A white radish harvester row recognition system and method based on machine vision and a harvester

Abstract

The present invention provides a machine vision-based white radish harvester row recognition system and method, and a harvester, comprising a control mechanism, an image acquisition mechanism, a plucking and harvesting device, a vehicle-mounted chassis, and a row adjustment device. The plucking and harvesting device and the row adjustment device are mounted on the vehicle-mounted chassis. The plucking and harvesting device is used to plucking and harvesting white radishes. The adjustment devices are located on both sides of the plucking and harvesting device and are used to adjust the left and right horizontal translation of the plucking and harvesting device. The control mechanism performs image processing to extract ridge boundary lines and white radish ridge row lines, uses the extracted ridge boundary lines and ridge row lines for verification, obtains a target navigation line, and controls the row adjustment device to adjust the left and right horizontal translation of the plucking and harvesting device for row alignment. The present invention can automatically adjust the harvesting components to harvest in row order, eliminating manual row alignment operations, reducing the driver's labor intensity, helping to ensure the harvest quality of white radishes under long-term operation, and improving the intelligent and efficient harvesting level of the white radish harvester.

Description

Technical Field

[0001] The present invention belongs to the technical field of root vegetable harvesting equipment, and in particular relates to a machine vision-based white radish harvester row recognition system and method and a harvester, which are mainly used for automatic row harvesting of white radish. Background Art

[0002] White radish is a common root vegetable rich in nutrients and widely cultivated in North China, Northeast my country, Northwest China, and Southwest China. Its production is highly seasonal and labor-intensive. Currently, white radish harvesting in my country relies primarily on manual labor, with a limited use of mechanical harvesting. Manual harvesting is labor-intensive and inefficient, while mechanical harvesting typically involves single-row pulling, requiring the driver to monitor the rows while driving.

[0003] The pull-and-pull combined harvester is a highly efficient method for harvesting radishes, completing multiple processes simultaneously, including loosening, plucking, conveying, cutting, and collecting. However, due to the unique growth patterns and row spacing of radishes, frequent manual adjustments to the rows are required during operation, which is crucial for ensuring harvest quality. However, this work requires long hours of manual labor, is labor-intensive, complex, and dangerous for the operator, and the quality of the harvest depends on the operator's experience.

[0004] Therefore, research on automatic row-aligned harvesting technology for pull-type white radish harvesters, achieving integrated automatic row alignment and adjustment operations, can reduce the driver's workload and the cost of white radish cultivation, and improve the intelligent and efficient harvesting of white radishes. Therefore, automatic row alignment for white radish harvesting is an important technology to ensure the long-term operation performance of white radish harvesters, reduce the workload of manual operation, and improve the intelligent and efficient harvesting of white radishes. Summary of the Invention

[0005] In response to the above technical problems, the present invention provides a white radish harvester row recognition system and method and a harvester based on machine vision, which can automatically identify the center line of the white radish planting row, obtain the position information of the center line of the planting row and adjust the harvesting components to harvest in the row, avoiding manual row operation, reducing the labor intensity of the driver, helping to ensure the harvest quality of white radish under long-term operation conditions, and improving the intelligent and efficient harvesting level of the white radish harvester.

[0006] Note that the inclusion of these objectives does not preclude the existence of other objectives. One embodiment of the present invention does not necessarily achieve all of the above objectives. Objectives other than the above objectives may be extracted from the description of the specification, drawings, and claims.

[0007] The present invention achieves the above technical objectives through the following technical means.

[0008] A machine vision-based row recognition system for a white radish harvester comprises a control mechanism, an image acquisition mechanism, a plucking and harvesting device, a vehicle-mounted chassis, and a row adjustment device; the plucking and harvesting device and the row adjustment device are mounted on the vehicle-mounted chassis, the plucking and harvesting device is used for plucking and harvesting white radishes, and the adjustment devices are located on both sides of the plucking and harvesting device and are used for adjusting the left and right horizontal translation of the plucking and harvesting device; the control mechanism is respectively connected to the image acquisition mechanism and the row adjustment device, the image acquisition mechanism is used for acquiring images of the white radish planting field in front of the harvester to obtain depth images and color images, and transmit the images to a control unit; the control mechanism performs image processing, extracts ridge boundary lines and white radish ridge row lines, uses the extracted ridge boundary lines and ridge row lines for verification, obtains a target navigation line, and controls the row adjustment device according to the target navigation line to adjust the left and right horizontal translation of the plucking and harvesting device for row alignment.

[0009] In the above scheme, the row adjustment device includes a first heavy-duty screw guide, a first displacement sensor, a stepper motor, a second heavy-duty screw guide and a second displacement sensor; the stepper motor includes a first stepper motor and a second stepper motor; the first heavy-duty screw guide and the first displacement sensor are installed in parallel at the front end of one side of the vehicle chassis, the first guide rail slide is installed on the first heavy-duty screw guide, the heavy-duty screw guide is connected to the first stepper motor, and the side end face of the first guide rail slide is connected to the measuring end of the first displacement sensor; the second heavy-duty screw guide and the second displacement sensor are installed in parallel on the upper end face of the mounting frame on the vehicle chassis, the second guide rail slide is installed on the second heavy-duty screw guide, and the second heavy-duty screw guide is connected to the second stepper motor The upper end of the second guide rail slide is hinged to the frame of the clamping and conveying mechanism of the plucking and harvesting device, and the side end surface of the second guide rail slide is connected to the measuring end of the second displacement sensor; the first stepper motor, the first displacement sensor, the second displacement sensor and the second stepper motor are respectively connected to the control mechanism, and the control mechanism controls the first stepper motor to drive the first heavy-loaded screw guide rail to slide on the first guide rail slide, and controls the second stepper motor to drive the second heavy-loaded screw guide rail to slide on the second guide rail slide. The first displacement sensor is used to measure the position of the first guide rail slide and feed it back to the control mechanism, and the second displacement sensor is used to measure the position of the second guide rail slide and feed it back to the control mechanism, so as to drive the overall horizontal translation of the plucking and harvesting device to realize the horizontal row adjustment of the plucking and harvesting device.

[0010] In the above scheme, the control mechanism includes an upper computer and a lower computer; the upper computer is connected to the image acquisition mechanism, and is used to receive the images acquired by the image acquisition mechanism and perform image processing, extract the ridge boundary line and the white radish ridge line, and use the extracted ridge boundary line and ridge line for verification to obtain the navigation tracking point navigation line; the lower computer controls the row adjustment device to adjust the harvesting device for row adjustment according to the navigation line of the upper computer.

