Farmland machinery navigation method based on visual correction

Abstract

The invention relates to the field of intelligent agriculture, and aims to provide a farmland machinery navigation method based on visual correction. Comprising the following steps: acquiring a real-time video in front of an advancing path by using a camera, taking an image center line as an initial operation navigation line, and extracting images in the video in real time according to a set time interval; after pixel coordinate correction is carried out on the image, crops and backgrounds in the image are separated based on feature colors of the target crops, and then binarization conversion is carried out on the image; calculating a crop distribution proportion, performing confidence evaluation, and judging the effectiveness of the image; performing optimization matching based on a sine template on the segmented image, confirming the position of the symmetric center of an actual crop row, and calculating to obtain an actual offset; and transmitting the offset to a lower computer for path correction. According to the invention, intelligent navigation and path correction of farmland mechanical equipment can be realized; the method still has strong robustness when the crop rows in the image are missing or deviated, and supports the navigation of various crop fields with different colors.

Description

Technical Field

[0001] The present invention relates to the field of intelligent agriculture, and in particular to a farm machinery navigation method based on visual deviation correction. Background Art

[0002] With the rapid development of smart agriculture and automation technology, farmland operation navigation technology is being used more and more widely in agricultural production. In order for farmland machinery to complete tasks such as automatic weeding, harvesting, fertilizing, pruning, and tilling in the operating environment, it is required to have the basic function of autonomous navigation, that is, farmland machinery perceives the surrounding environment through various sensors it carries, and after processing the perceived information, obtains its own position and posture information in real time, and then performs path planning and navigation control in real time to complete various operating tasks.

[0003] The satellite navigation system technology for farm machinery is generally mature and can improve the quality of work and extend the working time. The main advantages of this type of navigation system are that the structure is relatively simple, the technology is relatively mature, and its cost is constantly decreasing. The main problem is that satellite signals are sometimes greatly affected by the environment. When faced with complex terrain, severe weather conditions, random growth of crops and other variable environmental conditions, there are serious problems with navigation accuracy when relying solely on satellite navigation technology.

[0004] Compared with satellite navigation, machine vision navigation is more flexible, especially because machine vision images collect rich environmental information, a wide range, and complete target information. In farmland environments, crops are usually planted neatly in straight lines and parallel to each other. Therefore, the main task of machine vision navigation is to identify crop rows from images and detect path lines (navigation lines) for tracking, providing a basis for determining the relative position of agricultural machinery. However, due to the influence of many uncontrollable factors in farmland environments, such as changes in lighting, soil reflection, vegetation occlusion, and variable crop growth conditions, existing visual navigation technologies are often greatly affected by the external environment and have poor applicability and stability.

[0005] Therefore, new technical solutions need to be proposed to achieve accurate, stable and efficient navigation. Summary of the invention

[0006] The technical problem to be solved by the present invention is to overcome the deficiencies in the prior art and provide a farm machinery navigation method based on visual correction.

[0007] To solve the technical problem, the solution of the present invention is:

[0008] A farm machinery navigation method based on visual correction is provided, including the following contents:

[0009] (1) Use a camera fixed on the farm machinery to obtain real-time video of the front of the travel path, and use the center line of the image in the video as the initial operation navigation line; obtain the actual camera parameters and crop parameters, and extract the image in the video in real time according to the set time interval;

[0010] (2) Determine the transformation area based on the camera parameters and perform pixel coordinate correction on the image acquired in real time;

[0011] (3) Separate the crop and background in the image based on the characteristic color of the target crop, and then perform binary conversion on the image;

[0012] (4) Based on the distribution of crops in the binary image, calculate the crop distribution ratio and perform confidence assessment to determine the validity of the image;

[0013] (5) Determine the number of image segmentation regions based on crop parameters and perform uniform horizontal segmentation; then perform optimized matching based on a sinusoidal template for each segmented image to confirm the position of the symmetry center of the actual crop row in the image and calculate the actual offset;

[0014] (6) The offset is transmitted to the lower computer for path correction.

[0015] As a preferred embodiment of the present invention, the camera parameters in step (1) include: camera installation height, camera installation depression angle, camera horizontal field of view half angle, camera vertical field of view half angle; the crop parameters are derived from actual field crop conditions, including: actual crop row spacing, actual seedling strip spacing, actual crop width, crop color coding, the number of crop rows in the field of view, and the number of crop rows in the seedling strip.

