A method for extracting the navigation line of a seedling strip based on horizontal and vertical dual-slider scanning correction.

Abstract

This invention discloses a method for extracting curved seedling strip navigation lines based on horizontal and vertical dual-slider scanning correction, belonging to the field of agricultural engineering. It acquires near-end images of seedling strips in the field using a camera, and employs an improved image processing method to segment the seedling strips and land, reducing noise disturbance from weeds. The horizontal slider is used for vertical scanning to capture and collect the coordinate feature points of the left and right seedling strips, and the point set is corrected and optimized through horizontal scanning with the vertical slider. The coordinates of the left and right seedling strips are calculated by averaging the coordinates of the left and right seedling strips to obtain the navigation line coordinates. Polynomial fitting is performed on the navigation line coordinates to obtain the curved navigation line, obtaining the optimal solution. This provides a new method for realizing dynamic curved navigation in complex fields. This invention utilizes nonlinear navigation line acquisition and jitter reduction processing to improve the accuracy of near-end navigation in the field, while also improving processing speed. The processing time for one image is within 20ms, which can meet the needs of real-time field navigation for machinery.

Description

Technical Field:

[0001] This invention relates to the field of agricultural engineering, specifically to a method for extracting alignment navigation lines in intelligent agriculture. Background technology:

[0002] 1. With the development and application of intelligent agricultural equipment, the autonomous operation capability of unmanned equipment in the field is particularly important, and for row processing robots, the ability to align and navigate seedling strips with high efficiency and precision is of great significance.

[0003] 2. The core technology for aligning seedling strips in the field is improving the performance of navigation line extraction. More efficient and accurate navigation line acquisition will greatly improve the operation of intelligent inter-row equipment. Due to the complexity of the field environment, seedling strips are often irregular, twisted, uneven, and subject to noise interference, making visual navigation difficult. Currently, many studies and discussions have been conducted both domestically and internationally on this issue, resulting in numerous solutions. Swedish experts Bjprn Astrand et al. proposed a robust crop row recognition navigation algorithm based on Hough transform, but it only has straight line extraction capabilities; this was the earliest attempt. Now, navigation methods based on advanced technologies such as deep learning have emerged, but there is still room for improvement in processing speed and efficiency. Combining the research status of various teams at home and abroad, this invention proposes a curve seedling strip navigation line extraction method based on horizontal and vertical dual-slider scanning correction, providing efficient and reliable seedling strip alignment navigation for intelligent agricultural machinery. Summary of the Invention:

[0004] 1. This invention addresses the issues of slow navigation speed and low accuracy in row-based intelligent agricultural robots by proposing a curved seedling strip navigation line extraction method based on horizontal and vertical dual-slider scanning correction. The method involves acquiring and compressing field images, extracting the seedling strip range using the Region of Interest (ROI), extracting the green plant seedling strip region using an image color segmentation algorithm, optimizing feature point acquisition through horizontal and vertical scanning correction, and obtaining the curved navigation line by polynomial fitting of the navigation line coordinates. Simultaneously, dynamic optimization of the ROI is achieved to improve processing accuracy and obtain a better solution.

[0005] 2. To solve the above technical problems, the present invention adopts the following technical solution: a method for extracting curved seedling strip navigation lines based on horizontal and vertical dual-slider scanning correction, the steps of which are as follows: S1: Use a color camera to acquire field images and compress them; S2: Use ROI to extract the seedling strip range; S3: Extract the seedling strip area of ​​green plants through an image color segmentation extraction algorithm and perform morphological processing such as image binarization and opening / closing operations; S4: Use a horizontal slider to perform vertical scanning of the image, obtain the coordinates of the connected component boundary within the slider and perform mean calculation, divide the left and right seedling strip feature point coordinates and collect them; S5: Simultaneously, the image in step C is processed... S6: Perform perspective transformation to obtain a top-view binary image of the seedling strip; S7: Use the vertical slider to scan the image horizontally to obtain the pixel distribution curve and perform curve fitting and smoothing to obtain the maximum value coordinates to get the endpoint coordinates (x1,0)(x2,0)(x1,w)(x2,w) of the two seedling strip lines; S8: Project the two seedling strip lines y1, y2 back to the original image, use the lines y1, y2 to correct and optimize the point set, and calculate the coordinates of the navigation line by averaging the coordinates of the left and right seedling strips; S9: Perform polynomial fitting on the navigation line coordinates to obtain the curved navigation line; S10: Organize the coordinate information obtained in S4 and return it to correct the ROI region in S2; S11: Repeat steps S1 to S9.

[0006] 3. By using a horizontal slider to perform vertical scanning of the image and using a vertical slider to perform horizontal scanning to obtain the maximum coordinates, the endpoint coordinates of the two seedling strip lines are obtained. The two seedling strip lines y1 and y2 are projected back to the original image. Then, the point set is corrected and optimized using the straight lines y1 and y2. The coordinates of the navigation line are obtained by averaging the coordinates of the left and right seedling strips. The optimal solution of the feature is obtained, which improves the speed of navigation line extraction and fitting accuracy.

