, the first tow pin detection model is determined according to the model comprehensive coefficient, thereby realizing the numerical description of the performance of the initial tow pin detection model, and then realizing the acquisition of a more accurate first tow pin detection model to realize the auxiliary docking of the semi-trailer.

[0019] 3. By statistically analyzing the test time of the initial traction pin detection model, the corresponding extreme test time difference and test time coefficient are obtained. Then, the recognition accuracy of the traction pin is statistically analyzed to obtain the corresponding extreme test accuracy difference and test accuracy coefficient. Next, the intersection and union ratio of the traction pin is statistically analyzed to obtain the corresponding extreme test intersection and union ratio difference and test intersection and union ratio coefficient. Finally, the model detection coefficient of the initial traction pin detection model is obtained based on the obtained data, thereby realizing a numerical description of the performance of the initial traction pin detection model in the test set, and then realizing a more accurate description of the performance of the initial traction pin detection model in the test set. BRIEF DESCRIPTION OF THE DRAWINGS

[0020] Figure 1 A schematic diagram of the structure of a device for auxiliary docking of a semi-trailer provided in an embodiment of the present application;

[0021] Figure 2 A schematic diagram of a docking scenario provided in an embodiment of the present application;

[0022] Figure 3 A graph showing changes in the comprehensive coefficients of the model provided in the embodiments of the present application. DETAILED DESCRIPTION

[0023] The embodiment of the present application solves the problem of inefficiency of assisted docking of semi-trailers in the prior art by providing a device and method for assisted docking of semi-trailers. The experimental environment is designed by determining the experimental environment factors and the experimental environment range to obtain a first traction pin image, and an initial traction pin detection model is constructed according to the first traction pin image. The initial traction pin detection model is then tested by a test set and an experimental docking experiment to obtain the corresponding model detection coefficient and model field coefficient. The initial traction pin detection model is screened by the model comprehensive coefficient determined by the model detection coefficient and the model field coefficient to obtain the first traction pin detection model. Then, the designated semi-trailer is subjected to binocular vision before docking. The initial towing pin image obtained by the machine is input into the constructed first towing pin detection model for towing pin recognition. At the same time, the initial docking position data obtained by the inertial measurement unit is combined to obtain the initial position data of the towing pin. Then, the real-time towing pin image obtained by the binocular vision camera during the docking process of the designated semi-trailer is input into the constructed first towing pin detection model for towing pin recognition to obtain the real-time position data of the towing pin. Finally, it is judged whether the obtained real-time position data of the towing pin meets the requirements. If so, it indicates that the docking of the designated tractor and the designated semi-trailer is completed. Otherwise, the position of the designated semi-trailer is adjusted according to the real-time position data of the towing pin until the requirements are met, thereby improving the efficiency of the auxiliary docking of the semi-trailer.

[0024] The technical solution in the embodiment of the present application is to solve the problem of inefficient auxiliary docking of the above-mentioned semi-trailer, and the overall idea is as follows:

[0025] An initial towing pin detection model is constructed by acquiring the first towing pin image, and then the model comprehensive coefficient is obtained by the corresponding model detection coefficient and the model field coefficient to determine the first towing pin detection model. Subsequently, the initial towing pin image acquired by the binocular vision camera is input into the constructed first towing pin detection model to obtain the towing pin initial position data, and then the towing pin real-time position data is obtained based on the real-time towing pin image and the constructed first towing pin detection model. Finally, it is determined whether the obtained real-time position data of the towing pin meets the requirements. If so, it indicates that the semi-trailer assisted docking is completed. Otherwise, the position of the designated semi-trailer is adjusted until the requirements are met, thereby achieving the effect of improving the efficiency of the semi-trailer assisted docking.

[0026] In order to better understand the above technical solution, the above technical solution will be described in detail below in conjunction with the accompanying drawings and specific implementation methods.

[0027] like Figure 1 FIG. 1 is a schematic diagram of a structure of a device for auxiliary docking of a semitrailer provided in an embodiment of the present application, such as Figure 2As shown, it is a schematic diagram of the docking scenario provided in an embodiment of the present application. A device for auxiliary docking of a semi-trailer provided in an embodiment of the present application is used to connect with a designated tractor and is specifically installed in the middle position of the rear end beam of the designated tractor. It includes a binocular vision camera, an inertial measurement unit and a docking analysis module. The binocular vision camera has a communication connection with the docking analysis module, and the inertial measurement unit has a communication connection with the docking analysis module. Among them, the binocular vision camera is used to obtain the real-time traction pin image during the docking process of the designated semi-trailer and transmit it to the docking analysis module. The real-time traction pin image is used to record the real-time relative position of the traction pin in the designated semi-trailer within a preset time period. information; the inertial measurement unit is used to obtain the real-time docking position data of the designated tractor and the designated semi-trailer, and transmit it to the docking analysis module, the real-time docking position data is used to describe the spatial position information of the traction saddle of the designated tractor and the traction pin of the designated semi-trailer within a preset time period; the docking analysis module is used to obtain the real-time position data of the traction pin according to the received real-time traction pin image and real-time docking position data, the real-time position data of the traction pin is used to describe the three-dimensional spatial position of the traction pin in the designated semi-trailer and the spatial position information of the traction saddle relative to the designated tractor, specifically including the distance and azimuth, thereby realizing the docking of the designated semi-trailer with the designated tractor.

