Unmanned tractor control method and device, electronic equipment and medium
By using binocular stereoscopic cameras and object detection models on unmanned tractors to identify and locate obstacles, the problems of high measurement costs and low accuracy in the prior art are solved, and more efficient and accurate obstacle identification and control are achieved.
Patent Information
- Application Number
- CN202510125967.0
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2025-01-27
- Publication Date
- 2025-05-30
AI Technical Summary
Existing unmanned tractors have problems with high measurement costs and low measurement accuracy in obstacle identification and positioning.
A binocular stereoscopic vision camera is used to obtain color images and depth maps in the driving direction of the unmanned tractor, and the object detection model is used to identify obstacles, and the position of obstacles is determined by the fusion of color images and depth maps. Finally, the unmanned tractor is controlled based on this position.
It reduces the measurement cost of obstacle identification, and improves the measurement accuracy, solving the problem that the measurement cost and measurement accuracy cannot be taken into account during the control process of driverless tractors.
Smart Images
Figure CN120071297A_ABST
Abstract
Description
Technical Field
[0001] The present application relates to the technical field of image processing. Specifically, it relates to a control method, device, electronic device, and medium for an autonomous tractor. Background Technique
[0002] An autonomous tractor is a machine that can be used in agricultural production. It has the advantages of high tillage efficiency and labor cost savings. Using an autonomous tractor for tillage operations is an important direction for the development of high-end agricultural equipment and refined intelligent agriculture. Currently, positioning technology is usually used to achieve autonomous navigation and perform tillage operations in the farmland according to the path set by the navigation. During the operation of the autonomous tractor, the tractor travels along the planned path. To avoid collisions with obstacles on the traveling path, it is necessary to identify obstacles in the farmland.
[0003] However, existing autonomous tractors usually use lidar or a monocular intelligent camera for obstacle identification and positioning. However, the procurement and maintenance costs of lidar are relatively high, and it is prone to inaccurate measurement problems when affected by vehicle vibration and grain dust. Similarly, if obstacle identification is performed through a monocular intelligent camera, there is also a problem of low accuracy in position measurement. Summary of the Invention
[0004] In view of this, the purpose of the present application is to provide a control method, device, electronic device, and medium for an autonomous tractor to solve the problem that it is impossible to balance the measurement cost and measurement accuracy during the control process of the autonomous tractor.
[0005] In a first aspect, an embodiment of the present application provides a control method for an autonomous tractor, which is applied to an autonomous tractor control system. The autonomous tractor control system includes a binocular stereo vision camera. The binocular stereo vision camera is disposed outside the front upper part of the cockpit of the autonomous tractor and includes:
[0006] Using the binocular stereo vision camera, obtain a color image and a depth map in the driving direction of the autonomous tractor;
[0007] Using a target detection model, identify obstacles in the color image and mark the identified target obstacles in the color image;
[0008] Fuse the color image with the depth map marked with the target obstacles to determine the position of the target obstacles;
[0009] Based on the position of the target obstacles, control the autonomous tractor.
[0010] Optionally, determining the position of the target obstacle includes: determining the depth value between the target obstacle and the driverless tractor based on the fusion result of the color image and the depth map; determining the offset angle of the center of the target obstacle relative to the camera according to the conversion relationship between the image coordinate system and the world coordinate system; and determining the position of the target obstacle based on the offset angle and the depth value.
[0011] Optionally, determining the position of the target obstacle based on the offset angle and the depth value includes: multiplying the depth value by the tangent value of the offset angle to determine the vertical position of the target obstacle; and multiplying the depth value by the cotangent value of the offset angle to determine the horizontal position of the target obstacle.
[0012] Optionally, using a binocular stereo vision camera to obtain a color image and a depth map in the driving direction of the driverless tractor includes: using the binocular stereo vision camera to obtain a color image and processing the color image to obtain a grayscale image, where the grayscale image includes a first grayscale image and a second grayscale image; and obtaining a depth map in the driving direction of the driverless tractor based on the first grayscale image and the second grayscale image.