[0011] In the above solution, the control mechanism extracts the ridge boundary line based on the outermost ridge boundary line extraction algorithm of the depth image:

[0012] The control mechanism intercepts the target area and performs interception processing on the depth image:

[0013]

[0014] Among them, B (i,j) is the coordinate of the target area point in the intercepted depth image, I (i,j) is the coordinate of the corresponding pixel in the input original depth image, x, y are the coordinates of the upper left corner of the target area, W, H are the width and height of the target area respectively;

[0015] The fixed threshold binarization method is used to segment the intercepted depth image to obtain the outermost ridge boundary line and obtain a binary image:

[0016]

[0017] Among them, 0 represents a black background and 255 represents a white foreground;

[0018] Perform area filtering and morphological processing on the processed binary image to eliminate noise and edge hole interference, then read the coordinates of the first non-zero point from right to left in each row in the intercepted area and store them in the matrix P:

[0019]

[0020] Where (i k ,j k ) are the coordinates of points that are not 0;

[0021] The least squares method is used to fit the extracted points to obtain the ridge boundary line:

[0022] y=ax+b

[0023] Where a and b are the slope and intercept of the ridge boundary line extracted from the depth map, respectively.

[0024] In the above scheme, the control mechanism is based on the method for extracting the planting lines of white radish in the color picture of straight line clustering:

[0025] The super green feature grayscale method is used to enhance the green area features. First, the target interest area of ​​the color image is intercepted, and the remaining areas are set to 0. The super green feature method is used for grayscale conversion. The calculation method is:

[0026]

[0027] Among them, 2 means expanding the green channel pixels by 2 times, G means green channel, R means red channel, and B means blue channel;

[0028] Use Otsu method to perform image binarization processing;

[0029] The binary image is first dilated and then eroded to obtain the area of ​​each connected domain, and then the connected domain area filtering is performed;

[0030] Perform straight line clustering to extract planting row lines: select ridge lines in the intercepted area, distribute them evenly, calculate the distance between each white point in the binary image and the straight line (ridge line), compare them, classify the ones with the closest distance into the same category as the straight line, and obtain clustered straight lines by fitting:

[0031]

[0032] In the above solution, the navigation tracking point and navigation line are obtained by the tracking point fusion extraction algorithm:

[0033] First, the ridge boundary navigation line extracted from the depth image is converted according to the coordinate system, translated by half the standard ridge width distance L in the real coordinate system, and the translated straight line is recorded in the pixel coordinate system;

[0034] Then, the ridge boundary navigation line after translation and the ridge line extracted from the color image are calculated according to the coefficient of determination R. 2 And the comparison coefficient g is fused according to the normalization method, and the weight coefficients r1 and r2 are assigned. r1 is the weight coefficient assigned to the depth map ridge line, and r2 is the weight coefficient assigned to the color map ridge line:

[0035]

[0036] According to the weight coefficient fusion, the final target navigation line is obtained:

[0037] y=(k1r1+k2r2)x+(r1b1+r2b2)

[0038] Among them, k1 and b1 are the slope and intercept of the ridge line in the depth map, and k2 and b2 are the slope and intercept of the ridge line extracted from the color map.

[0039] A harvester comprises the white radish harvester row recognition system based on machine vision.

[0040] A control method for a row recognition system of a white radish harvester based on machine vision comprises the following steps:

[0041] The image acquisition mechanism collects images of the white radish planting field in front of the harvester to obtain depth images and color images, and transmits them to the control unit; the control mechanism performs image processing, extracts ridge boundary lines and white radish ridge lines, uses the extracted ridge boundary lines and ridge lines for verification, obtains target navigation lines, and controls the row adjustment device to adjust the harvesting device to move horizontally left and right according to the target navigation lines to harvest the rows.

[0042] In the above scheme, the row harvesting specifically includes the following steps:

[0043] S1. Start the equipment and reset the harvesting device to the set initial starting point through the row adjustment device. Check the working status of the equipment and confirm whether the safety function can work.

[0044] S2. Adjust the pulling height of the end of the pulling and harvesting device through the hydraulic cylinder to prepare for harvesting;

[0045] S3, adjusting the field of view of the image acquisition mechanism to capture images of the white radish planting field in front of the harvester to obtain a depth image and a color image;

[0046] S4, performing interception, filtering, expansion and corrosion processing on the depth image and color image obtained by shooting;

[0047] S5, extracting ridge boundary lines and ridge planting row lines from the pre-processed depth image and color image respectively;

[0048] S6. Fusing the extracted ridge boundary line and ridge row line to obtain a target navigation line, thereby obtaining a navigation tracking point;

[0049] S7, converting the navigation tracking point in the image coordinate system into a point in the real coordinate system, and calculating the position error of the harvesting device relative to the point;

[0050] S8, the lower computer uses the position error as input and adopts the fuzzy control algorithm to control the movement of the line adjustment device to complete the line adjustment. After the adjustment is completed, the above process is repeated to refresh and realize real-time line adjustment;

[0051] S9. After the target row is harvested, proceed to the next row for harvesting.

[0052] In the above solution, the control mechanism uses a fuzzy control algorithm to control the row adjustment device to adjust the row, including the following steps:

[0053] The position error between the current position and the target position is set as the input, and the pulse frequency of the stepper motor is set as the output. The position error is fuzzified and divided into fuzzy sets. A set of fuzzy rule tables is established to map the fuzzified input to the fuzzified output. Fuzzy reasoning is performed based on the input fuzzy variables and fuzzy rules to obtain fuzzy output, which is then converted into a specific pulse frequency. By adjusting the pulse frequency, the line adjustment speed can be controlled.

[0054] Compared with the prior art, the present invention has the following beneficial effects:

[0055] 1. Improve harvesting efficiency: Through machine vision technology, automatic row-by-row harvesting of white radish planting rows is achieved, reducing manual operations and greatly improving harvesting efficiency.

[0056] 2. Reduce labor intensity: It reduces the labor intensity of the driver and no longer requires the driver to perform frequent manual adjustments, thereby reducing operational risks and the driver's workload.

[0057] 3. Improve harvest quality: Automatic row harvesting can ensure the accuracy and consistency of harvesting, avoid errors caused by human operation, and thus improve the harvest quality of white radish.

[0058] 4. Intelligent operation: It adopts a combination of upper and lower computers, combines image processing algorithms to identify planting rows in real time, and adjusts rows through fuzzy control algorithms, making the harvesting process more intelligent and automated.

[0059] 5. Realize long-term continuous operation: Through automatic row harvesting technology, the white radish harvester can realize long-term continuous operation, improving the continuity and stability of the operation.