[0028] In this embodiment, an inertial measurement unit (IMU) is a device used to measure and calculate the direction, velocity and acceleration of an object in space. It is usually composed of a gyroscope, an accelerometer and a magnetometer. The various parts of the inertial measurement unit usually sample data at a certain frequency to determine the spatial position of the towing pin relative to the binocular vision camera and the towing saddle, thereby realizing the docking of a designated tractor with a designated semi-trailer; the spatial position information of the towing pin is accurately determined through the images captured by the binocular vision camera, and the spatial position information is monitored in real time by continuously capturing images during the docking process. Problems in the docking process can be discovered in the shortest time, thereby realizing efficient docking of the designated tractor with the designated semi-trailer, and effectively solving the problem of inefficient auxiliary docking of semi-trailers in the prior art.

[0029] An embodiment of the present application provides a method for assisted docking of semi-trailers, which is applied to a device for assisted docking of semi-trailers, and the method includes the following steps: acquiring a first traction pin image through a binocular vision camera, and constructing a first traction pin detection model based on the acquired first traction pin image, wherein the first traction pin image is used to record the position information of the traction pin in a specified semi-trailer under an experimental environment, and the first traction pin detection model is used to identify the traction pin in the specified semi-trailer; before docking with the specified semi-trailer, an initial traction pin image is acquired through a binocular vision camera, and inputted into the constructed first traction pin detection model for traction pin identification, and at the same time, the traction pin initial position data is obtained in combination with the initial docking position data acquired by the inertial measurement unit, and the acquisition of the initial traction pin image is performed at a certain frequency until the first traction pin is passed. The detection model successfully identifies the towing pin and is within the initially set time range. The towing pin initial position data describes the initial spatial position information of the towing pin of the specified semi-trailer, including distance and direction angle. During the docking process of the specified semi-trailer, a real-time towing pin image is obtained through a binocular vision camera and input into the constructed first towing pin detection model for towing pin identification. At the same time, the real-time docking position data obtained by the inertial measurement unit is combined to obtain the towing pin real-time position data. It is judged whether the obtained towing pin real-time position data meets the requirements. If so, it indicates that the docking of the specified tractor and the specified semi-trailer is completed. Otherwise, the position of the specified semi-trailer is adjusted according to the towing pin real-time position data until the requirements are met. The requirement here refers to successful docking, which is specifically manifested in the docking of the towing pin with the center of the towing saddle.

[0030] In this embodiment, the experimental environment is used to simulate all position conditions of the designated semi-trailer relative to the designated tractor, the initial docking position data is used to describe the initial spatial position information of the designated tractor and the designated semi-trailer within a preset time period, the initial traction pin image is used to record the relative position information of the traction pin in the designated semi-trailer before the designated semi-trailer is docked within a preset time period, and the position adjustment is used to reduce the position difference between the traction pin of the designated semi-trailer and the traction saddle of the designated tractor to achieve rapid docking; more accurate identification of the spatial position information of the traction pin is achieved, thereby achieving improved docking efficiency between the designated semi-trailer and the designated tractor, that is, the efficiency of semi-trailer assisted docking is improved, which effectively solves the problem of inefficient semi-trailer assisted docking in the prior art.

[0031] Specifically, the specific method for acquiring the first traction pin image is as follows: determine the experimental environment factors and the experimental environment range in combination with the actual docking environment of the designated tractor and the designated semi-trailer and the parameters of the binocular vision camera, wherein the experimental environment factors represent the amount of change in the experimental environment, and the experimental environment factors specifically include angle, distance, light intensity and resolution; the experimental environment range includes angle range, distance range, light intensity range and resolution range, wherein the angle range refers to the angle change range of the designated semi-trailer relative to the designated tractor, the distance range refers to the distance change range of the designated semi-trailer relative to the designated tractor, the light intensity change range refers to the light intensity change range in the actual docking environment, and the resolution change range refers to the resolution change range of the binocular vision camera. The angle range, distance range and light intensity range can all be obtained based on the response data statistics of the historical time period, and the resolution range is obtained based on the parameters of the binocular vision camera; design the experimental environment based on the experimental environment factors and the experimental environment range, and use binocular vision to The camera collects the corresponding first traction pin image, wherein the experimental environment includes the first experimental environment, the second experimental environment, the third experimental environment and the fourth experimental environment. In the first experimental environment, the angle varies within the angle range and the distance, light intensity and resolution are the same as the corresponding preset values. In the second experimental environment, the distance varies within the distance range and the angle, light intensity and resolution are the same as the corresponding preset values. In the third experimental environment, the light intensity varies within the light intensity range and the angle, distance and resolution are the same as the corresponding preset value. In the fourth experimental environment, the resolution varies within the resolution range and the angle, distance and light intensity are the same as the corresponding preset value. Specifically, the preset values ​​corresponding to the angle, distance and light intensity can be obtained by taking the average value of the results after the corresponding data statistics of the historical time period, and the preset value corresponding to the resolution is obtained according to the parameters of the binocular vision camera; the diversity of the first traction pin image is improved, which is closer to the docking situation of the semi-trailer in the real situation, and then a first traction pin detection model with more accurate traction pin identification is constructed.

[0032] Furthermore, a specific method for acquiring the initial position data of the traction pin is as follows: A1, judging whether the traction pin in the initial traction pin image is detected based on the result of the traction pin recognition. If the traction pin in the initial traction pin image is detected, execute A2, otherwise continue to acquire a new initial traction pin image for traction pin recognition until the traction pin is recognized; A2, judging whether it is the first time that the traction pin in the initial traction pin image is detected. If so, execute A3 after initializing the timer, otherwise execute A3 directly; A3, obtaining the position information of the traction pin relative to the traction saddle through the point cloud data of the image corresponding to the binocular vision camera and combining the result of the traction pin recognition and the initial docking position data acquired by the inertial measurement unit, thereby obtaining the initial position data of the traction pin.

