Automatic film breaking method based on combination of machine vision and laser technology

Abstract

The invention discloses a cotton seedling detection and laser membrane rupture control method based on improved YOLOv11 and image processing, and belongs to the technical field of agricultural automatic detection and control. According to the method, aiming at the problems of low precision and large damage of traditional mechanical film breaking, deep learning and image processing technologies are combined, and precise positioning and efficient film breaking of cotton seedlings are realized. The method comprises the following specific steps: carrying out target detection by adopting an improved YOLOv11 model integrated with a PConv-CGLU hybrid module, and optimizing bounding box regression precision in combination with an MPDIOU loss function; highlighting a green region of the cotton seedling by using HSV color space conversion and an ultra-green feature extraction algorithm, and calculating a centroid coordinate through contour analysis; setting a threshold value based on the contour area to determine the size of the cotton seedling, and dynamically adjusting the membrane rupture radius; and finally, the laser control unit controls the laser to accurately rupture the membrane according to the center point coordinate and the contour information. The method improves the film breaking precision and efficiency, reduces the damage rate of cotton seedlings, and is suitable for automatic film breaking operation of field plastic film mulching crops.

Description

Technical Field

[0001] This invention relates to an automatic film breaking method based on machine vision combined with laser technology, belonging to the field of agricultural automation technology. Specifically, it is a method for automatically identifying and cutting cotton mulch film using machine vision and laser technology, applicable to scenarios such as automated film breaking of cotton seedlings and mulch film by agricultural robots. Background Technology

[0002] Mulching technology is widely used in cotton cultivation. Cotton mulch effectively retains heat and moisture, creating a suitable temperature and humidity environment for cotton seedling growth. Simultaneously, it suppresses weed growth, reducing competition for nutrients and sunlight between weeds and cotton, thus promoting healthy cotton growth. However, as the cotton growth cycle progresses, prolonged use of mulch can lead to numerous problems. For example, it hinders rainwater infiltration, affecting the cotton roots' absorption of water; it also restricts gas exchange between the soil and the outside environment, resulting in poor soil permeability, which is detrimental to root respiration and normal growth and development. Therefore, timely removal of the mulch is crucial for the sustained growth and high yield of cotton. However, traditional mulch removal usually requires significant manual intervention, which is inefficient and labor-intensive. Furthermore, the inconsistent pore sizes created by manual mulch removal result in poor heat and moisture retention.

[0003] While some automated film-breaking devices exist, most rely on mechanical means to tear or cut the mulch film, often resulting in irregular tears and significant damage to the soil and cotton plants. These devices typically lack precise cotton seedling identification and cutting technology, leading to poor film-breaking efficiency. Therefore, there is an urgent need for an automated film-breaking method that can accurately identify cotton seedlings beneath the mulch film and precisely cut the film. Summary of the Invention

[0004] The purpose of this invention is to provide an automatic film-breaking method based on machine vision combined with laser technology, which can accurately identify cotton seedlings under the mulch film and determine the film-breaking radius according to the size of the cotton seedlings. This method is used for real-time detection and precise film breaking of cotton seedlings under mulch film in precision agriculture. The technical solution of this invention includes using an improved YOLOv11 deep learning algorithm for preliminary cotton seedling detection, combining image processing methods to extract the center point position and size of the cotton seedlings, and using a laser system to cut the mulch film, thereby improving the film-breaking efficiency.

[0005] This invention provides an automatic membrane breaking method, comprising the following steps: (1) Use the improved YOLOv11 deep learning model to perform target detection on the collected cotton seedling images under the film. Obtain the position coordinates of the detection box; (2) Feature extraction: Extract the feature regions of the cotton seedlings effectively by converting the detected cotton seedlings according to the specific color space conversion and threshold processing ultra-green feature extraction method; (3) Transmit the relevant data of cotton seedling identification to the laser control unit. The laser control unit automatically generates the corresponding laser trajectory according to the size and position of each cotton seedling and controls the laser to cut the film.

[0006] As a further technical solution, cotton seedling target detection based on YOLOv11 includes: using an improved YOLOv11 deep learning model integrating a PConv-GULU combined module and an MPDIoU loss function to perform target detection on acquired images of cotton seedlings under plastic film. The trained YOLOv11 model can effectively identify cotton seedlings under plastic film in complex environments. YOLOv11 can quickly extract the location information of cotton seedlings from images and output the location coordinates of detection boxes, improving the separation effect between plastic film and cotton seedlings.

[0007] As a further technical solution, based on the morphological feature extraction of cotton seedlings, it includes: extracting ultragreen features from detected cotton seedlings according to the specific color space conversion and threshold processing method, such as... Figure 2 As shown, the algorithm accurately extracts the green portion of the cotton seedling area within each segmentation box. Through specific color space conversion and thresholding methods, the algorithm effectively extracts the green area, thereby enhancing the accuracy of cotton seedling identification and overcoming the influence of changes in ambient lighting.