[0007] 4. Compared with the prior art, the present invention has the following advantages: it is lighter in size, faster in processing speed, and can meet the seedling alignment and navigation requirements of intelligent machinery at high speeds, while also providing more accurate curve navigation capabilities for intelligent agricultural machinery. Attached image description:

[0008] Figure 1 This is a technical flowchart of the present invention;

[0009] Figure 2 This is the original map of the cornfield of this invention;

[0010] Figure 3 This is the image-processed and optimized binary image of the seedling strips of the present invention;

[0011] Figure 4 This is a diagram showing the plant characteristic points and navigation lines of the present invention;

[0012] Figure 5 This invention provides an optimized fitting curve diagram of the two seedling strip correction lines obtained after the longitudinal slider performs a transverse scan. Detailed implementation method:

[0013] The present invention will now be described in detail with reference to the accompanying drawings.

[0014] like Figure 1 As shown, this invention proposes a curve seedling strip navigation algorithm based on horizontal and vertical dual slider scanning extraction. The algorithm includes (1) acquiring field images using a color camera and compressing them; (2) extracting the seedling strip range using ROI; (3) extracting the seedling strip area of ​​green plants using an image color segmentation extraction algorithm and performing morphological processing such as image binarization and opening / closing operations; (4) using a horizontal slider to perform vertical scanning of the image, obtaining the boundary coordinates of the connected components within the slider and performing mean calculation, dividing the left and right seedling strip feature point coordinates and collecting them;

[0015] (5) Simultaneously perform perspective transformation on the image in step C to obtain a top-view binary image of the seedling strip; (6) Use the vertical slider to scan the image horizontally to obtain the pixel distribution curve and perform curve fitting and smoothing processing to obtain the maximum value coordinates to obtain the endpoint coordinates (x1,0)(x2,0)(x1,w)(x2,w) of the two seedling strip lines; (7) Project the two seedling strip lines y1, y2 back to the original image, use the straight lines y1, y2 to fit and optimize the point set, and calculate the average coordinates of the left and right seedling strips to obtain the coordinates of the navigation line; (8) Perform polynomial fitting on the coordinates of the navigation line to obtain the curved navigation line; (9) Organize the coordinate information obtained in S4 and return it to correct the ROI region in S2; (10) Repeat steps S1 to S9.

Claims

1. A method for extracting curved seedling guide lines based on horizontal and vertical dual-slider scanning correction, characterized in that, The steps are as follows: S1: Use a color camera to acquire field images and compress them; S2: Use ROI to extract the seedling strip extent; S3: Extract the seedling strip region of green plants using an image color segmentation extraction algorithm, and perform image binarization and opening / closing operations for morphological processing; S4: Use the horizontal slider to scan the image vertically, obtain the coordinates of the connected domain boundaries within the slider and perform mean calculation, divide the left and right seedling band feature point coordinates and collect them; S5: Simultaneously perform perspective transformation on the image from step 3 to obtain a top-view binary image of the seedling strip; S6: Use the vertical slider to scan the image horizontally to obtain the pixel distribution curve and perform curve fitting and smoothing. Obtain the coordinates of the maximum value to get the coordinates of the endpoints of the two seedling strips: (x1,0) (x2,0) (x1,w) (x2,w); S7: Project the two seedling strip lines y1 and y2 back to the original image, use the lines y1 and y2 to fit and optimize the point set, and calculate the coordinates of the navigation line by averaging the coordinates of the left and right seedling strips; S8: Perform polynomial fitting on the navigation line coordinates to obtain the curved navigation line; S9: Organize the coordinate information obtained in S4 and return it to correct the ROI region in S2.

2. The method for extracting the navigation line of a seedling strip based on horizontal and vertical dual-slider scanning correction according to claim 1, characterized in that, In S4, the image is scanned vertically using a w×10 horizontal slider.

3. The method for extracting the navigation line of a seedling strip based on horizontal and vertical dual-slider scanning correction according to claim 1, characterized in that, In S6, the image is scanned horizontally using a 10×h vertical slider.

4. The method for extracting the navigation line of a seedling strip based on horizontal and vertical dual-slider scanning correction according to claim 1, characterized in that, In S6, the pixel distribution curve is smoothed by polynomial fitting to obtain a better solution with the maximum value.

5. The method for extracting the navigation line of a seedling strip based on horizontal and vertical dual-slider scanning correction according to claim 1, characterized in that, In S7, the feature points are corrected and optimized using linear seedling lines to fit a more accurate nonlinear navigation line.

6. The method for extracting the navigation line of a seedling strip based on horizontal and vertical dual-slider scanning correction according to claim 1, characterized in that, In S9, the seedling data from the previous frame obtained in S4 is used to return the ROI extraction process in S2 to correct the ROI range and improve the extraction accuracy and adaptability of the ROI region.

Images