[0016] As a preferred embodiment of the present invention, the step (2) specifically comprises:

[0017] (2.1) Calculate the process quantity according to the following formula:

[0018]

[0019]

[0020] (2.2) Determine the transformation area for correction based on the process quantity:

[0021]

[0022] (2.3) The coordinates of the four vertices of the transformation area are:

[0023] (2.4) Determine the target area of ​​the transformation:

[0024]

[0025] (2.5) Use the cv::getPerspectiveTransform() function in the standard image processing library OpenCV to calculate the transformation matrix M; perform the transformation as follows to obtain the corrected image;

[0026]

[0027] In the above formulas: The camera's vertical field of view half angle and the camera's horizontal field of view half angle Is the camera internal parameter; camera installation depression angle and camera installation height is the camera extrinsic parameter; is the image width, is the image height; is the bottom distortion ratio, is the top distortion ratio; is the image pixel coordinate before transformation; is the pixel coordinate of the transformed image.

[0028] As a preferred embodiment of the present invention, the step (3) specifically comprises:

[0029] Based on the hue, saturation and brightness of the target crop in the image, the crop color is encoded; according to the encoding difference, the image is extracted using the inRange() function in OpenCV, and then the image is converted into a binary image in which the crop is separated from the background.

[0030] As a preferred solution of the present invention, in the binary image, the crop is a white pixel, and the coordinates in the image matrix are Corresponding value ; The background is black pixels, with coordinates in the image matrix Corresponding value .

[0031] As a preferred embodiment of the present invention, the step (4) specifically comprises:

[0032] (4.1) The confidence score is calculated according to the following formula: :

[0033]

[0034]

[0035]

[0036] In the above formulas: is the number of crop rows; is the total width of the row of white pixels corresponding to the crop in the binary image; is the estimated line width obtained through image analysis; is the actual line width calculated by angle; is the actual crop width;

[0037] (4.2) If the confidence score ∈[0.5,1], the binarized image is considered valid; otherwise, it is considered invalid and the confidence evaluation and judgment are continued for the next binarized image.

[0038] As a preferred embodiment of the present invention, the step (5) specifically comprises:

[0039] (5.1) Determine the value of the number of image segmentation regions roinum according to the camera field of view, and then divide the image evenly horizontally, roinum∈[5,10];

[0040] (5.2) Constructing crop row distribution template based on sine function ,by For the template symmetry center coordinates, calculate the crop parameters in the image:

[0041]

[0042]

[0043]

[0044]

[0045] and All are process quantities:

[0046]

[0047]

[0048] In the above formulas: is the spacing between seedling strips in the image; is the actual spacing between seedling strips; is the crop row spacing in the image; is the actual crop row spacing; is the actual crop width; is the number of seedling strips; is the number of crop rows in the field of view; The number of crop rows in each strip; is the seedling strip count variable used for summation; is the crop row count variable used for summing; is the gate function, used to filter the target area; is the offset parameter, used to adjust the template height;

[0049] (5.3) For each image region after segmentation, for each horizontal coordinate i, accumulate the counts of all horizontal coordinates White pixels , traverse i from 1 to W, and get the length of the image width The crop row distribution series ; Use the gradient descent method to distribute the crop row template and crop row distribution series Fitting is performed and the position with the smallest offset between the two is obtained by adjusting the value of midaxis:

[0050]

[0051] The midaxis value at this time is used as the abscissa, and the mean ordinate value of the region is used as the ordinate to obtain the coordinate position of the symmetry center of the actual crop row in the image region.

[0052] As a preferred solution of the present invention, the offset between the crop row distribution template and the crop row distribution series is determined in the following manner:

[0053]

[0054]

[0055]

[0056]

[0057]

[0058]

[0059]

[0060] In the above formulas: is the center position of the crop row at the bottom of the image, The vertical coordinate of the image is The center position of the crop row; The four standard parameters are obtained by fitting the symmetric center coordinates of each area of ​​the crop row using the standard library function fitline(); is the image center offset, The bottom offset of the image; is the image center projection ratio, is the bottom projection ratio of the image; is the offset.

[0061] As a preferred embodiment of the present invention, the step (6) specifically comprises:

[0062] The offset Transmitted to the lower computer, according to the offset Sending adaptive control signals to the driving actuators to correct the travel path of the farm machinery.

[0063] The present invention further provides a computer device, comprising: at least one processor, and a memory communicatively connected to the at least one processor, wherein the memory stores instructions executed by the at least one processor, and the instructions are executed by the at least one processor so that the at least one processor executes the aforementioned farm machinery navigation method based on visual correction.