[0060] In summary, the present invention can automatically identify the center line of the white radish planting row, obtain the position information of the center line of the planting row and adjust the harvesting components to harvest the row, thereby improving the harvesting efficiency, reducing labor intensity, improving the harvesting quality, intelligent operation and realizing long-term continuous operation and other advantages.

[0061] Note that the description of these effects does not preclude the existence of other effects. One embodiment of the present invention does not necessarily have all of the above effects. Effects other than the above can be clearly seen and extracted from the description of the specification, drawings, claims, etc. BRIEF DESCRIPTION OF THE DRAWINGS

[0062] Figure 1 It is a schematic diagram of coordinate system conversion according to one embodiment of the present invention.

[0063] Figure 2 It is a schematic axial view of the entire machine according to one embodiment of the present invention.

[0064] Figure 3 It is a schematic diagram of the process of ridge boundary extraction based on depth image according to one embodiment of the present invention.

[0065] Figure 4 It is a schematic diagram of the process of extracting planting lines based on straight line clustering according to one embodiment of the present invention.

[0066] Figure 5 It is a schematic diagram of the principle of selecting navigation line tracking points in one embodiment of the present invention.

[0067] Figure 6 It is a fuzzy control rule table according to one embodiment of the present invention.

[0068] Figure 7 This is a system hardware diagram of an embodiment of the present invention.

[0069] Figure 8 This is a workflow diagram for row harvesting according to one embodiment of the present invention.

[0070] In the figure: 1. Harvesting device, 2. Vehicle-mounted chassis, 3. Row adjustment device, 4. Mounting frame, 5. Motor, 6. Gantry, 7. Camera, 10. Clamping conveyor belt, 11. Supporting mechanism, 12. Clamping conveying mechanism, 13. Hydraulic cylinder, 14. First guide rail slide, 15. First heavy-duty screw guide, 16. First stepper motor, 17. First displacement sensor, 18. Collecting bucket, 19. Cutting mechanism, 20. Frame, 21. Transmission mechanism, 22. Motor, 23. Second displacement sensor, 24. Second stepper motor, 25. Second heavy-duty screw guide, 26. Second guide rail slide. DETAILED DESCRIPTION

[0071] The following describes embodiments of the present invention in detail, examples of which are shown in the accompanying drawings, wherein the same or similar reference numerals throughout represent the same or similar elements or elements having the same or similar functions. The embodiments described below with reference to the accompanying drawings are exemplary and are intended to be used to explain the present invention, and are not to be construed as limiting the present invention.

[0072] In the description of the present invention, it should be understood that the terms "center", "longitudinal", "lateral", "length", "width", "thickness", "front", "back", "left", "right", "up", "down", "axial", "radial", "vertical", "horizontal", "inside", "outside" and the like indicate orientations or positional relationships based on the orientations or positional relationships shown in the accompanying drawings, and are only for the convenience of describing the present invention and simplifying the description, rather than indicating or implying that the device or element referred to must have a specific orientation, be constructed and operated in a specific orientation, and therefore cannot be understood as limiting the present invention. In addition, the terms "first" and "second" are only used for descriptive purposes and cannot be understood as indicating or implying relative importance or implicitly indicating the number of the indicated technical features. Therefore, the features defined as "first" and "second" may explicitly or implicitly include one or more of the features. In the description of the present invention, "multiple" means two or more, unless otherwise clearly and specifically defined.

[0073] In the present invention, unless otherwise expressly specified or limited, the terms "mounted," "connected," "connect," "fixed," etc. should be understood broadly. For example, they may refer to fixed, detachable, or integral connections; mechanical or electrical connections; direct or indirect connections through an intermediary; or internal communication between two components. Those skilled in the art will understand the specific meanings of the above terms in the present invention based on specific circumstances.

[0074] Figure 1 and Figure 2 The figure shows a preferred embodiment of the white radish harvester row recognition system based on machine vision, which includes a control mechanism, an image acquisition mechanism, a pulling and harvesting device 1, a vehicle-mounted chassis 2, a row adjustment device 3 and a mounting frame 4.

[0075] In a specific embodiment of the present invention, the mounting frame 4 is located at the right rear end of the vehicle chassis 2, the row adjustment device 3 is installed on the chassis 2 and the mounting frame 4, the plucking and harvesting device 1 is installed on the vehicle chassis 2, the plucking and harvesting device 1 is used to plucking and harvesting white radish, and the adjustment device 3 is located on both sides of the plucking and harvesting device 1, and is used to adjust the horizontal translation of the plucking and harvesting device 1 to the left and right.

[0076] The control mechanism is connected to the image acquisition mechanism and the row adjustment device 3 respectively. The image acquisition mechanism is used to collect images of the white radish planting field in front of the harvester to obtain depth images and color images, and transmit them to the control unit; the control mechanism performs image processing, extracts ridge boundary lines and white radish ridge lines, uses the extracted ridge boundary lines and ridge lines for verification, obtains navigation tracking point navigation lines, and controls the row adjustment device 3 according to the navigation lines to adjust the horizontal translation of the extraction harvesting device 1 left and right to perform row alignment.

[0077] The harvesting device 1 includes a clamping conveyor belt 10, a tassel-supporting mechanism 11, a clamping conveying mechanism 12, a tassel-cutting mechanism 19, a transmission mechanism 21 and a motor 22; the tassel-supporting mechanism 11 is located at the front end of the clamping conveying mechanism 12, and the transmission mechanism 21 is located at the rear end of the clamping conveying mechanism 12; the tassel-cutting mechanism 19 is located at the lower end of the clamping conveying mechanism 12 close to the transmission mechanism 21, and the motor 22 is located on one side of the transmission mechanism 21, and is used to drive the transmission mechanism 21 to drive the clamping conveyor belt 10 to rotate.