[0013] Optionally, obtaining a depth map in the driving direction of the driverless tractor based on the first grayscale image and the second grayscale image includes: determining the disparity value of each pixel point in the first grayscale image and the second grayscale image; generating a disparity map based on the disparity value of each pixel point, and obtaining a depth map according to the disparity map.
[0014] Optionally, the method further includes: splitting the training data set to obtain multiple groups of training data subsets, where each training data subset includes multiple training images; for each training data subset, randomly cropping and splicing the multiple training images to generate a training extended image corresponding to the training data subset; and training the target detection model using the training extended image and the training data set.
[0015] Optionally, the method further includes: using a preset tracking algorithm to track the position of the target obstacle in adjacent frame images to continuously label the target obstacle in different frame images.
[0016] In a second aspect, an embodiment of the present application further provides a driverless tractor control device, which is applied to a driverless tractor control system. The driverless tractor control system includes a binocular stereo vision camera, and the binocular stereo vision camera is disposed at the outer upper front of the driver's cab of the driverless tractor. The device includes:
[0017] An image acquisition module, configured to use a binocular stereo vision camera to obtain a color image and a depth map in the driving direction of the driverless tractor;
[0018] An obstacle recognition module, configured to use an object detection model to recognize obstacles in a color image and mark the recognized target obstacles in the color image;
[0019] A position calculation module, configured to fuse the color image with the marked target obstacles and the depth map to determine the position of the target obstacles;
[0020] A device control module, configured to control the driverless tractor based on the position of the target obstacles.
[0021] In a third aspect, an embodiment of the present application further provides an electronic device, including: a processor, a memory, and a bus. The memory stores machine-readable instructions executable by the processor. When the electronic device runs, the processor communicates with the memory through the bus. When the machine-readable instructions are executed by the processor, the steps of the driverless tractor control method as described above are executed.
[0022] In a fourth aspect, an embodiment of the present application further provides a computer-readable storage medium, on which a computer program is stored. When the computer program is run by a processor, the steps of the driverless tractor control method as described above are executed.
[0023] The embodiments of the present application bring the following beneficial effects:
[0024] A driverless tractor control method, device, electronic device, and medium provided by an embodiment of the present application can use a binocular stereo vision camera to obtain a color image and a depth map, so as to accurately calculate the position of a target obstacle through the color image and the depth map, and thus control the driverless tractor according to the position of the target obstacle, avoiding the problems of high measurement cost and low measurement accuracy when using a lidar or a monocular camera to control a driverless tractor. Compared with the driverless tractor control method in the prior art, the problem that the measurement cost and measurement accuracy cannot be taken into account in the control process of the driverless tractor is solved.
[0025] To make the above objects, features, and advantages of the present application more obvious and understandable, the following specific embodiments are given in conjunction with the accompanying drawings and described in detail as follows. Description of the Drawings
[0026] To more clearly illustrate the technical solutions of the embodiments of the present application, the accompanying drawings required for the embodiments will be briefly introduced below. It should be understood that the following drawings only show some embodiments of the present application and should not be regarded as limiting the scope. For those of ordinary skill in the art, other related drawings can be obtained based on these drawings without creative efforts.
[0027] Figure 1The flowchart of the driverless tractor control method provided by the embodiments of the present application is shown;
[0028] Figure 2 The structural schematic diagram of the driverless tractor control device provided by the embodiments of the present application is shown;
[0029] Figure 3 The structural schematic diagram of the electronic device provided by the embodiments of the present application is shown. Detailed implementation manners
[0030] To make the objectives, technical solutions, and advantages of the embodiments of the present application clearer, the technical solutions in the embodiments of the present application will be clearly and completely described below with reference to the accompanying drawings in the embodiments of the present application. Apparently, the described embodiments are only some of the embodiments of the present application, rather than all of the embodiments. Usually, the components of the embodiments of the present application described and illustrated in the accompanying drawings here can be arranged and designed in various different configurations. Therefore, the following detailed description of the embodiments of the present application provided in the accompanying drawings is not intended to limit the scope of the present application claimed, but merely represents the selected embodiments of the present application. Based on the embodiments of the present application, every other embodiment obtained by those skilled in the art without creative efforts belongs to the scope of protection of the present application.