[0064] The present invention also provides a computer-readable storage medium, wherein the computer-readable storage medium stores computer instructions, and the computer instructions are used to enable the computer to execute the aforementioned farm machinery navigation method based on visual correction.

[0065] Compared with the prior art, the present invention has the following beneficial effects:

[0066] 1. The method provided by the present invention can enable autonomous driving farm machinery and equipment to achieve intelligent navigation and path correction in the farmland.

[0067] 2. Compared with the existing crop row recognition technology based on Hough transform, the present invention uses a template fitting algorithm in the calculation, which still has strong robustness when the crop rows in the image are missing or offset, and will not cause recognition errors due to the missing of some crop rows.

[0068] 3. Compared with the existing crop row recognition technology based on deep learning, the present invention adopts the traditional visual algorithm, which has less computational complexity and can run at a high frame rate on agricultural machinery.

[0069] 4. Due to the unique color screening algorithm based on the HSV color space, the present invention can also support navigation of crop fields of various colors. BRIEF DESCRIPTION OF THE DRAWINGS

[0070] Figure 1 A system architecture diagram of a navigation method operating environment shown in an embodiment of the present invention;

[0071] Figure 2 This is a diagram illustrating crop parameters of the farm machinery in the present invention when operating in the field;

[0072] Figure 3 It is a diagram illustrating the image segmentation method in the present invention;

[0073] Figure 4 This is a flow chart of a farm machinery navigation method according to an embodiment of the present invention;

[0074] Figure 5 is a binary image containing crops and background;

[0075] Figure 6 This is the position diagram of the actual crop rows in the original image after fitting optimization processing.

[0076] Reference numerals: crop 110 , camera 120 , processor 130 , agricultural machine 140 . DETAILED DESCRIPTION

[0077] First of all, it should be explained that the present invention relates to image processing technology, which is an application of computer technology in the field of autonomous driving technology. In the process of implementing the present invention, the application of multiple software functional modules will be involved. The applicant believes that after carefully reading the application documents and accurately understanding the implementation principle and purpose of the present invention, in combination with the existing known technology, those skilled in the art can fully use their software programming skills to implement the present invention. All those mentioned in the application documents of the present invention belong to this category, and the applicant will not list them one by one.

[0078] Example embodiments will be described more fully hereinafter with reference to the accompanying drawings.

[0079] An exemplary embodiment of the present invention discloses a farm machinery navigation method based on visual deviation correction. Figure 1 The system architecture diagram of the operating environment of the navigation method is shown.

[0080] refer to Figure 1 As shown, the system architecture 100 includes crops 110, cameras 120, processors 130, and agricultural machinery 140. Among them, the crop 100 can be any crop, the camera 120 can be any device with an image acquisition function, the processor 130 can exist independently or attached to the camera 120 or the agricultural machinery 140 in various forms, and the agricultural machinery 140 can be any device with a mobile function. The crop 110 is located within the sensing range of the camera 120, and the area includes but is not limited to farmland, mountains, and any other area where crops can be grown. The camera 120 and the processor 130 can be connected by wire or wireless means, so that the image taken by the camera is transmitted to the processor for calculation and processing. The agricultural machinery 140 is a device connected to the processor 130 by wire or wireless means, and moves based on the signal transmitted by the processor.

[0081] The farm machinery navigation method based on visual deviation correction according to this exemplary embodiment is described below.

[0082] The basic application of the method includes: the user is in the farmland, and after completing the basic connection between the camera 120, the processor 130, and the agricultural machinery 140, the camera 120 is used to shoot the current condition of the farmland, and then the farm machinery navigation method of this exemplary embodiment is executed to obtain the direction offset data of the agricultural machinery 140, which is transmitted to the agricultural machinery 140 by the processor 130, and the agricultural machinery 140 performs a steering operation to achieve the purpose of correcting the direction.

[0083] Figure 4 An exemplary process of a farm machinery navigation method with visual correction is shown, and each step is described in detail below.

[0084] 1. Use the camera fixed on the farm machinery to obtain real-time video in front of the travel path, and use the center line of the image in the video as the initial operation navigation line; obtain the actual camera parameters and crop parameters, and extract the image in the video in real time according to the set time interval.