[0008] As a further technical solution, based on the determination of the cotton seedling membrane breakage radius, the following steps are taken: morphological denoising is performed on the extracted green area, and then the denoising is performed according to the formula... N is The total number of pixels in the region, calculated Area of ​​target pixels within the region According to the area To approximate the size of cotton seedlings, 5000 < For large cotton seedlings, 5000< <1000 indicates medium-sized cotton seedlings. <1000 are small cotton seedlings and according to the formula , Find the coordinates of the center point of the cotton seedling. , In the binary image, the pixel value of a given pixel (x, y) is such that if that pixel belongs to the target region, then... =1 otherwise =0. and These represent the summation of all x and y coordinates in the image, respectively.

[0009] As a further technical solution, the generation of laser trajectory includes: calculating the radius of the circle based on the obtained area and coordinates of the center point, and sending the command to the control card. The control card controls the laser generator and galvanometer module to draw the corresponding film breaking trajectory to break the film. Beneficial effects

[0010] Compared with existing technologies, the beneficial effects of this invention are as follows: 1. By using the improved YOLOv11 model, image recognition accuracy can be effectively improved, noise interference can be reduced, accurate separation of cotton seedlings and mulch film can be ensured, and the stability and accuracy of the film breaking process can be improved; 2. By utilizing the Supergreen feature extraction algorithm, this invention can accurately extract the feature parts of the cotton seedling area, improving the reliability of cotton seedling identification and scale determination; 3. By using real-time adjustment of laser parameters and path during the laser cutting process, the accuracy and efficiency of the cutting process can be ensured, damage to cotton plants can be reduced, and physical damage to soil and plants that may be caused by traditional mechanical cutting methods can be avoided. Attached Figure Description

[0011] The accompanying drawings and tables are provided to further illustrate the invention and form part of the specification. They are used in conjunction with examples of the invention to explain it, but do not constitute a limitation thereof. In the drawings and tables: Figure 1 This is a schematic diagram of the steps of a membrane breaking method provided in an example of the present invention; Figure 2 The image obtained using object detection and super-green feature extraction algorithms; Figure 3 The center point, parameter coordinates, and membrane rupture radius results are provided for the examples of this invention. Detailed Implementation

[0012] The present invention will be further described in detail below with reference to the accompanying drawings.

[0013] The automatic film-breaking method of laser technology of the present invention combines YOLOv11 deep learning target detection with traditional image processing technology. It aims to accurately locate the center point coordinates and shape and size of cotton seedlings, and is especially suitable for crops covered by plastic film in field environment, ensuring stable and effective positioning in complex environment.

[0014] like Figure 1As shown, the automatic film-breaking method of the present invention, which combines machine vision with laser technology, first uses an improved YOLOv11 deep learning model to perform target detection on crop images, quickly identifying the target. Then, background interference is removed through specific color space conversion and thresholding, and feature information of the cotton seedlings is extracted. The center point coordinates of each cotton seedling are calculated based on the feature information; the relevant data of the identified cotton seedlings are transmitted to the laser control unit, which automatically generates the corresponding laser trajectory based on the size and position of each cotton seedling and controls the laser to cut the mulch film.

[0015] Step 1. Perform random brightness and contrast adjustments and add Gaussian noise to the acquired sub-membrane images to enhance the model's generalization ability. An improved YOLOv11 model is used for target detection: its backbone network integrates a hybrid PConv-CGLU module to enhance the ability to capture features of the target region; the loss function is MPDIoU, whose calculation method considers information such as overlapping regions, center point distance, and structural differences, expressed as: ; ; ; In the formula, A and B are the ground truth bounding box and the predicted bounding box, respectively, h is the height of the input image, and w is the width of the input image. This is the Euclidean distance between the top-left corners of the ground truth bounding box and the predicted bounding box. This is the Euclidean distance between the bottom right corners of the ground truth bounding box and the predicted bounding box. , , , These represent the coordinates of the top-left and bottom-right corners of the actual bounding box, respectively. , , , These represent the coordinates of the top left and bottom right corners of the prediction box, respectively.

[0016] Step 2. Based on the bounding box coordinates obtained in Step 1, the outline of the cotton seedling is extracted by analyzing the green features of the cotton seedling leaves. The green parts of the cotton seedling have significant color differences in the image, which can be effectively enhanced by using a color threshold to improve the outline recognition of the cotton seedling. Specifically, the input image is converted from the RGB color space to the HSV color space, making the separation of the green areas clearer. By setting appropriate chromaticity thresholds and saturation conditions, the cotton seedling is separated from the background, thus obtaining a clear outline of the cotton seedling. The calculation formula for the super-green feature extraction algorithm is: ExG=2 G - R - B； in, R ,G , B These represent the red, green, and blue channel values ​​of the image in the RGB color space, respectively. This formula highlights green vegetation areas by weighted combination of RGB channel values, while suppressing interference from other colors (such as soil, shadows, and withered grass).