[0031] It should be noted that before the present application was proposed, a driverless tractor is a machine that can be used for agricultural production. It has the advantages of high tillage efficiency and labor cost savings, and can promote the development of agricultural productivity. Using a driverless tractor for tillage operations is an important direction for the development of high-end agricultural equipment and the development of refined intelligent agriculture. Currently, the global navigation satellite system (GNSS) and real-time kinematic (RTK) technology are usually used to achieve driverless navigation to perform tillage operations in the farmland according to the path set by the navigation. During the operation of the driverless tractor, the tractor travels along the planned path. To avoid collisions with obstacles on the travel path, it is necessary to identify the obstacles in the farmland. However, existing driverless tractors usually use lidar or monocular intelligent cameras for obstacle identification and positioning. However, the procurement and maintenance costs of lidar are relatively high, and it is prone to inaccurate measurement problems when affected by vehicle vibration and grain dust. Similarly, if obstacle identification is performed through a monocular intelligent camera, there is also a problem of low accuracy in position measurement.
[0032] Based on this, the embodiments of the present application provide a driverless tractor control method to improve the accuracy of obstacle measurement while reducing the measurement cost.
[0033] Please refer to Figure 1 , Figure 1 , which is a flowchart of a control method for an autonomous tractor provided by an embodiment of the present application. As Figure 1 shown, the control method for an autonomous tractor provided by the embodiment of the present application includes:
[0034] Step S101: Use a binocular stereo vision camera to obtain a color image and a depth map in the driving direction of the autonomous tractor;
[0035] Step S102: Use an object detection model to identify obstacles in the color image and mark the identified target obstacles in the color image;
[0036] Step S103: Fuse the color image with the depth map marked with the target obstacles to determine the position of the target obstacles;
[0037] Step S104: Control the autonomous tractor based on the position of the target obstacles.
[0038] The control method for an autonomous tractor provided by the embodiment of the present application can use a binocular stereo vision camera to obtain a color image and a depth map, so as to accurately calculate the position of the target obstacles through the color image and the depth map, and thus control the autonomous tractor according to the position of the target obstacles, avoiding the problems of high measurement cost and low measurement accuracy when using a lidar or a monocular camera to control the autonomous tractor, and solving the problem that it is impossible to balance the measurement cost and the measurement accuracy in the control process of the autonomous tractor.
[0039] For the convenience of understanding this embodiment, the following takes the application of the control method for an autonomous tractor to an autonomous tractor control system as an example to separately describe the above exemplary steps provided by the embodiment of the present application. The autonomous tractor control system includes a binocular stereo vision camera, a domain controller, and an autonomous driving system.
[0040] The binocular stereo vision camera is arranged at the outer upper front of the cab of the autonomous tractor and is used to sense the environmental information of the autonomous tractor in the forward area; the domain controller is installed at the roof position inside the cab of the autonomous tractor and is used to receive and process the image and video information transmitted by the binocular stereo vision camera, execute the perception strategy algorithm, and communicate with the vehicle CAN signal. Among them, the binocular stereo vision camera and the domain controller are connected through Gigabit Ethernet to transmit information, and the domain controller and the autonomous driving system are connected through a CAN bus for vehicle-wide communication.
[0041] In step S101, use a binocular stereo vision camera to obtain a color image and a depth map in the driving direction of the autonomous tractor.
[0042] In this step, the binocular stereo vision camera includes two cameras, namely a left camera and a right camera. Both of these cameras are color imaging units, and these two cameras are respectively referred to as the first camera and the second camera. The two cameras continuously acquire images in the traveling direction of the driverless tractor, and this image is a color image.