[0085] The camera parameters include: camera installation height, camera installation pitch, camera horizontal field of view, camera vertical field of view; the crop parameters are the actual crop parameters in the field, including: crop row spacing, seedling strip spacing, crop width, crop color coding, the number of crop rows in the field of view, the number of crop rows in the seedling strip, and the center of the seedling strip spacing is used as the symmetry center of the crop row. The crop row spacing can be defined according to the actual field crop planting conditions. The naming instructions of some crop parameters are as follows: Figure 2 shown.

[0086] Since the internal and external parameters of the camera basically do not change during operation, the camera parameters are generally input manually once, without repeated measurements. Even if the crop type or crop batch is changed, historical data can be directly called.

[0087] Since field crops are generally sown in batches, the growth conditions of crops in the same batch are basically the same. Therefore, they can be measured manually before operation, and the measurement can be input into the control program for direct calculation in the navigation process of the present invention; when operating on the same batch of crops, it can be imported from the historical database. Only when switching crop types or changing batch operations, it is necessary to re-measure and input crop parameters. Since the allowable error has been considered when designing the calculation method in the navigation process of the present invention, the crop parameters do not need to be very accurate.

[0088] Extracting images from real-time video at set time intervals for subsequent processing can significantly reduce the amount of data processing and improve the computational efficiency of farm machinery navigation and correction.

[0089] 2. Determine the transformation area according to the camera parameters and perform pixel coordinate correction on the real-time acquired image.

[0090] Each camera has different intrinsic parameters, and the installation angle of each camera can be adjusted manually. In order to facilitate processing after the camera acquires the image, the image needs to be corrected according to the camera's own parameters (intrinsic parameters) and its installation parameters (extrinsic parameters). After performing calculations based on the physical space angle height, the image is corrected through the functions in the OpenCV library. The details are as follows:

[0091] (2.1) Calculate the process quantity according to the following formula:

[0092]

[0093]

[0094] (2.2) Determine the transformation area for correction based on the process quantity:

[0095]

[0096] (2.3) The coordinates of the four vertices of the transformation area are:

[0097] (2.4) Determine the target area of ​​the transformation:

[0098]

[0099] (2.5) Use the cv::getPerspectiveTransform() function in the standard image processing library OpenCV to calculate the transformation matrix M; then perform the transformation as follows to obtain the corrected image;

[0100]

[0101] In the above formulas: The camera's vertical field of view half angle and the camera's horizontal field of view half angle Is the camera internal parameter; camera installation depression angle and camera installation height is the camera extrinsic parameter; is the image width, is the image height; is the bottom distortion ratio, is the top distortion ratio; is the image pixel coordinate before transformation; is the pixel coordinate of the transformed image.

[0102] 3. Based on the characteristic color of the target crop, separate the crop and background in the image and perform binary conversion on the image.

[0103] After the image is corrected, the color type of the crop is first encoded in the HSV (hue, saturation, value) space according to the characteristic color of the target crop. As an example, the encoding includes but is not limited to: 0-green, 1-blue, 2-red, 3-green + yellow, 4-green + red, 5-green + blue. Then, based on the encoding difference, the inRange() function in the OpenCV library is used to extract the crop image to separate the crop from the background. The image is transformed to obtain a binary image (such as Figure 5 In the figure, the crop is the white pixel, whose coordinates in the image matrix are Corresponding value ; The background is black pixels, whose coordinates in the image matrix are Corresponding value .

[0104] 4. According to the distribution of crops in the binary image, calculate the crop distribution ratio and perform confidence assessment to determine the validity of the image.

[0105] (4.1) The confidence score is calculated according to the following formula: :

[0106]

[0107]

[0108]

[0109] In the above formulas: is the number of crop rows; is the total width of the row of white pixels corresponding to the crop in the binary image; is the estimated line width obtained through image analysis; is the actual line width calculated by angle;

[0110] (4.2) If the confidence score ∈[0.5,1], the binarized image is considered valid; otherwise, it is considered invalid and the confidence evaluation and judgment are continued for the next binarized image.

[0111] If the image is determined to be valid, the image processing operation continues.

[0112] 5. Optimize and match the effective binary image based on the sinusoidal template, confirm the position of the actual crop row in the image, and calculate the actual offset;

[0113] (5.1) According to the camera field of view, determine the appropriate number of image segmentation regions roinum and perform uniform horizontal segmentation.

[0114] The more horizontal segmentation areas, the more conducive it is to tracking the deviation of the navigation line. However, considering that too many segmentation areas will increase the amount of calculation and thus prolong the calculation time, roinum∈[5,10] is generally taken. The specific method of horizontal segmentation is as follows: Figure 3 shown.