[0078] The row adjustment device 3 includes a first heavy-duty screw guide 15, a first displacement sensor 17, a stepper motor, a second heavy-duty screw guide 25 and a second displacement sensor 23; the stepper motor includes a first stepper motor 16 and a second stepper motor 24; the first heavy-duty screw guide 15 and the first displacement sensor 17 are installed in parallel at the front end of one side of the vehicle chassis 2, the first guide rail slide 14 is installed on the first heavy-duty screw guide 15, the heavy-duty screw guide 15 is connected to the first stepper motor 16, the upper end surface of the first guide rail slide 14 is connected to the base end of the hydraulic cylinder 13, and the side end surface of the first guide rail slide 14 is connected to the measuring end of the first displacement sensor 17; the second heavy-duty screw guide 25 and the second displacement sensor 23 are installed in parallel on the upper end surface of the mounting frame 4 on the vehicle chassis 2, the second guide rail slide 26 is installed on the second heavy-duty screw guide 25, the second heavy-duty screw guide 25 It is connected to the second stepper motor 24, and the upper end of the second guide rail slide 26 is hinged to the frame 20 of the clamping and conveying mechanism 12 of the plucking and harvesting device 1, and the side end face of the second guide rail slide 26 is connected to the measuring end of the second displacement sensor 23; the first stepper motor 16, the first displacement sensor 17, the second displacement sensor 23 and the second stepper motor 24 are respectively connected to the control mechanism, and the control mechanism controls the first stepper motor 16 to drive the first heavy-loaded screw guide 15 to slide on the first guide rail slide 14, and controls the second stepper motor 24 to drive the second heavy-loaded screw guide 25 to slide on the second guide rail slide 26. The first displacement sensor 17 is used to measure the position of the first guide rail slide 14 and feed it back to the control mechanism, and the second displacement sensor 23 is used to measure the position of the second guide rail slide 26 and feed it back to the control mechanism, so as to drive the overall horizontal translation of the plucking and harvesting device 1 to realize the horizontal row adjustment of the plucking and harvesting device 1.

[0079] The telescopic end of the hydraulic cylinder 13 is connected to the lower end of the frame 20 of the clamping and conveying mechanism 12 for adjusting the harvesting height of the harvesting device 1 .

[0080] The second heavy-loaded screw guide rail 25 and the second displacement sensor 23 are installed in parallel on the upper end surface of the mounting frame 4. The upper end of the second guide rail slide 26 is hinged to the frame 20 of the clamping and conveying mechanism 12, so that the harvesting device 1 can rotate around the hinge axis. The side end surface of the second guide rail slide 26 is connected to the measuring end of the second displacement sensor 23.

[0081] When the whole machine is harvesting in rows, the transmission mechanism 21 is used to drive the clamping conveyor belt 10 to rotate, and the radish leaves are lifted by the supporting mechanism 11 and sent to the clamping conveying mechanism 12. Then the white radish is picked and transported backward to the radish leaves cutting mechanism 19 for cutting. After cutting, the white radish falls into the collecting bucket 18, and the radish leaves continue to move backward and are removed.

[0082] When row adjustment is required, the first stepper motor 16 and the second stepper motor 24 simultaneously drive the first heavy-duty screw guide 15 and the second heavy-duty screw guide 25 to move synchronously, thereby driving the overall horizontal translation of the harvesting device 1, so that the first heavy-duty screw guide 15 and the second heavy-duty screw guide 25 move in the same direction at the same time, thereby realizing horizontal row adjustment of the harvesting device 1.

[0083] The present invention uses a binocular vision camera as an image acquisition mechanism to capture the working environment in front of a pull-type combine harvester, obtaining a depth image and a color image within the same field of view. The depth image and color image target interest domains are selected based on actual conditions, the depth image is processed to extract the outer boundary lines of the white radish ridges, a line clustering algorithm is used on the color image to extract the white radish planting ridge lines, and the white radish planting row lines extracted from the color image are verified based on the outer boundary lines of the ridges extracted from the depth image. Pixel tracking points are selected for the verified white radish planting row lines as navigation lines, which are converted into real tracking points through coordinate system mapping. The current position of the harvesting device is read and the relative distance deviation from the real starting tracking point is calculated. A fuzzy control algorithm is used to adjust the harvesting device to a specified position to achieve row-by-row harvesting. The algorithm is then refreshed and the above process is repeated, thereby achieving integrated automatic row alignment and adjustment.

[0084] Preferably, the image acquisition mechanism is a camera 7 .

[0085] Camera 7 installation position and coordinate system mapping conversion:

[0086] like Figure 1As shown; in actual use, the camera 7's field of view forms a certain angle with the planting ground, so it is necessary to convert the pixel coordinate system to the real coordinate system through coordinate system mapping conversion. The camera 7 is fixedly mounted on the gantry 6 at the front end above the plucking and harvesting device 1 and is in the same vertical plane as the plucking and harvesting device 1 in the reset state. The camera 7 is tilted downward to ensure that the front field of view is clear of obstacles.

[0087] In the picture captured by the camera 7 field of view, the midpoint of the bottom of the picture is taken as the origin O of the coordinate system, the horizontal direction of the bottom is the U axis, and the vertical direction is the V axis, and the picture pixel coordinate system UOV is established.

[0088] The world coordinate system is established on the horizontal plane at the average height h of the white radish above the ground, with the machine's forward direction as Y w Axis, perpendicular to the Y plane w The axis direction is X w Axis, vertical projection of camera 7 installation position along the Y w The distance c in the axis direction is set as the origin of the world coordinate system O w , establish the world coordinate system X w O w Y w .

[0089] Method for determining the translation distance c; when the machine is in normal working condition, the distance between the midpoint of the width of the end of the support mechanism 11 and the vertical projection point of the camera 7 is c;

[0090] According to Zhang Zhengyou's calibration method, the pixel coordinate system and the world coordinate system satisfy the mapping relationship: A is the intrinsic parameter matrix of the camera, and R*T is the extrinsic parameter matrix.

[0091]

[0092] The method of extracting the boundary line of white radish planting ridge based on depth image is as follows: Figure 3 As shown;

[0093] Preferably, the camera 7 uses a binocular camera to collect images of the white radish planting field in front of the harvester. Figure 3 As shown, the system employs binocular ranging principles, using left and right infrared cameras to obtain the parallax of pixel coordinates at the same location in the photo. Binocular triangulation is then used to determine the depth distance of each pixel and construct a depth image. Extracting row boundaries from depth images fully utilizes plant features with varying heights, is unaffected by changes in lighting and color, and is more stable and reliable. However, this method is susceptible to interference from physical terrain obstacles.

[0094] The distance calculation relationship is as follows; Z is the distance, d is the parallax of the photo at the same point, D is the binocular center distance, and f is the camera focal length.

[0095]

[0096] Since the outer ridge boundary of the white radish planting area grows higher than the ground, there will be a significant height difference at the outer ridge boundary position, which makes the outer ridge boundary line form an obvious gradient dividing line on the depth map, so it can be used as the basis for feature line extraction.