[0033] In this embodiment, the three-dimensional position information of the traction pin in the camera coordinate system is obtained through the point cloud data of the corresponding image of the binocular vision camera and the area information of the traction pin in the initial traction pin image, and the position information of the traction pin relative to the traction saddle is obtained in combination with the relative position information of the binocular vision camera and the traction saddle of the designated traction vehicle, thereby obtaining the traction pin positioning data within the initial set time range and the traction pin initial position data; the traction pin initial position data includes the distance and azimuth of the traction pin relative to the traction saddle; and more accurate acquisition of the traction pin initial position data is achieved.

[0034] Specifically, to obtain the position information of the traction pin relative to the traction saddle For example, the three-dimensional position information of the traction pin in the reference system is obtained based on the relative position and posture relationship between the binocular vision camera and the traction saddle. ,in, is the rotation coordinate transformation between the camera coordinate system and the saddle coordinate system. The center of the camera coordinate system refers to the binocular vision camera, and the center of the saddle coordinate system refers to the traction saddle. is the translation coordinate transformation of the camera coordinate system relative to the saddle coordinate system; within the initial set time range, a set of identified traction pin positioning data is obtained through the above steps , where r is the total number of initial traction pin images within the initial set time range. By calculating the average value, a more accurate initial position information of the traction pin relative to the saddle coordinate system is obtained. , calculated , and the initial distance of the traction pin relative to the saddle coordinate system And azimuth information , the initial distance and azimuth information is the initial position data, where The initial position information is also the initial position of the traction pin relative to the global fixed coordinate system.

[0035] Furthermore, the specific method for obtaining the real-time position data of the traction pin is as follows: B1, judging whether the distance data of the traction pin relative to the traction saddle is within a preset distance range based on the initial position data of the traction pin. If so, execute B2, otherwise execute B4. The preset distance range is represented by the shortest distance and the closest distance obtained after multiple experiments by the preset driver. During the experiment, except for the change in the distance from the designated tractor, other factors remain the same; B2, judging whether the traction pin is detected by the traction pin recognition corresponding to the input real-time traction pin image. If the traction pin is detected, execute B3, otherwise execute B4; B3, obtaining the position information of the traction pin relative to the binocular vision camera through the point cloud data corresponding to the real-time traction pin image and combining the area of ​​the traction pin recognition, and obtaining the real-time position data of the traction pin by combining the relative position information of the binocular vision camera and the traction saddle; B4, obtaining the position information of the traction pin relative to the binocular vision camera based on the real-time docking position data obtained by the inertial measurement unit and combining the initial position data of the traction pin, and then obtaining the real-time position data of the traction pin by combining the relative position information of the binocular vision camera and the traction saddle.

[0036] In this embodiment, For example, if the current distance between the traction pin and the saddle coordinate system is The traction pin is detected within the preset distance range, and the 3D position information of the traction pin in the camera coordinate system is obtained by combining the point cloud data collected by the binocular vision camera and the area information of the traction pin in the image. Then, according to the relative position and posture relationship between the binocular vision camera and the saddle coordinate system, the three-dimensional position information of the traction pin in the saddle coordinate system is obtained. And the distance of the traction pin relative to the saddle coordinate system And azimuth information On the contrary, the position information of the camera coordinate system relative to the global fixed coordinate system is obtained through the binocular vision camera and the inertial measurement unit , combined with the initial position of the traction pin relative to the global fixed coordinate system Get the position of the traction pin in the camera coordinates ,in, It is the rotation coordinate transformation between the global fixed coordinate system and the camera coordinate system, and the three-dimensional position information of the traction pin in the saddle coordinate system is obtained based on it. And the distance of the traction pin relative to the saddle coordinate system and azimuth information ; It realizes more accurate acquisition of the real-time position data of the towing pin, thereby realizing the auxiliary docking of the semi-trailer more quickly and accurately.

[0037] Furthermore, the specific process of adjusting the position of the designated semi-trailer according to the real-time position data of the towing pin is as follows: D1, determine whether the azimuth angle in the real-time position data of the towing pin is 0. If the azimuth angle is 0, execute D2, otherwise execute D3; D2, send a reverse command to the driver of the designated tractor until the corresponding real-time position data of the towing pin meets the requirements; D3, send a steering wheel control command to the driver of the designated tractor, and monitor whether the azimuth angle decreases. If it decreases, continue to monitor until the corresponding azimuth angle is 0 and then execute D2, otherwise execute D4; D4, send a reverse steering wheel control command to the driver of the designated tractor, and monitor the change of the azimuth angle until the corresponding azimuth angle is 0 and then execute D2.

[0038] In this embodiment, the distance between the traction pin and the saddle coordinate system is , and real-time estimation of azimuth The goal of guiding the tractor to connect with the semi-trailer is to achieve ; Specifically, if ,and If the driver is expected to turn the steering wheel in a certain direction, If the absolute value of becomes larger, the driver is prompted to turn in the opposite direction, and the driver slowly reverses until , then the steering wheel returns to the center position. If , but there is a trend of decreasing, then repeat and remind the preset driver to move in the direction as well as Reversing, finally The docking is completed in a timely manner; the docking between the designated semi-trailer and the designated tractor is completed more efficiently and comprehensively.