[0115] (5.2) According to the actual crop growth conditions, if the white pixels of the crops in the binary image are counted vertically, the obtained statistical data often presents a sinusoidal distribution; therefore, the present invention adopts a template defined based on a sinusoidal function to fit the crop rows to eliminate errors.

[0116] Construction of crop row distribution template based on sine function , The symmetric center coordinates of the template are used to calculate the crop-related parameters in the image:

[0117]

[0118]

[0119]

[0120] Then there is

[0121]

[0122] in For process quantity:

[0123]

[0124] in For process quantity:

[0125]

[0126] In the above formulas: is the spacing between seedling strips in the image; is the actual spacing between seedling strips; is the crop row spacing in the image; is the actual crop row spacing; is the actual crop width; is the number of seedling strips; is the number of crop rows in the field of view; The number of crop rows in each strip; is the seedling strip count variable used for summation; is the crop row count variable used for summing; is the gate function, used to filter the target area; is the offset parameter, used to adjust the template height;

[0127] (5.3) For each image region after segmentation, for each horizontal coordinate i, accumulate the counts of all horizontal coordinates White pixels , traverse i from 1 to W, and get the length of the image width The crop row distribution series ; Use the gradient descent method to distribute the crop row template and crop row distribution series Fitting is performed and the position with the smallest offset between the two is obtained by adjusting the value of midaxis:

[0128]

[0129] The midaxis value at this time is used as the abscissa, and the mean ordinate value of the region is used as the ordinate to obtain the coordinate position of the symmetry center of the actual crop row in the image region.

[0130] Figure 6 This is the position diagram of the actual crop rows in the original image after fitting optimization processing. The three red lines in the figure represent the symmetry centers of the three crop rows after optimization processing.

[0131] Furthermore, the offset between the crop row distribution template and the crop row distribution series is determined according to the following method:

[0132]

[0133]

[0134]

[0135]

[0136]

[0137]

[0138]

[0139] In the above formulas: is the center position of the crop row at the bottom of the image, The vertical coordinate of the image is The center position of the crop row; The four standard parameters are obtained by fitting the symmetric center coordinates of each area of ​​the crop row using the standard library function fitline(); is the image center offset, The bottom offset of the image; is the image center projection ratio, is the bottom projection ratio of the image; is the offset.

[0140] 6. Set the offset Transmitted to the lower computer for path correction.

[0141] The offset Transmitted to the lower computer, according to the offset Sending adaptive control signals to the driving actuators to correct the travel path of the farm machinery.

[0142] Those skilled in the art to which the present invention belongs will appreciate that the various aspects of the present invention may be specifically implemented in the following forms, namely: a complete hardware implementation, a complete software implementation (including firmware, microcode, etc.), or a combination of hardware and software implementations, which may be collectively referred to as "modules" or "systems". This application is intended to cover any variations, uses, or adaptive changes that follow the general principles of the present invention and include common knowledge or customary technical means in the art that are not disclosed in the present invention. The description and implementations are to be regarded as exemplary only, and the true scope and spirit of the present invention are indicated by the claims.

Claims

1. A farm machinery navigation method based on visual correction, characterized in that: Includes the following: (1) Use a camera fixed on the farm machinery to obtain real-time video of the front of the travel path, and use the center line of the image in the video as the initial operation navigation line; Get the actual camera parameters and crop parameters, and extract images from the video in real time according to the set time interval; (2) Determine the transformation area based on the camera parameters and perform pixel coordinate correction on the image acquired in real time; (3) Separate the crop and background in the image based on the characteristic color of the target crop, and then perform binary conversion on the image; (4) Based on the distribution of crops in the binary image, calculate the crop distribution ratio and perform confidence assessment to determine the validity of the image; (5) Determine the number of image segmentation regions based on crop parameters and perform uniform horizontal segmentation; then perform optimized matching based on a sinusoidal template for each segmented image to confirm the position of the symmetry center of the actual crop row in the image and calculate the actual offset; (6) The offset is transmitted to the lower computer for path correction.

2. The method according to claim 1, characterized in that: The camera parameters in step (1) include: camera installation height, camera installation depression angle, camera horizontal field of view half angle, camera vertical field of view half angle; the crop parameters are derived from actual field crop conditions, including: actual crop row spacing, actual seedling strip spacing, actual crop width, crop color coding, the number of crop rows in the field of view, and the number of crop rows in the seedling strip.