[0043] In the embodiment of the present application, a binocular stereo vision camera can be used to acquire color images, and these color images include a first color image and a second color image. Then, the color images are processed to obtain grayscale images, and these grayscale images include a first grayscale image and a second grayscale image. For example: the first color image is subjected to grayscale processing to obtain the first grayscale image, and the second color image is subjected to grayscale processing to obtain the second grayscale image.
[0044] Then, based on the first grayscale image and the second grayscale image, a depth map in the traveling direction of the driverless tractor is obtained. Here, there is a pixel position disparity between the first grayscale image and the second grayscale image. This position disparity is the pixel difference between the column coordinates of the first grayscale image and the column coordinates of the second grayscale image at the corresponding points (the same position points), and this difference is the disparity value. At this time, the first grayscale image is the image captured by the left camera, and the second grayscale image is the image captured by the right camera. The disparity value is a numerical value in pixels. After determining the disparity values of each corresponding point in the first grayscale image and the second grayscale image, a disparity map can be generated based on the disparity value of each pixel point, and the three-dimensional point cloud depth information is calculated according to the disparity map to obtain a depth map, and this depth map includes the depth value Z of each pixel point.
[0045] In step S102, a target detection model is used to identify obstacles in the color image and mark the identified target obstacles in the color image.
[0046] In this step, the target detection model can refer to an object recognition model for identifying obstacles. As an example, the target detection model can be the YOLOv5 model.
[0047] In one example, when using a target detection model for obstacle recognition, the target detection model needs to be trained first. At this time, a training data set can be obtained and split into multiple training data subsets, and each training data subset includes multiple training images. Then, for each training data subset, the multiple training images in the training data subset are randomly cropped and scaled, and then the cropped and scaled images are spliced to generate a new image, which is the training extended image corresponding to the training data subset. This can increase small-sample targets and improve the model training speed. The original training data set is extended using the training extended image, and the target detection model is trained using the training extended image and the training data set. Each time during training, the extended training data set is divided into multiple training sets, and the optimal anchor box values in different training sets are adaptively calculated. After backpropagation update, the optimal anchor box values are iteratively saved into the network parameters of the target detection model for subsequent invocation from the target detection model.
[0048] In one example, the target detection model includes a backbone network Backbone, a neck network Neck, and a head network Head. Among them, the BackBone network includes four parts: CSP (Cross Stage Partial Network), Dropblock, Mish, and SPPF.
[0049] The BackBone network is responsible for extracting high-dimensional feature maps from the input image. First, the feature map of the target obstacle in the input image is extracted through the BackBone network, and the feature information in the feature map is extracted. Then, the feature map containing feature information of different dimensions is passed into the Neck network.
[0050] The Neck network is connected to the BackBone network. The Neck network includes an FPN module and a PAN module. The FPN module is respectively connected to the BackBone network and the PAN module. Feature fusion is performed using FPN and PAN to obtain the target fusion feature, and the target fusion feature is transmitted to the next layer of the network.
[0051] Through the FPN module, downsampling fusion can be performed on the obstacle feature maps of different dimensions to obtain the first fused obstacle feature map. Based on the first fused obstacle feature map, upsampling fusion is performed on the first fused obstacle feature map through the PAN module to obtain the second fused obstacle feature map. Finally, the target obstacle in the input image and the target box of the target obstacle on the input image are identified on the second fused obstacle feature map to complete the detection of the target obstacle.
[0052] The FPN module can improve the object detection effect through feature fusion, achieving a top-down semantic enhancement. The PAN module realizes a bottom-up localization enhancement based on the FPN module. The combination of the two is used to extract the feature map of obstacles, which can strengthen the network features.
[0053] Finally, the Head network outputs a feature vector, and the output feature vector includes information such as the category of the target obstacle, the obstacle bounding box, and the confidence. Among them, the obstacle bounding box is used to mark the target obstacle.