[0039] Furthermore, a specific method for constructing a first traction pin detection model is as follows: construct a recognition sample data set based on the collected first traction pin image, and divide the recognition sample data set into a training set, a verification set and a test set according to a preset ratio, wherein the recognition sample data set is a collection of first traction pin images, and the preset ratio is 8:1:1; the manually annotated training set is input into the candidate detection model for model training, and the result of the model training is verified and optimized through the verification set to obtain an initial traction pin detection model, where manual annotation refers to annotating the traction pin in the image with a rectangular frame, and the candidate detection model has a target detection function, such as YOLO (You Only Look Once) series models, the initial traction pin detection model is a converged model, and the purpose of optimization is to further improve the performance of the model; the initial traction pin detection model is tested for the first time through the test set to obtain the test time and test results, and the model detection coefficient of the initial traction pin detection model is obtained on this basis. The model detection coefficient is used to numerically describe the performance of the initial traction pin detection model in the test set; the initial traction pin detection model is tested for the second time through the field docking experiment to obtain the field docking time, and the model field coefficient of the initial traction pin detection model is obtained on this basis. The model field coefficient is used to numerically describe the performance of the initial traction pin detection model in the field docking experiment. The field docking time refers to the preset driving The time taken by the driver to complete the docking of the designated tractor and the designated semi-trailer each time. The field docking experiment refers to the preset driver conducting multiple field dockings of the designated tractor and the designated semi-trailer. The corresponding model comprehensive coefficient is obtained by combining the model detection coefficient and the model field coefficient of the initial traction pin detection model, and the initial traction pin detection model is screened according to the model comprehensive coefficient to obtain the first traction pin detection model. The model comprehensive coefficient is used to comprehensively describe the performance of the initial traction pin detection model. During the screening process, if the model comprehensive coefficients of multiple initial traction pin detection models are the same, then the model parameters can be combined for re-screening, and the model with the smallest model parameters is selected as the first traction pin detection model.

[0040] In this embodiment, in order to further improve the robustness and generalization ability of the first traction pin detection model, it also involves the operation of data enhancement on the first traction pin image to increase the amount of data in the recognition sample data set, wherein the data enhancement method mainly refers to the geometric transformation of the data, such as rotation, scaling, etc.; manual annotation is equivalent to giving the reference number and reference position of the traction pins in the image, the number of rectangular boxes represents the reference number of traction pins, and the pixel position of the rectangular box in the image represents the reference position of the traction pin, and the spatial position information of the traction pin can be determined in combination with the camera parameters of the binocular vision camera; the numerical description of the performance of the initial traction pin detection model is more conducive to model screening, and the first traction pin detection model should be the model with the best performance in the initial traction pin detection model. Theoretically, as the number of training samples increases, the performance of the model will be further improved; a more accurate first traction pin detection model is constructed to realize the recognition of the traction pin, and then the semi-trailer assisted docking is realized more accurately and efficiently.

[0041] Furthermore, the results of model training are verified and optimized through the validation set to obtain the initial traction pin detection model. The specific process is as follows: Step 1, input the validation set into the result of model training to obtain the first validation image, and perform statistics on the recognition deviation in the first validation image. The recognition deviation includes the missed detection ratio and the false detection ratio; Step 2, determine whether the statistical recognition error is 0. If it is 0, optimize the training result through the test set, otherwise execute step 3; Step 3, optimize the prior frame size of the model training result in combination with the clustering algorithm, and optimize the prediction frame screening in combination with the suppression method. After the optimization, the model needs to be retrained until the model converges. The prior frame size optimization refers to optimizing the prior frame of the selected detection model to make it more consistent with the size of the traction pin.

[0042] In this embodiment, the first verification image has traction pin marks annotated by the trained model, and the recognition deviation can be obtained by comparing the first verification image with the result of manual annotation; the missed detection ratio refers to the proportion of the number of missed traction pins in the total number of traction pins in the verification set, and the false detection ratio refers to the proportion of the number of falsely detected traction pins in the total number of traction pins in the verification set; the prior frame optimization can be specifically clustered through k-means, k-means++, etc. Taking k-mesns++ clustering as an example, first, the length and width features of the traction pin are extracted according to the results of manual annotation, and then the extracted features are clustered using the K-means++ algorithm. For each cluster cluster, the number of The eigenvalues ​​of the center point (i.e., the average bounding box size) can be used as the prior box size of the YOLO model, and finally the cluster center, that is, the size of the prior box, is obtained. The prior box size of the candidate detection model is replaced with the clustering result and the model is retrained; the suppression methods include non-maximum suppression and flexible non-maximum suppression. For example, the role of non-maximum suppression is to screen out the boxes with the largest scores belonging to the same category in a certain area. Looping over the categories can help us perform non-maximum suppression on each category separately. The core is to screen out the prediction box according to the overlap degree score; it realizes the construction of a more accurate initial tow pin detection model, and then screens out a more accurate first tow pin detection model for tow pin identification.

[0043] Furthermore, in addition to being directly expressed by the F1 score of the initial traction pin detection model, the model comprehensive coefficient can also be calculated more accurately; the model comprehensive coefficient is calculated using the following formula: ;

[0044] Where n is the number of the initial traction pin detection model, , N is the total number of initial traction pin detection models, is the comprehensive model coefficient of the nth initial traction pin detection model, is the model detection coefficient of the nth initial traction pin detection model, is the left model detection coefficient of the initial traction pin detection model, is the right model detection coefficient of the initial traction pin detection model, is the model field coefficient of the nth initial traction pin detection model, is the left model field coefficient of the initial traction pin detection model, is the right model field coefficient of the initial traction pin detection model, is the detection weight, is the field weight, is a logarithmic factor, and , the value range of the detection weight and the real weight is 0 to 1, and the sum of the detection weight and the real weight is 1, that is, .