[0097] The control mechanism extracts the ridge boundary line based on the outermost ridge boundary line extraction algorithm of the depth image, specifically comprising the following steps:

[0098] Step 1: The control mechanism intercepts the target area and processes the depth image to eliminate the interference of unnecessary radish ridges in the field of view. The interception principle is as follows;

[0099]

[0100] Among them, B (i,j) is the coordinate of the target area point in the intercepted depth image, I (i,j) is the coordinate of the corresponding pixel in the input original depth image, x, y is the coordinate of the upper left corner of the target area, W, H are the width and height of the target area respectively, the effect is as follows Figure 3 As shown;

[0101] Step 2: Considering that the camera's field of view is not perpendicular to the planting ground, the intercepted depth map is depth preprocessed to make the camera coordinate system plane parallel to the planting ground. Since the depth map is a single-channel grayscale image, the grayscale of a single pixel represents the distance from the real point to the camera coordinate system plane. The farther the distance, the lower the grayscale. For points that cannot be measured, the default is infinity and the grayscale is set to 0. Since the ridge height of the radish planting plot is significantly higher than the ground, a fixed threshold binarization method is used to segment the intercepted depth image to obtain the outermost ridge boundary line, resulting in a binary image:

[0102]

[0103] Among them, 0 represents black (background) and 255 represents white (foreground), as shown in Figure 3 As shown, where T is a selected fixed threshold;

[0104] Step 3: Perform area filtering and morphological processing on the processed binary image to eliminate noise and edge hole interference, then read the coordinates of the first non-zero point from right to left in each row in the intercepted area and store them in the matrix P, where (i k ,j k ) are the coordinates of points that are not 0:

[0105]

[0106] Step 4: Use the least squares method to fit the extracted points to obtain the ridge boundary line, as shown in Figure 3 As shown; Figure 3 The position of the extracted ridge boundary line in the image is:

[0107] y=ax+b

[0108] Where a and b are the slope and intercept of the ridge boundary line extracted from the depth map, respectively.

[0109] White radish planting row line extraction method based on line clustering in color images

[0110] like Figure 4 As shown; the selected binocular vision camera also has a color camera, through which a color picture with rich color information can be obtained. Figure 4 ①, extract the radish ridge lines based on the obtained image using color features.

[0111] The control mechanism is based on the method for extracting white radish planting lines in a color picture based on straight line clustering, which specifically includes the following steps:

[0112] Step 1: Use the super green feature grayscale method to enhance the green area features.

[0113] The super green feature grayscale method is an image processing method for vegetation areas. It extracts the super green area in the image and converts it into a grayscale image, thereby highlighting the characteristics of the vegetation. The super green feature grayscale method selected by this invention is 2G-RB (Two Green-Red-Blue). First, the target interest area of ​​the color image is intercepted to ensure that there are three clear ridges in the field of view, and the rest of the area is set to 0. Figure 4 ② As shown, the super green feature method is then used to grayscale, as shown in Figure 4 As shown in ③, the calculation method is:

[0114]

[0115] Among them, 2 means expanding the green channel pixels by 2 times, G means green channel, R means red channel, and B means blue channel;

[0116] Step 2: Use Otsu method to perform image binarization processing, such as Figure 4 As shown in ④, the Otsu method is an algorithm that automatically determines the image threshold and is used for binarization processing in image segmentation. By analyzing the grayscale histogram of the image, a threshold is found to maximize the inter-class variance of the image into two classes (foreground and background). The mask effect of the color image after binarization is shown in the figure below. Figure 4⑩As shown;

[0117] Step 3: The binarized image is first dilated and then eroded to reduce the impact of noise. The dilation operation can expand the area of ​​the brighter areas in the image, and may also connect gaps or fill small holes. Preferably, the convolution kernel size used for dilation is 5×5. The subsequent erosion operation can reduce the area of ​​the brighter areas in the image, and may also disconnect or weaken small protrusions to restore the original image outline size. Preferably, the convolution kernel size is still 5×5;

[0118] Step 4: Perform area filtering on the binary image to obtain the area of ​​each connected domain, perform area filtering on the connected domain, and remove interference points in small areas. Figure 4 ⑤As shown;

[0119] Step 5: Line clustering is performed to extract planting rows. In the processed binary image, due to the presence of furrows between the radish planting ridges and the overlap of leaves between the furrows, traditional clustering methods cannot effectively separate the clusters between different furrows, making subsequent navigation line fitting and extraction impossible. However, the radish rows on the same ridge are clustered on the same straight line, their planting lines are straight, and the number of radish rows within the field of view of the photo is known. Therefore, a new clustering method is proposed using this feature:

[0120] First, select three ridge lines in the intercepted area and distribute them evenly. Figure 4 As shown in ⑥, the number of clusters is 3, which is consistent with the number of ridges within the target interest threshold. Then, the distance between each white point in the binary image and the three straight lines (ridge lines) is calculated and compared. The ones with the closest distance are classified into the same cluster as the straight lines, and the cluster lines are obtained by fitting:

[0121]

[0122] The first clustering effect Figure 4 ⑦As shown;

[0123] Then, after the white coordinate points are classified, the least squares fitting straight line of each type of coordinate point is calculated, and a straight line roughly consistent with the direction of the navigation line can be extracted. Then the straight line of the first fitting is used as a new clustering straight line for clustering again. Repeating the above steps can further refine the clustering of the last rough classification and extract a more accurate clustering straight line. The second clustering effect is better. Figure 4 ⑧As shown.

[0124] Finally, in the third clustering, distance judgment is added. According to practical experience, when the distance is greater than d, it can be considered that the point is far from the row center line and is an unreliable point, so it is removed. Finally, the straight line extracted by the fourth clustering is the planting row line. Figure 4 ⑨As shown.

[0125]

[0126] Navigation line tracking point selection:

[0127] like Figure 5 The figure below shows the principle of selecting navigation line tracking points. During actual operation, the machine moves at a constant speed v, and the algorithm refreshes at a frequency of f. Therefore, within one algorithm refresh, the actual distance the machine moves is z = v / f. The pixel width used in the image is h. Since the machine's forward speed is 0.2-0.3 m / s and the algorithm refreshes at a frequency of 2-5 Hz, the distance z is 0.06-0.15 m. However, within a field of view of 4-5 m, its pixel width can be considered a single point in the image. Therefore, the tracking point is the starting point of the extracted line for each image refresh.

[0128] Since color images are easily affected by outdoor lighting, while depth images are not easily affected by lighting due to the use of binocular infrared light measurement, but can only extract the outer ridge boundary line; therefore, the depth map is used to extract the navigation line to assist in the verification of the color map extraction navigation line, so that the two navigation lines are fused to obtain the final target navigation line.

[0129] In order to facilitate the fusion of two navigation lines, it is necessary to reference relevant evaluation indicators to judge the reliability of the navigation line extraction effect and to allocate weight ratios in subsequent fusion.