[0017] Step 3. Based on the green area extracted in Step 2, calculate the centroid coordinates of the cotton seedling outline using the contour analysis method. The final precise center point coordinates of the cotton seedlings were calculated using a weighted formula. The formula for calculating coordinates is: ; ; ; ; In the formula , σ represents the offset relative to the network unit in the x and y directions; σ is the sigmoid activation function that ensures the output value is between [0,1]. and The coordinates of the top-left corner of the current network cell; , ) represents the coordinates of each point on the contour; n is the number of contour points; is a weighting coefficient used to balance the influence of the YOLOv11 model detection results and the supergreen feature extraction results.

[0018] Step 4. Calculate the outline area A using the number of pixels in the cotton seedling outline region from Step 3. The outline area A is represented by counting the number of pixels within the outline. The formula for calculating the outline area is: In the formula, R is the number of pixels and A is the area of ​​the cotton seedling outline, both in pixels.

[0019] Step 5. Based on the area A calculated in Step 4, two threshold values ​​can be set. , To determine the size and shape of cotton seedlings. When A < At that time, it was determined to be a small cotton seedling. <A< , determined to be medium-sized cotton seedlings, when A> At that time, it was determined to be a large cotton seedling.

[0020] Step 6. Based on the outline area and cotton seedling size obtained in Steps 4 and 5, approximate the size of the membrane rupture. Membrane rupture size. The approximate definition of is: In the formula It is an adjustment factor that indicates that during laser film breaking operation, the laser treatment area may be slightly larger than the actual contour.

[0021] Step 7. Figure 3 As shown, cotton seedlings in different areas may require different film-breaking radii, therefore the laser processing range and film-breaking size need to be adjusted accordingly. The precise definition of is: ;in, This is a dynamic adjustment factor used to adjust the size of the area where the film breaks down according to the actual situation. Assuming the presence of cotton seedlings under the film is confirmed, the improved YOLOv11 model, combined with the Supergreen feature extraction algorithm, can more accurately control the area where the film breaks down, thereby improving operational efficiency, reducing resource waste, and providing a theoretical basis and technical support for subsequent automated operations.

[0022] Step 8. The obtained cotton seedling position coordinates are transmitted to the laser control unit through coordinate transformation. The laser control unit determines the film breaking radius based on the center point coordinates, feature contours, and the pre-set film breaking radius calculation formula, and precisely controls the laser to perform the film breaking operation, thus completing the precise removal of the mulch film around the cotton seedlings.

[0023] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, and not to limit them; although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some or all of the technical features therein; and these modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the technical solutions of the embodiments of the present invention.

Claims

1. An automatic membrane breaking method based on machine vision combined with laser technology, characterized in that, include: An improved YOLOv11 deep learning model integrating a PConv-CGLU combined module and the MPDIoU loss function was used to perform target detection on the acquired cotton seedling images under the mulch film, obtaining the bounding box coordinates of the cotton seedling targets. Based on the regions located by the target boxes, features were extracted from these regions using the super-green feature extraction algorithm. The size of each cotton seedling was determined based on the area of ​​the green region, and the center point coordinates of each cotton seedling were calculated. The relevant data of the cotton seedling identification was transmitted to the laser control unit, which automatically generated the corresponding laser trajectory based on the size and position of each cotton seedling and controlled the laser to cut the mulch film.

2. The automatic film breaking method using laser technology as described in claim 1, characterized in that, The improved YOLOv11 deep learning model, which combines the PConv-GULU module with the MPDIoU loss function, specifically includes: replacing the C3K2 module in the YOLOv11 backbone and neck network with the PConv-GULU module; enhancing feature extraction for small-scale objects by combining separable convolutions with global context learning; and replacing the original CIoU loss function with the MPDIoU loss function, thereby improving the detection accuracy for small targets and effectively enhancing the performance of YOLOv11.

3. The automatic film breaking method using laser technology as described in claim 1, characterized in that, The super-green feature extraction algorithm extracts the outline of cotton seedlings, specifically by converting the image from the RGB color space to the HSV color space, and using the H and S components of the H (hue), S (saturation), and V (luminance) components to set a specific threshold range to extract the green areas. This method helps reduce background interference, thereby reducing missed detections and false detections.

4. The automatic film breaking method using laser technology as described in claim 1, characterized in that, The method of determining the size and center point coordinates of the cotton seedling based on the area of ​​pixels occupied by the green region specifically includes estimating the size of the cotton seedling based on the ratio of the pixel area of ​​the green region within the detection box to the total area. Then, by weighted summing of each pixel in the green region according to its position, the pixel positions in the X and Y directions are normalized to obtain the average coordinates of the cotton seedling area. These coordinates are the center point coordinates of the cotton seedling, avoiding inaccurate estimations caused by edge blurring or noise interference.

5. The automatic film breaking method using laser technology as described in claim 1, characterized in that, Based on the center point coordinates of the cotton seedling and the area of ​​the green region, a suitable circular path for cutting is calculated. This circular trajectory data is then transmitted as an instruction to the laser control unit. The control unit adjusts the laser's movement trajectory according to these parameters to ensure that the laser accurately cuts the mulch film and avoids the cotton seedling itself.

Images