[0045] In this embodiment, the left model detection coefficient is the minimum value among the model detection coefficients of all initial traction pin detection models, and the right model detection coefficient is the maximum value among the model detection coefficients of all initial traction pin detection models; similarly, the left model field coefficient is the minimum value among the model field coefficients of all initial traction pin detection models, and the right model field coefficient is the maximum value among the model field coefficients of all initial traction pin detection models; the detection weight and the field weight are determined by the hierarchical analysis method, wherein the detection weight represents the influence of the model detection coefficient on the model comprehensive coefficient, and the field weight represents the influence of the model field coefficient on the model comprehensive coefficient; a numerical description of the performance of the initial traction pin detection model is realized, thereby realizing a more accurate determination of the first traction pin detection model to realize traction pin identification.

[0046] Specifically, with the help of and The calculation formula of the comprehensive coefficient of the model can be simplified to For example, , , the formula of the simplified model comprehensive coefficient is: ,in , , is the model detection score of the nth initial tow pin detection model, is the model ground truth score of the nth initial traction pin detection model, such as Figure 3 As shown, it is a change diagram of the model comprehensive coefficient provided in an embodiment of the present application. It can be seen from the figure that the model comprehensive coefficient is directly proportional to the model detection score and is also directly proportional to the model field score. When the Fu Shou Nian Gao model detection score and the model field score are both 0, the model comprehensive coefficient is the minimum value 0. Because the model detection score and the model field score have different degrees of influence on the model comprehensive coefficient, the corresponding change relationship is also different; a more intuitive description of the performance of the initial traction pin detection model is achieved, thereby achieving the determination of a more accurate first traction pin detection model to assist in the docking of the semi-trailer.

[0047] Furthermore, in addition to being directly expressed by accuracy, the model detection coefficient of the initial traction pin detection model can also be calculated more accurately: the specific method for obtaining the model detection coefficient is as follows: the test time of the initial traction pin detection model is statistically analyzed to obtain the left test time and the right test time, thereby obtaining the limit test time difference and the test time coefficient of the initial traction pin detection model, wherein the left test time represents the minimum test time of the initial traction pin detection model, the right test time represents the maximum test time of the initial traction pin detection model, the limit test time difference is the difference between the right test time and the left test time, and the test time coefficient is the ratio of the difference between the actual test time and the left test time to the limit test time difference; the recognition accuracy of the traction pin in the test results of the initial traction pin detection model is statistically analyzed to obtain the left test accuracy and the right test accuracy, thereby obtaining the limit test accuracy difference and the test accuracy coefficient of the initial traction pin detection model, wherein the left test accuracy represents the minimum accuracy of the initial traction pin detection model, and the right test accuracy represents the maximum accuracy of the initial traction pin detection model. The accuracy of the limit test is the difference between the right test accuracy and the left test accuracy. The test accuracy coefficient is the ratio of the difference between the actual test accuracy and the left test accuracy to the limit test accuracy difference. The I / O ratio of the traction pin in the test results of the initial traction pin detection model is statistically obtained to obtain the left test I / O ratio and the right test I / O ratio, and the limit test I / O ratio difference and the test I / O ratio coefficient of the initial traction pin detection model are obtained accordingly. Among them, the left test I / O ratio represents the minimum I / O ratio of the initial traction pin detection model, and the right test I / O ratio represents the minimum I / O ratio of the initial traction pin detection model. The maximum intersection-and-union ratio of the guide pin detection model, the limit test intersection-and-union ratio difference is the difference between the right test intersection-and-union ratio and the left test intersection-and-union ratio, the test intersection-and-union ratio coefficient is the ratio of the difference between the actual test intersection-and-union ratio and the left test intersection-and-union ratio and the limit test intersection-and-union ratio difference, and the intersection-and-union ratio is the result of dividing the intersection of the detection frame area and the real frame area by the union of the detection frame area and the real frame area; the model detection coefficient of the initial traction pin detection model is obtained according to the obtained test time coefficient, test accuracy coefficient and test intersection-and-union ratio coefficient, and the model detection coefficient is calculated using the following formula: ;

[0048] Where n is the number of the initial traction pin detection model, , N is the total number of initial traction pin detection models, is the model detection coefficient of the nth initial traction pin detection model, is the test time coefficient of the nth initial traction pin calibration model, is the test accuracy coefficient of the nth initial traction pin detection model, is the test intersection-combination ratio coefficient of the nth initial traction pin calibration model.

[0049] In this embodiment, the test time coefficient, the test accuracy coefficient and the test intersection-to-combination ratio coefficient all have a variation range of 0 to 1. The variation statistics of the model detection coefficient are shown in Table 1 below:

[0050] Table 1 Statistics of changes in model detection coefficients <![CDATA[Test time coefficient CT n > <![CDATA[Test accuracy coefficient CT n > <![CDATA[Test the coefficient of intersection over union CJ n > <![CDATA[Model detection coefficient QM n > 0.517 0.846 0.563 2.676 0.652 0.829 0.785 2.523 0.768 0.874 0.324 2.520 0.996 0.545 0.599 2.519 0.870 0.747 0.513 2.549 ... ... ... ...