[0130] For the ridge navigation line extracted from the color map, the outermost ridge line is selected and the goodness of fit is used as the evaluation index. The goodness of fit refers to the degree of fit of the regression line to the observed value. The statistic that measures the goodness of fit is the coefficient of determination R. 2 . R 2 The maximum value is 1, R 2 The closer the value of R is to 1, the better the regression line fits the observed value; on the contrary, 2 The smaller the value of , the worse the regression line fits the observed value. i is the i-th observation value, The mean of all observations, The i-th predicted value;

[0131]

[0132] For the ridge-row boundary navigation line extracted from the depth map, the ridge boundary line is obtained by using the height difference between the planted ridges and the ground for boundary segmentation, and then the obtained boundary line is fitted to obtain the ridge boundary navigation line. Therefore, the accuracy of the ridge boundary line segmentation becomes an important index for evaluating the reliability of the ridge boundary navigation line. For the boundary line, calculate the average height difference between the upper pixel points of the boundary line in the depth map and the ground, and compare this average height difference with the standard height difference to obtain the comparison coefficient g as the evaluation index of the reliability degree. The larger the comparison coefficient, the greater the height difference between the ridge boundary line and the ground. If it is greater than the standard height difference, it is considered reliable, and the reliability coefficient is set to 1. The formula is as follows;

[0133]

[0134] The navigation tracking point navigation line is obtained through the tracking point fusion extraction algorithm, which specifically includes the following steps:

[0135] First, convert the ridge boundary navigation line extracted from the depth image according to the coordinate system, translate it by a distance of half the standard ridge width L in the real coordinate system, and record the translated straight line in the pixel coordinate system.

[0136] Subsequently, for the translated ridge boundary navigation line and the ridge-row line extracted from the color image, according to the coefficient of determination R 2 and the comparison coefficient g, perform fusion according to the normalization method, and assign the weight coefficients r1 and r2. r1 is the weight coefficient assigned to the ridge-row line in the depth map, and r2 is the weight coefficient assigned to the ridge-row line in the color map:

[0137]

[0138] According to the fusion of the weight coefficients, obtain the final target navigation line. Assume that k1 and b1 are the slope and intercept of the ridge-row line in the depth map, and assume that k2 and b2 are the slope and intercept of the ridge-row line extracted from the color map. The fusion formula using the weight coefficients is:

[0139] y = (k1r1 + k2r2)x + (r1b1 + r2b2)

[0140] Finally, the intersection point of the target navigation line at the pixel h / 2 at the bottom end of the color image is the target navigation line tracking point, which is converted into a point in the real coordinate system through the coordinate system. This point is the horizontal target position adjustment point required for the row alignment adjustment device. Specifically, obtain the X-axis coordinates of the intersection points of all navigation lines, which are X1, X2, X3, X4, and X5 from right to left (X5 < X4 < X3 < X2 < X1) in sequence. Obtain the intersection point Z of the target navigation line at h / 2 as the target navigation line tracking point, which is converted into a point in the real coordinate system through the coordinate system. This point is the horizontal position adjustment point required for the row alignment adjustment device 3.

[0141] Fuzzy control row alignment adjustment method

[0142] like Figure 6 As shown in the figure, it is a fuzzy control rule table for controlling row adjustment; according to the pixel position of the tracking point obtained, it is converted into a coordinate point in the real coordinate system through the coordinate system, and the position of the harvesting device is measured by the displacement sensor, and then the difference is calculated according to the distance, which is the relative distance that needs to be moved to achieve row tracking.

[0143] The relative distance is obtained, and the fuzzy control algorithm is used to control the movement of the heavy-duty screw guide rail in the row adjustment mechanism to drive the pulling and harvesting device 1 to adjust the row.

[0144] The input is defined as the position error, that is, the distance difference between the current position and the target position; the output is the pulse frequency, that is, the pulse frequency of the lower computer PLC controlling the stepper motor.

[0145] The position error is fuzzified and divided into fuzzy sets, such as "far, farther, moderate, closer, near". A set of fuzzy rules is established to map the fuzzified input to the fuzzified output.

[0146] Speed ​​fuzzy control: the stepper motor controls the speed of the motor by adjusting the pulse frequency;

[0147] If the position error is far, then the pulse frequency of the stepper motor is the largest;

[0148] If the position error is large, then the pulse frequency of the stepper motor is medium;

[0149] If the position error is moderate, then keep the stepper motor pulse frequency in the middle;

[0150] If the position error is close, then the pulse frequency of the stepper motor is slightly lower;

[0151] If the position error is small, the pulse frequency of the stepper motor is minimized.

[0152] According to the input fuzzy variables and fuzzy rules, fuzzy reasoning is performed to obtain fuzzy output, which is then converted into a specific pulse frequency. The pulse frequency is adjusted by PLC to achieve control of the line adjustment speed.

[0153] System hardware structure design:

[0154] Preferably, the control mechanism includes a host computer and a slave computer; the host computer is connected to the image acquisition mechanism, and is used to receive images acquired by the image acquisition mechanism and perform image processing, extract ridge boundary lines and white radish ridge lines, and use the extracted ridge boundary lines and ridge lines for verification to obtain navigation tracking point navigation lines; the slave computer controls the row adjustment device 3 to adjust the harvesting device 1 for row adjustment according to the navigation line of the host computer.

[0155] like Figure 7 As shown in the figure, the system hardware selects distributed processing combining the upper computer and the lower computer. The two transmit data through the 485 communication protocol, which can effectively ensure the real-time performance and efficiency of the system.

[0156] The host computer uses an embedded control terminal with Linux system as the processor for running the image processing algorithm, and the lower computer uses a high-reliability PLC for terminal drive control;

[0157] The system's human-computer interaction display terminal is connected to the host computer through a USB port for human-computer interaction and terminal operation display;

[0158] The embedded control terminal collects the image signal output by the camera 7 through the USB port and completes the calculation work such as image processing, interface interaction, distance calculation, and lower computer communication.

[0159] The lower computer PLC transmits the data read from the two displacement sensors to the upper computer to calculate the position distance, and controls the two heavy-loaded screw guides to run to the specified position according to the moving distance calculated by the upper computer, thereby driving the harvesting device 1 to adjust the rows.

[0160] A control method for a row recognition system of a white radish harvester based on machine vision comprises the following steps:

[0161] The image acquisition mechanism collects images of the white radish planting field in front of the harvester to obtain depth images and color images, and transmits them to the control unit; the control mechanism performs image processing, extracts ridge boundary lines and white radish ridge lines, uses the extracted ridge boundary lines and ridge lines for verification, obtains target navigation lines, and controls the row adjustment device 3 to adjust the horizontal translation of the extraction and harvesting device 1 left and right according to the target navigation lines to harvest the rows.