[0054] In an example, to avoid the situation where the same obstacle is not detected in two consecutive frames of images during the obstacle recognition process due to vehicle vibration or pitch angle change, a preset tracking algorithm can be used to track the position of the target obstacle in adjacent frame images, so as to continuously mark the target obstacle in different frame images. Among them, the preset tracking algorithm can refer to a tracking algorithm based on the Hungarian matching algorithm, and the tracking algorithm is part of the post-processing and is used for target tracking after the target detection model detects the target obstacle.
[0055] In step S103, the color image marked with the target obstacle is fused with the depth map to determine the position of the target obstacle.
[0056] In this step, multiple color images are sequentially passed into a queue, aligned with multiple collected depth maps according to the time sequence, and then the aligned color image and depth map are fused together. The obstacle bounding box is matched with the obstacle in the depth map to determine the depth value between the target obstacle and the driverless tractor based on the fusion result of the color image and the depth map.
[0057] Then, according to the conversion relationship between the image coordinate system and the world coordinate system, the offset angle of the center of the target obstacle relative to the camera is determined. Here, the conversion relationship can be expressed as:
[0058]
[0059] In the above formula, w represents the pixel width of the image, HFOV represents the horizontal field of view angle of the binocular stereo vision camera, f represents the camera focal length, and y represents the y-axis coordinate of the target obstacle in the image coordinate system.
[0060] From the above conversion relationship, it can be obtained that: θ = arctan(tan(HFOV / 2) * y / (w / 2)).
[0061] Then, based on the offset angle and the depth value, the position coordinates of the target obstacle can be determined. Here, since y / f = Y / Z, where Y and Z are coordinate values in the camera coordinate system, Y represents the vertical distance from the camera to the ground, and Z represents the horizontal distance from the camera to the obstacle. According to the Pythagorean theorem, the product of the depth value and the tangent value of the offset angle can be determined as the vertical position of the target obstacle, that is, Y = Z × tanθ; similarly, the product of the depth value and the cotangent value of the offset angle is determined as the horizontal position of the target obstacle, that is, X = Z × secθ. The position coordinates of the target obstacle are (X, Y, Z).
[0062] In step S104, the driverless tractor is controlled based on the position of the target obstacle.
[0063] In this step, after determining the position of the target obstacle, it can be judged whether the target obstacle is on the traveling path of the driverless tractor. If it is on the traveling path or the operation road of the driverless tractor, it means that it will affect the operation of the driverless tractor. Then, the position and type of the target obstacle can be transmitted to the driverless system in the form of a CAN message, and the device control scheme can be determined according to the relative position relationship between the target obstacle and the driverless tractor, and the device control scheme can be executed. For example: if the target obstacle is a movable device, the driverless tractor is controlled to sound the horn to signal the target obstacle to leave the current path; if the target obstacle is a non-movable device, the traveling path can be re-planned or an emergency brake can be implemented according to the distance between the target obstacle and the driverless tractor.
[0064] In the embodiment of the present application, a binocular stereo vision camera is used to collect color images, and the collected color images are processed by the domain control to obtain a depth map, and the domain controller executes step S102 and step S103 to determine the position and type of the target obstacle. The domain controller judges whether the target obstacle is on the traveling path of the driverless tractor, and when it is determined that the target obstacle affects the operation of the driverless tractor, the information is transmitted to the driverless system in the form of a CAN message, so that the driverless system controls the driverless tractor.
[0065] Based on the same inventive concept, an embodiment of the present application also provides a driverless tractor control device corresponding to the driverless tractor control method. Since the principle of solving problems by the device in the embodiment of the present application is similar to the above driverless tractor control method in the embodiment of the present application, the implementation of the device can refer to the implementation of the method, and the repeated parts will not be described again.