[0051] Specifically, the larger the test accuracy coefficient, the better, and the smaller the test time coefficient and the test intersection-and-combination ratio coefficient, the better. However, it does not mean that the model detection coefficient of the initial traction pin detection model with a large test accuracy coefficient will be larger. For example, the test accuracy coefficient in the third set of data reaches 0.874, but the corresponding model detection coefficient is not the smallest data among these sets of data. In short, it is necessary to comprehensively consider the values ​​of the test time coefficient, the test accuracy coefficient and the test intersection-and-combination ratio coefficient to obtain an accurate model detection coefficient, which achieves a more intuitive description of the performance of the initial traction pin detection model in the test set. Combined with the model field coefficient, a more comprehensive and intuitive description of the performance of the initial traction pin detection model can be achieved, thereby determining a more accurate first traction pin detection model to assist the docking of the designated semi-trailer with the designated tractor.

[0052] Furthermore, the model field coefficient can also be obtained by scoring the preset driver according to the docking assistance effect in the field docking experiment. Specifically, it can be represented by the relative deviation between the time for the preset driver to complete the docking in the field docking experiment and the time for completing the docking based on experience. The specific method for obtaining the model field coefficient is as follows: the field docking time is obtained by the preset driver performing the field docking experiment in the experimental area within the preset time period. The field docking time includes the experience docking time and the model docking time. The field docking experiment includes the experience docking experiment and the model docking experiment. Among them, the experience docking time represents the docking time of the preset driver in each experience docking experiment, and the model docking time represents the docking time of the preset driver in each model docking experiment; The total docking time of the preset driver is obtained by statistics of the field docking time through the integral operation. The total docking time includes the total experience docking time and the total model docking time. The total experience docking time represents the total docking time of the preset driver in the experience docking experiment, and the total model docking time represents the total docking time of the preset driver in the model docking experiment; the field docking time of the preset driver is sequentially sorted to obtain the experience docking sequence and the model docking sequence, and the experience docking extreme value and the model docking extreme value are obtained accordingly, wherein the experience docking extreme value represents the range of the experience docking time, and the model docking extreme value represents the range of the model docking time, and the model field coefficient of the initial traction pin detection model is obtained in combination with the obtained total docking time. The model field coefficient is calculated using the following formula: ;

[0053] Where n is the number of the initial traction pin detection model, , N is the total number of initial traction pin detection models, is the model field coefficient of the nth initial traction pin detection model, q is the preset driver number, , Q is the total number of preset drivers, x is the number of times of experience docking experiment, , A is the total number of empirical docking experiments, y is the number of field docking experiments, , B is the total number of field docking experiments. The total number of empirical docking experiments is the same as the total number of field docking experiments. represents the experience docking time function of the qth preset driver in the xth experience docking experiment, represents the model docking time function in the yth model docking experiment performed by the qth preset driver through the nth initial traction pin detection model, the corresponding time obtained by the empirical docking time function and the model docking function, i.e., the timer, is the minimum value of the experience docking of the qth preset driver, is the maximum value of the experience connection of the qth preset driver, is the model docking minimum value corresponding to the model docking experiment performed by the qth preset driver through the nth initial traction pin detection model, It is the model docking maximum value corresponding to the model docking experiment performed by the qth preset driver through the nth initial traction pin detection model.

[0054] In this embodiment, the experience docking experiment refers to the preset driver realizing the docking of the designated semi-trailer with the designated tractor according to his own experience in the experimental area, and the model docking experiment refers to the preset driver realizing the docking of the designated semi-trailer with the designated tractor according to the initial traction pin detection model; specifically, the experience docking extreme value includes the experience docking time minimum value and the experience docking time maximum value, and the model docking extreme value includes the model docking time minimum value and the model docking time maximum value; wherein the experience docking time function and the model docking time function are the corresponding experience docking time and model docking time; the model field coefficient is inversely proportional to the total model docking time, and the longer the total model docking time, the worse the actual application performance; the model field docking coefficient is inversely proportional to the model docking minimum value, and the larger the model docking minimum value, the smaller the corresponding model field coefficient, so as to achieve a more accurate description of the actual application performance of the initial traction pin detection model, and the combination of the model detection coefficient can realize a comprehensive description of the performance of the initial traction pin detection model, and then the traction pin recognition can be realized through the first traction pin detection model with the best performance, thereby further improving the efficiency of the semi-trailer auxiliary docking.

[0055] To summarize, the embodiment of the present application constructs an initial towing pin detection model by acquiring the first towing pin image, and then obtains the model comprehensive coefficient through the corresponding model detection coefficient and the model field coefficient to determine the first towing pin detection model, and then inputs the acquired initial towing pin image into the constructed first towing pin detection model to obtain the towing pin initial position data, and finally combines the real-time towing pin image to obtain the real-time position data of the towing pin to achieve the docking of the designated semi-trailer with the designated tractor, thereby achieving accurate docking of the designated semi-trailer, and then achieving improved efficiency of semi-trailer assisted docking, effectively solving the problem of inefficient semi-trailer assisted docking in the prior art.

[0056] It will be appreciated by those skilled in the art that embodiments of the present invention may be provided as methods, systems, or computer program products. Therefore, the present invention may take the form of a complete hardware embodiment, a complete software embodiment, or an embodiment combining software and hardware. Furthermore, the present invention may take the form of a computer program product implemented on one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0057] The present invention is described with reference to flowcharts and / or block diagrams of systems, devices (systems), and computer program products according to embodiments of the present invention. It should be understood that each process and / or block in the flowchart and / or block diagram, as well as the combination of processes and / or blocks in the flowchart and / or block diagram, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, a special-purpose computer, an embedded processor, or other programmable data processing device to produce a machine, so that the instructions executed by the processor of the computer or other programmable data processing device generate instructions for implementing the processes in the flowchart and / or block diagram. Figure 1 A process or multiple processes and / or boxes Figure 1 A device that provides the functions specified in a block or multiple blocks.