[0162] like Figure 8 As shown, the row harvesting specifically includes the following steps:

[0163] S1. Start the equipment and reset the harvesting device 1 to the set initial starting point through the row adjustment device 3. Check the working status of the equipment to confirm whether the safety function can work;

[0164] S2, adjust the pulling height of the end of the pulling and harvesting device 1 by the hydraulic cylinder 13, and prepare for harvesting;

[0165] S3, adjusting the field of view of the image acquisition mechanism to capture images of the white radish planting field in front of the harvester to obtain a depth image and a color image;

[0166] S4, performing pre-processing such as interception, filtering, expansion and corrosion on the depth image and color image obtained by shooting;

[0167] S5, extracting ridge boundary lines and ridge planting row lines from the pre-processed depth image and color image respectively;

[0168] S6. Fusing the extracted ridge boundary line and ridge row line to obtain a target navigation line, thereby obtaining a navigation tracking point;

[0169] S7, converting the navigation tracking point in the image coordinate system into a point in the real coordinate system, and calculating the position error of the harvesting device 1 relative to the point;

[0170] S8, the lower computer uses the position error as input and adopts the fuzzy control algorithm to control the line adjustment device 3 to move and complete the line adjustment. After the adjustment is completed, the above process is repeated to refresh and realize real-time line adjustment;

[0171] S9, after the target row is harvested, proceed to the next row for harvesting;

[0172] S10: All harvesting is completed and the machine is shut down.

[0173] The control mechanism adopts fuzzy control algorithm to control the row adjustment device 3 to adjust the row, including the following steps:

[0174] The position error between the current position and the target position is set as the input, and the pulse frequency of the stepper motor is set as the output. The position error is fuzzified and divided into fuzzy sets: "far, relatively far, moderate, relatively close, and near". A set of fuzzy rule tables is established to map the fuzzified input to the fuzzified output. Fuzzy reasoning is performed based on the input fuzzy variables and fuzzy rules to obtain fuzzy output, which is then converted into a specific pulse frequency. By adjusting the pulse frequency, the line adjustment speed can be controlled.

[0175] The control mechanism of the present invention includes an algorithm module for obtaining the position of white radish rows. The algorithm module includes an algorithm for extracting the outermost ridge boundary line based on a depth image, an algorithm for extracting the white radish row line based on line clustering, a tracking point fusion extraction algorithm, and a fuzzy control algorithm for the row adjustment device. When the driver harvests along the outer edge of the ridge, the image acquisition mechanism captures depth images and color images in real time. The algorithm processes the ridge boundary line and the ridge planting row line, and the tracking point verification and extraction algorithm obtains tracking points. The fuzzy control algorithm then controls the row adjustment device to adjust the pulling and harvesting device to the tracking point position for row-by-row harvesting. This enables the pulling and harvesting device to harvest rows simultaneously. The present invention solves the problems of frequent and labor-intensive manual row alignment in traditional white radish harvesting processes, realizes automatic row alignment adjustment of the harvesting components, and improves the intelligent level of white radish combined harvesting.

[0176] A harvester includes the white radish harvester row recognition system based on machine vision, and thus has the above-mentioned beneficial effects, which will not be described in detail here.

[0177] It should be understood that although this specification is described according to various embodiments, not every embodiment contains only one independent technical solution. This narrative method of the specification is only for the sake of clarity. Those skilled in the art should regard the specification as a whole. The technical solutions in each embodiment can also be appropriately combined to form other implementation methods that can be understood by those skilled in the art.

[0178] The series of detailed descriptions listed above are only specific descriptions of feasible embodiments of the present invention. They are not intended to limit the scope of protection of the present invention. Any equivalent embodiments or changes that do not deviate from the technical spirit of the present invention should be included in the scope of protection of the present invention.

Claims

1. A white radish harvester row recognition system based on machine vision, characterized in that: It includes a control mechanism, an image acquisition mechanism, a plucking and harvesting device (1), a vehicle chassis (2) and a row adjustment device (3); The plucking and harvesting device (1) and the row adjustment device (3) are installed on the vehicle chassis (2); the plucking and harvesting device (1) is used for plucking and harvesting white radish; the row adjustment device (3) is located on both sides of the plucking and harvesting device (1) and is used for adjusting the plucking and harvesting device (1) to move horizontally left and right; The control mechanism is connected to the image acquisition mechanism and the row adjustment device (3) respectively. The image acquisition mechanism is used to acquire an image of the white radish planting field in front of the harvester to obtain a depth image and a color image, and transmit the image to the control unit; the control mechanism performs image processing, extracts ridge boundary lines and white radish ridge row lines, uses the extracted ridge boundary lines and ridge row lines for verification, obtains a target navigation line, and controls the row adjustment device (3) according to the target navigation line to adjust the pulling and harvesting device (1) to move horizontally left and right for row alignment; The control mechanism includes an upper computer and a lower computer; The host computer is connected to the image acquisition mechanism, and is used to receive images acquired by the image acquisition mechanism and perform image processing, extract ridge boundary lines and white radish ridge lines, and use the extracted ridge boundary lines and ridge lines for verification to obtain navigation tracking point navigation lines; The navigation tracking point and navigation line are obtained by the tracking point fusion extraction algorithm: First, the ridge boundary navigation line extracted from the depth image is converted according to the coordinate system, translated by half the standard ridge width distance L in the real coordinate system, and the translated straight line is recorded in the pixel coordinate system; Then, the translated ridge boundary navigation line and the ridge line extracted from the color image are fused according to the determination coefficient R² and the comparison coefficient g, and the weight coefficient is assigned. r 1 and r 2, r1 is the weight coefficient assigned to the depth map ridge line, r2 is the weight coefficient assigned to the color map ridge line: ； According to the weight coefficient fusion, the final target navigation line is obtained: ； in, k 1 and b 1 is the slope and intercept of the depth map ridge line, k 2 and b 2 is the slope and intercept of the ridge line extracted from the color map.