[0066] Please refer to Figure 2 , Figure 2The figure is a schematic structural diagram of a control device for an autonomous tractor provided by an embodiment of the present application. As Figure 2 shown in the figure, the control device 200 for the autonomous tractor is applied to an autonomous tractor control system. The autonomous tractor control system includes a binocular stereo vision camera, and the binocular stereo vision camera is arranged at the outer upper front of the cockpit of the autonomous tractor. The control device 200 for the autonomous tractor includes:
[0067] An image acquisition module 201, configured to use the binocular stereo vision camera to acquire a color image and a depth map in the driving direction of the autonomous tractor;
[0068] An obstacle recognition module 202, configured to use a target detection model to recognize obstacles in the color image and mark the recognized target obstacles in the color image;
[0069] A position calculation module 203, configured to fuse the color image with the depth map marked with the target obstacle to determine the position of the target obstacle;
[0070] A device control module 204, configured to control the autonomous tractor based on the position of the target obstacle.
[0071] Please refer to Figure 3 , Figure 3 which is a schematic structural diagram of an electronic device provided by an embodiment of the present application. As Figure 3 shown in the figure, the electronic device 300 includes a processor 310, a memory 320, and a bus 330.
[0072] The memory 320 stores machine-readable instructions executable by the processor 310. When the electronic device 300 runs, the processor 310 communicates with the memory 320 through the bus 330. When the machine-readable instructions are executed by the processor 310, the steps of the control method for the autonomous tractor in the method embodiment as shown above can be executed. The specific implementation manner can refer to the method embodiment and will not be elaborated here. Figure 1 shown.
[0073] An embodiment of the present application further provides a computer-readable storage medium. A computer program is stored on the computer-readable storage medium. When the computer program is run by a processor, the steps of the control method for the autonomous tractor in the method embodiment as shown above can be executed. The specific implementation manner can refer to the method embodiment and will not be elaborated here. Figure 1 shown.
[0074] Those skilled in the art can clearly understand that for the convenience and brevity of description, the specific working processes of the systems, devices, and units described above can refer to the corresponding processes in the foregoing method embodiments and will not be elaborated here.
[0075] In several embodiments provided in this application, it should be understood that the disclosed systems, devices, and methods can be implemented in other ways. The device embodiments described above are merely illustrative. For example, the division of the units is only a logical function division. In actual implementation, there may be other division methods. For another example, multiple units or components can be combined or integrated into another system, or some features can be ignored or not executed. Another point is that the displayed or discussed couplings or direct couplings or communication connections to each other can be through some communication interfaces. The indirect couplings or communication connections of the devices or units can be in electrical, mechanical, or other forms.
[0076] The units described as separate components may or may not be physically separated. The components displayed as units may or may not be physical units, that is, they can be located in one place or distributed to multiple network units. Some or all of the units can be selected according to actual needs to achieve the purpose of the solution of this embodiment.
[0077] In addition, in each embodiment of this application, the functional units can be integrated in a processing unit, or each unit can exist physically alone, or two or more units can be integrated in one unit.
[0078] If the function is implemented in the form of a software functional unit and sold or used as an independent product, it can be stored in a non-volatile computer-readable storage medium executable by a processor. Based on this understanding, the technical solution of this application, in essence, or the part that contributes to the prior art, or a part of this technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions for causing a computer device (which can be a personal computer, a server, or a network device, etc.) to execute all or part of the steps of the methods described in each embodiment of this application. The foregoing storage medium includes: various media such as USB flash drives, mobile hard disks, read-only memories (ROM), random access memories (RAM), magnetic disks, or optical discs that can store program codes.
[0079] Finally, it should be noted that the above-described embodiments are only specific embodiments of the present application, which are used to illustrate the technical solutions of the present application, rather than limiting them. The protection scope of the present application is not limited thereto. Although the present application has been described in detail with reference to the foregoing embodiments, those of ordinary skill in the art should understand that any person skilled in the art within the technical scope disclosed by the present application can still modify the technical solutions described in the foregoing embodiments, or can easily think of changes, or perform equivalent replacements on some of the technical features; and these modifications, changes or replacements do not make the essence of the corresponding technical solutions deviate from the spirit and scope of the technical solutions of the embodiments of the present application, and should all be covered within the protection scope of the present application. Therefore, the protection scope of the present application shall be subject to the protection scope of the claims.