[0058] These computer program instructions may also be stored in a computer-readable memory capable of directing a computer or other programmable data processing device to operate in a specific manner, so that the instructions stored in the computer-readable memory produce an article of manufacture comprising an instruction device, which implements the process Figure 1 A process or multiple processes and / or boxes Figure 1 A function specified in one or more boxes.

[0059] These computer program instructions can also be loaded onto a computer or other programmable data processing device so that a series of operating steps are executed on the computer or other programmable device to produce a computer-implemented process, thereby providing instructions for implementing the process. Figure 1A process or multiple processes and / or boxes Figure 1 The steps for the functions specified in one or more boxes.

[0060] Although the preferred embodiments of the present invention have been described, those skilled in the art may make other changes and modifications to these embodiments once they have learned the basic creative concept. Therefore, the appended claims are intended to be interpreted as including the preferred embodiments and all changes and modifications that fall within the scope of the present invention.

[0061] Obviously, those skilled in the art can make various changes and modifications to the present invention without departing from the spirit and scope of the present invention. Thus, if these modifications and variations of the present invention fall within the scope of the claims of the present invention and their equivalents, the present invention is also intended to include these modifications and variations.

Claims

1. A device for auxiliary docking of a semitrailer, used for connecting with a designated tractor, characterized by: It includes a binocular vision camera, an inertial measurement unit and a docking analysis module, wherein the binocular vision camera is communicatively connected to the docking analysis module, and the inertial measurement unit is communicatively connected to the docking analysis module; Wherein, the binocular vision camera is used to obtain the real-time towing pin image of the designated semi-trailer during docking, and transmit it to the docking analysis module; The inertial measurement unit is used to obtain real-time docking position data of the designated tractor and the designated semi-trailer, and transmit the data to the docking analysis module; The docking analysis module is used to obtain the real-time position data of the towing pin according to the received real-time towing pin image and the real-time docking position data, thereby achieving the docking of the designated semi-trailer with the designated tractor.

2. A method for auxiliary docking of a semitrailer, applied to the device for auxiliary docking of a semitrailer as claimed in claim 1, characterized in that: The following steps are involved: Collecting a first traction pin image through a binocular vision camera, and constructing a first traction pin detection model according to the collected first traction pin image; Before docking with a designated semi-trailer, an initial towing pin image is obtained through a binocular vision camera and input into the constructed first towing pin detection model for towing pin recognition. At the same time, the towing pin initial position data is obtained by combining the initial docking position data obtained by the inertial measurement unit; During the docking process of the designated semi-trailer, a real-time towing pin image is obtained through a binocular vision camera, and is input into the constructed first towing pin detection model for towing pin recognition, and the real-time docking position data obtained by the inertial measurement unit is combined to obtain the real-time position data of the towing pin; It is determined whether the obtained real-time position data of the towing pin meets the requirements. If so, it indicates that the docking of the designated tractor and the designated semi-trailer is completed. Otherwise, the position of the designated semi-trailer is adjusted according to the real-time position data of the towing pin until the requirements are met.

3. A device for auxiliary docking of a semitrailer as claimed in claim 2, characterized in that: The specific method of constructing the first traction pin detection model is as follows: Constructing a recognition sample data set according to the collected first traction pin image, and dividing the recognition sample data set into a training set, a verification set, and a test set according to a preset ratio; The manually annotated training set is input into the candidate detection model for model training, and the model training result is verified and optimized through the verification set to obtain the initial traction pin detection model; Performing a first test on the initial traction pin detection model through the test set to obtain the test time and test results, and obtaining the model detection coefficient of the initial traction pin detection model based on the test set, wherein the model detection coefficient is used to numerically describe the performance of the initial traction pin detection model in the test set; The initial traction pin detection model is tested for the second time through the field docking experiment to obtain the field docking time, and the model field coefficient of the initial traction pin detection model is obtained based on the field docking time. The model field coefficient is used to numerically describe the performance of the initial traction pin detection model in the field docking experiment; The corresponding model comprehensive coefficient is obtained by combining the model detection coefficient and the model field coefficient of the initial traction pin detection model, and the initial traction pin detection model is screened according to the model comprehensive coefficient to obtain the first traction pin detection model. The model comprehensive coefficient is used to comprehensively describe the performance of the initial traction pin detection model.

4. A device for auxiliary docking of a semitrailer as claimed in claim 3, characterized in that: The specific process of verifying and optimizing the model training results through the validation set to obtain the initial tow pin detection model is as follows: Step 1: Input the verification set into the result of model training to obtain a first verification image, and perform statistics on the recognition deviation in the first verification image; Step 2: determine whether the statistical recognition error is 0. If it is 0, optimize the training result through the test set, otherwise execute step 3; Step three: Combine the clustering algorithm to optimize the prior box size of the model training results, and combine the suppression method to perform prediction box screening optimization.