3. The method according to claim 1, characterized in that: The step (2) specifically includes: (2.1) Calculate the process quantity according to the following formula: ; ; (2.2) Determine the transformation area for correction based on the process quantity: ; (2.3) The coordinates of the four vertices of the transformation area are: ; (2.4) Determine the target area of ​​the transformation: ; (2.5) Use the cv::getPerspectiveTransform() function in the standard image processing library OpenCV to calculate the transformation matrix M; perform the transformation as follows to obtain the corrected image; ; In the above formulas: The camera's vertical field of view half angle and the camera's horizontal field of view half angle Is the camera internal parameter; camera installation depression angle and camera installation height is the camera extrinsic parameter; is the image width, is the image height; is the bottom distortion ratio, is the top distortion ratio; is the image pixel coordinate before transformation; is the pixel coordinate of the transformed image.

4. The method according to claim 1, characterized in that: The step (3) specifically includes: Based on the hue, saturation and brightness of the target crop in the image, the crop color is encoded; according to the encoding difference, the image is extracted using the inRange() function in OpenCV, and then the image is converted into a binary image in which the crop is separated from the background.

5. The method according to claim 1, characterized in that: In the binary image, the crop is white pixels, with coordinates in the image matrix Corresponding value ; The background is black pixels, with coordinates in the image matrix Corresponding value .

6. The method according to claim 1, characterized in that: The step (4) specifically includes: (4.1) The confidence score is calculated according to the following formula: : ; ; ; In the above formulas: is the number of crop rows; is the total width of the row of white pixels corresponding to the crop in the binary image; is the estimated line width obtained through image analysis; is the actual line width calculated by angle; is the actual crop width; (4.2) If the confidence score ∈[0.5,1], the binarized image is considered valid; otherwise, it is considered invalid and the confidence evaluation and judgment are continued for the next binarized image.

7. The method according to claim 1, characterized in that: The step (5) specifically includes: (5.1) Determine the value of the number of image segmentation regions roinum according to the camera field of view, and then divide the image evenly horizontally, roinum∈[5,10]; (5.2) Based on the segmented image, a crop row distribution template is constructed based on the sine function ,by Compute crop parameters in the image for the template symmetry center coordinates: ; ; ; ; and All are process quantities: ; ; In the above formulas: is the spacing between seedling strips in the image; is the actual spacing between seedling strips; is the crop row spacing in the image; is the actual crop row spacing; is the actual crop width; is the number of seedling strips; is the number of crop rows in the field of view; The number of crop rows in each strip; is the seedling strip count variable used for summation; is the crop row count variable used for summing; is the gate function, used to filter the target area; is the offset parameter, used to adjust the template height; (5.3) For each image region after segmentation, for each horizontal coordinate i, accumulate the counts of all horizontal coordinates White pixels , traverse i from 1 to W, and get the length of the image width The crop row distribution series ; Use the gradient descent method to distribute the crop row template and crop row distribution series Fitting is performed and the position with the smallest offset between the two is obtained by adjusting the value of midaxis: ; The midaxis value at this time is used as the abscissa, and the mean ordinate value of the region is used as the ordinate to obtain the coordinate position of the symmetry center of the actual crop row in the image region.

8. The method according to claim 7, characterized in that: Determine the offset between the crop row distribution template and the crop row distribution array as follows: ; ; ; ; ; ; ; In the above formulas: is the center position of the crop row at the bottom of the image, The vertical coordinate of the image is The center position of the crop row; The four standard parameters are obtained by fitting the symmetric center coordinates of each area of ​​the crop row using the standard library function fitline(); is the image center offset, The bottom offset of the image; is the image center projection ratio, is the bottom projection ratio of the image; is the offset.

9. The method according to claim 1, characterized in that: The step (6) specifically includes: The offset Transmitted to the lower computer, according to the offset Sending adaptive control signals to the driving actuators to correct the travel path of the farm machinery.

10. A computer device, characterized in that: include: At least one processor, and a memory communicatively connected to the at least one processor, wherein the memory stores instructions executed by the at least one processor, and the instructions are executed by the at least one processor so that the at least one processor executes the farm machinery navigation method based on visual correction as described in any one of claims 1 to 8.

11. A computer-readable storage medium, characterized in that: The computer-readable storage medium stores computer instructions, and the computer instructions are used to enable the computer to execute the farm machinery navigation method based on visual correction as described in any one of claims 1 to 8.

Images