Claims
1. A method for controlling an unmanned tractor, characterized in that: Applied to an unmanned tractor control system, the unmanned tractor control system includes a binocular stereo vision camera, the binocular stereo vision camera is arranged on the upper front outside of the driver's cabin of the unmanned tractor, and includes: Using the binocular stereo vision camera, a color image and a depth map in the driving direction of the unmanned tractor are obtained; Using the target detection model, identifying obstacles in the color image, and marking the identified target obstacles in the color image; Merging the color image with the target obstacle marked with the depth map to determine the position of the target obstacle; The unmanned tractor is controlled based on the position of the target obstacle.
2. The method according to claim 1, characterized in that The determining the position of the target obstacle comprises: Determining a depth value between the target obstacle and the unmanned tractor based on a fusion result of the color image and the depth map; According to the conversion relationship between the image coordinate system and the world coordinate system, the offset angle of the center of the target obstacle relative to the camera is determined; Based on the offset angle and the depth value, a position of the target obstacle is determined.
3. The method according to claim 2, characterized in that The determining the position of the target obstacle based on the offset angle and the depth value includes: The vertical position of the target obstacle is determined by multiplying the depth value by the tangent value of the offset angle; The product of the depth value and the cotangent value of the offset angle is determined as the horizontal position of the target obstacle.
4. The method according to claim 1, characterized in that: The method of using the binocular stereo vision camera to obtain a color image and a depth map in the driving direction of the unmanned tractor includes: Acquire a color image using the binocular stereo vision camera, and process the color image to obtain a grayscale image, wherein the grayscale image includes a first grayscale image and a second grayscale image; A depth map in the driving direction of the unmanned tractor is acquired based on the first grayscale image and the second grayscale image.
5. The method according to claim 4, characterized in that The step of acquiring a depth map in a driving direction of the unmanned tractor based on the first grayscale map and the second grayscale map includes: Determine a disparity value of each pixel point in the first grayscale image and the second grayscale image; A disparity map is generated based on the disparity value of each pixel, and a depth map is acquired according to the disparity map.
6. The method according to claim 1, characterized in that The method further comprises: The training data set is split to obtain multiple sets of training data subsets, each training data subset includes multiple training images; For each training data subset, the plurality of training images are randomly cut and then spliced to generate a training extended image corresponding to the training data subset; The target detection model is trained using the training extended image and the training data set.
7. The method according to claim 1, characterized in that The method further comprises: The position of the target obstacle in adjacent frame images is tracked by using a preset tracking algorithm, so as to continuously mark the target obstacle in different frame images.
8. An unmanned tractor control device, characterized in that: Applied to an unmanned tractor control system, the unmanned tractor control system includes a binocular stereo vision camera, the binocular stereo vision camera is arranged on the upper front outside of the driver's cabin of the unmanned tractor, and includes: An image acquisition module, used to acquire a color image and a depth map in the driving direction of the unmanned tractor using the binocular stereo vision camera; An obstacle recognition module, used to recognize obstacles in the color image using a target detection model, and mark the recognized target obstacles in the color image; A position calculation module, used to fuse the color image in which the target obstacle is marked with the depth map to determine the position of the target obstacle; The equipment control module is used to control the unmanned tractor based on the position of the target obstacle.
9. An electronic device, characterized in that: include: A processor, a memory and a bus, wherein the memory stores machine-readable instructions executable by the processor, and when the electronic device is running, the processor communicates with the memory via the bus, and the processor executes the machine-readable instructions to perform the steps of the unmanned tractor control method according to any one of claims 1 to 7.
10. A computer-readable storage medium, characterized in that: The computer-readable storage medium stores a computer program, and when the computer program is executed by a processor, the steps of the unmanned tractor control method according to any one of claims 1 to 7 are executed.