5. A method for auxiliary docking of a semitrailer as claimed in claim 3, characterized in that: The comprehensive coefficient of the model is calculated using the following formula: ; Where n is the number of the initial traction pin detection model, , N is the total number of initial traction pin detection models, is the comprehensive model coefficient of the nth initial traction pin detection model, is the model detection coefficient of the nth initial traction pin detection model, is the left model detection coefficient of the initial traction pin detection model, is the right model detection coefficient of the initial traction pin detection model, is the model field coefficient of the nth initial traction pin detection model, is the left model field coefficient of the initial traction pin detection model, is the right model field coefficient of the initial traction pin detection model, is the detection weight, is the field weight, is a logarithmic factor, and .

6. A device for auxiliary docking of a semitrailer as claimed in claim 3, characterized in that: The specific method for obtaining the model detection coefficient is as follows: The test time of the initial traction pin detection model is statistically obtained to obtain the left test time and the right test time, and the limit test time difference and the test time coefficient of the initial traction pin detection model are obtained accordingly; The recognition accuracy of the traction pin in the test results of the initial traction pin detection model is statistically obtained to obtain the left test accuracy and the right test accuracy, and the limit test accuracy difference and the test accuracy coefficient of the initial traction pin detection model are obtained accordingly; The I / O ratio of the traction pin in the test results of the initial traction pin detection model is statistically obtained to obtain the left test I / O ratio and the right test I / O ratio, and the limit test I / O ratio difference and the test I / O ratio coefficient of the initial traction pin detection model are obtained accordingly; The model detection coefficient of the initial traction pin detection model is obtained according to the obtained test time coefficient, test accuracy coefficient and test intersection-to-combination ratio coefficient.

7. A method for auxiliary docking of a semitrailer as claimed in claim 3, characterized in that: The specific method for obtaining the field coefficient of the model is as follows: The on-site docking time is obtained by performing an on-site docking experiment in the experimental area by a preset driver within a preset time period; The actual docking time is counted by the integration operation to obtain the total docking time of the preset driver; The preset driver's actual docking time is sorted in sequence to obtain the empirical docking sequence and the model docking sequence, and the empirical docking extreme value and the model docking extreme value are obtained based on this. The model actual coefficient of the initial traction pin detection model is obtained by combining the total docking time obtained. The model actual coefficient is calculated using the following formula: ; Where n is the number of the initial traction pin detection model, , N is the total number of initial traction pin detection models, is the model field coefficient of the nth initial traction pin detection model, q is the preset driver number, , Q is the total number of preset drivers, x is the number of times of experience docking experiment, , A is the total number of empirical docking experiments, y is the number of field docking experiments, , B is the total number of field docking experiments, represents the experience docking time function of the qth preset driver in the xth experience docking experiment, represents the model docking time function of the qth preset driver in the yth model docking experiment through the nth initial tow pin detection model, is the minimum value of the experience docking of the qth preset driver, is the maximum value of the experience connection of the qth preset driver, is the model docking minimum value corresponding to the model docking experiment performed by the qth preset driver through the nth initial traction pin detection model, It is the model docking maximum value corresponding to the model docking experiment performed by the qth preset driver through the nth initial traction pin detection model.

8. The device for auxiliary docking of a semitrailer as claimed in claim 2, characterized in that: The specific method for obtaining the initial position data of the traction pin is as follows A1, judging whether the traction pin in the initial traction pin image is detected according to the result of traction pin recognition, if the traction pin in the initial traction pin image is detected, executing A2, otherwise continuing to obtain the initial traction pin image for traction pin recognition; A2, determining whether the traction pin in the initial traction pin image is detected for the first time, if so, executing A3 after initializing the timer, otherwise directly executing A3; A3, the position information of the traction pin relative to the traction saddle is obtained by combining the point cloud data of the corresponding image of the binocular vision camera with the result of traction pin recognition and the initial docking position data obtained by the inertial measurement unit, and the initial position data of the traction pin is obtained accordingly.

9. The device for auxiliary docking of a semitrailer as claimed in claim 2, characterized in that: The specific method for obtaining the real-time position data of the traction pin is as follows: B1, judging whether the distance data of the traction pin relative to the traction saddle is within a preset distance range according to the initial position data of the traction pin, if it is within the preset distance range, executing B2, otherwise executing B4; B2, judging whether the traction pin recognition corresponding to the input real-time traction pin image detects the traction pin, if the traction pin is detected, executing B3, otherwise executing B4; B3, obtain the position information of the traction pin relative to the binocular vision camera through the point cloud data corresponding to the real-time traction pin image and the area identified by the traction pin, and obtain the real-time position data of the traction pin by combining the relative position information of the binocular vision camera and the traction saddle; B4, based on the real-time docking position data obtained by the inertial measurement unit and combined with the initial position data of the traction pin, the position information of the traction pin relative to the binocular vision camera is obtained, and then the real-time position data of the traction pin is obtained by combining the relative position information of the binocular vision camera and the traction saddle.

10. The device for auxiliary docking of a semitrailer as claimed in claim 2, characterized in that: The specific process of adjusting the position of a designated semi-trailer according to the real-time position data of the traction pin is as follows: D1, determine whether the azimuth angle in the real-time position data of the traction pin is 0. If the azimuth angle is 0, execute D2, otherwise execute D3; D2, sending a reverse command to the driver of the designated tractor until the corresponding tractor pin real-time position data meets the requirements; D3, sending a steering wheel control command to the driver of the designated tractor, and monitoring whether the azimuth angle decreases. If it decreases, continue monitoring until the corresponding azimuth angle is 0 and then execute D2, otherwise execute D4; D4, sends a reverse steering wheel control command to the driver of the designated tractor, and monitors the change of the azimuth angle until the corresponding azimuth angle is 0 and then executes D2.