2. The machine vision-based white radish harvester row recognition system according to claim 1, characterized in that: The row adjustment device (3) comprises a first heavy-duty screw guide rail (15), a first displacement sensor (17), a stepping motor, a second heavy-duty screw guide rail (25) and a second displacement sensor (23); The stepper motor includes a first stepper motor (16) and a second stepper motor (24); The first heavy-duty screw guide rail (15) and the first displacement sensor (17) are installed in parallel at the front end of one side of the vehicle chassis (2), the first guide rail slide (14) is installed on the first heavy-duty screw guide rail (15), the first heavy-duty screw guide rail (15) is connected to the first stepper motor (16), and the side end surface of the first guide rail slide (14) is connected to the measuring end of the first displacement sensor (17); The second heavy-duty screw guide rail (25) and the second displacement sensor (23) are installed in parallel on the upper end surface of the mounting frame (4) on the vehicle chassis (2), the second guide rail slide (26) is installed on the second heavy-duty screw guide rail (25), the second heavy-duty screw guide rail (25) is connected to the second stepping motor (24), the upper end of the second guide rail slide (26) is hinged to the frame (20) of the clamping and conveying mechanism (12) of the picking and harvesting device (1), and the side end surface of the second guide rail slide (26) is connected to the measuring end of the second displacement sensor (23); The first stepper motor (16), the first displacement sensor (17), the second displacement sensor (23) and the second stepper motor (24) are respectively connected to the control mechanism. The control mechanism controls the first stepper motor (16) to drive the first heavy-loaded screw guide rail (15) to slide on the first guide rail slide (14), and controls the second stepper motor (24) to drive the second heavy-loaded screw guide rail (25) to slide on the second guide rail slide (26). The first displacement sensor (17) is used to measure the position of the first guide rail slide (14) and feed it back to the control mechanism. The second displacement sensor (23) is used to measure the position of the second guide rail slide (26) and feed it back to the control mechanism, thereby driving the overall horizontal translation of the plucking and harvesting device (1) to achieve horizontal row adjustment of the plucking and harvesting device (1).

3. The machine vision-based white radish harvester row recognition system according to claim 1, characterized in that: The lower computer controls the row adjustment device (3) to adjust the plucking and harvesting device (1) to perform row adjustment according to the navigation line of the upper computer.

4. The machine vision-based white radish harvester row recognition system according to claim 1, characterized in that: The control mechanism extracts the ridge boundary line based on the outermost ridge boundary line extraction algorithm of the depth image: The control mechanism intercepts the target area and performs interception processing on the depth image: ； in, B (i,j) is the coordinate of the target area point in the intercepted depth image, I (i,j) is the coordinate of the corresponding pixel in the input original depth image, x, y is the coordinate of the upper left corner of the target area, W 、 H are the width and height of the target area respectively; the fixed threshold binarization method is used to segment the intercepted depth image to obtain the outermost ridge boundary line and obtain a binary image: ； Where 0 represents a black background and 255 represents a white foreground; T is the selected fixed threshold; Perform area filtering and morphological processing on the processed binary image to eliminate noise and edge hole interference, then read the coordinates of the first non-zero point from right to left in each row in the intercepted area and store it in the matrix P middle: ； in( i k , j k ) are the coordinates of points that are not 0; The least squares method is used to fit the extracted points to obtain the ridge boundary line: y=ax+b in, a , b are the slope and intercept of the ridge boundary line extracted from the depth map, respectively.

5. The machine vision-based white radish harvester row recognition system according to claim 1, characterized in that: The control mechanism is based on the method for extracting the planting lines of white radish in the color picture of straight line clustering: The super green feature grayscale method is used to enhance the green area features. First, the target interest area of ​​the color image is intercepted, and the remaining areas are set to 0. The super green feature method is used for grayscale conversion. The calculation method is: ； Among them, 2 means expanding the green channel pixels by 2 times, G means green channel, R means red channel, and B means blue channel; Use Otsu method to perform image binarization processing; The binary image is first dilated and then eroded to obtain the area of ​​each connected domain, and then the connected domain area filtering is performed; Perform straight line clustering to extract planting row lines: select ridge lines in the intercepted area, distribute them evenly, calculate the distance between each white point in the binary image and the straight line (ridge line), compare them, classify the ones with the closest distance into the same category as the straight line, and obtain clustered straight lines by fitting: 。 6. A harvester, characterized in that: The invention comprises the row recognition system for a white radish harvester based on machine vision as described in any one of claims 1 to 5.

7. A control method for a row recognition system of a white radish harvester based on machine vision according to any one of claims 1 to 5, characterized in that: The following steps are involved: The image acquisition mechanism acquires an image of the white radish planting field in front of the harvester to obtain a depth image and a color image, and transmits the image to the control unit; the control mechanism processes the image, extracts ridge boundary lines and white radish ridge lines, uses the extracted ridge boundary lines and ridge lines for verification, obtains a target navigation line, and controls the row adjustment device (3) to adjust the horizontal translation of the extraction and harvesting device (1) to the left and right according to the target navigation line to harvest the rows.

8. The control method of the row recognition system of the white radish harvester based on machine vision according to claim 7, characterized in that: The specific steps for row harvesting include: S1. The equipment is turned on and the harvesting device (1) is reset to the set initial starting point through the row adjustment device (3). The working status of the equipment is checked to confirm whether the safety function can work; S2, adjusting the pulling height of the end of the pulling and harvesting device (1) by means of the hydraulic cylinder (13), and preparing for harvesting; S3, adjusting the field of view of the image acquisition mechanism to capture images of the white radish planting field in front of the harvester to obtain a depth image and a color image; S4, performing interception, filtering, expansion and corrosion processing on the depth image and color image obtained by shooting; S5, extracting ridge boundary lines and ridge planting row lines from the pre-processed depth image and color image respectively; S6. Fusing the extracted ridge boundary line and ridge row line to obtain a target navigation line, thereby obtaining a navigation tracking point; S7, converting the navigation tracking point in the image coordinate system into a point in the real coordinate system, and calculating the position error of the harvesting device (1) relative to the point; S8, the lower computer uses the position error as input and adopts the fuzzy control algorithm to control the line adjustment device (3) to move and complete the line adjustment. After the adjustment is completed, the above process is repeated and refreshed to achieve real-time line adjustment; S9. After the target row is harvested, proceed to the next row for harvesting.

9. The control method of the row recognition system of the white radish harvester based on machine vision according to claim 7, characterized in that: The control mechanism adopts a fuzzy control algorithm to control the row adjustment device (3) to adjust the row, including the following steps: The position error between the current position and the target position is set as the input, and the pulse frequency of the stepper motor is set as the output. The position error is fuzzified and divided into fuzzy sets. A set of fuzzy rule tables is established to map the fuzzified input to the fuzzified output. Fuzzy reasoning is performed based on the input fuzzy variables and fuzzy rules to obtain fuzzy output, which is then converted into a specific pulse frequency. By adjusting the pulse frequency, the line adjustment speed can be controlled.

Images