An online detection method for seeding performance

Abstract

The present invention discloses an on-line detection method for seeding performance, which comprises the following steps: loading an image and preprocessing the image; determining adhered seeds; segmenting the adhered seeds; analyzing the seeding uniformity, accuracy and dispersion of the segmented seeds, so as to realize on-line detection of the seeding performance of a seeder; using the present invention can quickly and accurately segment the adhered seeds, and has high accuracy in performance detection.

Description

Technical Field

[0001] The present invention relates to the technical field of sowing detection, in particular to an on-line detection method for sowing performance. Background Art

[0002] During the sowing process of a seeder, the seeds in the seed box are discharged downward onto the ground, and then the subsequent soil covering device covers the seeds with soil to meet the agronomic sowing requirements. The detection of seed sowing is an important link in agricultural production and has a very important impact on the subsequent crop growth. However, the current sowing devices cannot on-line detect sufficient sowing information of the sown seeds on the ground.

[0003] For example, in the prior art, in the paper published in the journal Transactions of the Chinese Society of Agricultural Engineering, titled "Performance Detection Experiment of Precision Seed Metering Device Based on High-Speed Camera System", during the experiment, the falling seeds were photographed by a high-speed camera system, the tape moved at a constant speed, and the castor oil was evenly sprayed onto the tape surface by an oil pump to reduce the rolling and bouncing of the seeds after they fell onto the tape. The positions of the precision seed metering device and the camera were kept fixed. It could only photograph the seeds during the falling process and could not detect the sowing performance of the seeds after they landed. Summary of the Invention

[0004] The purpose of this part is to outline some aspects of the embodiments of the present invention and briefly introduce some preferred embodiments. Some simplifications or omissions may be made in this part, as well as in the abstract and title of the specification of this application, to avoid obscuring the purpose of this part, the abstract, and the title. However, such simplifications or omissions cannot be used to limit the scope of the present invention.

[0005] In view of the above and / or existing problems in sowing detection, the present invention is proposed.

[0006] Therefore, the purpose of the present invention is to provide an on-line detection method for sowing performance, which can quickly and accurately segment the adhered seeds to achieve accurate counting of the seeds, with high detection accuracy, and solves the technical problem of low accuracy in sowing performance detection in the prior art.

[0007] To solve the above technical problems, the present invention provides the following technical solution: An on-line detection method for sowing performance, comprising the following steps:

[0008] Loading an image and preprocessing the image;

[0009] Determining the adhered seeds;

[0010] Segmenting the adhered seeds;

[0011] Analyzing the sowing uniformity, accuracy, and dispersion of the segmented seeds, so as to realize the on-line detection of the sowing performance of the seeder.

[0012] As a preferred solution of the online detection method for seeding performance in the present invention, wherein: the step of determining sticky seeds specifically is that,

[0013] For the connected regions in the image, the area S of the connected region s and the perimeter L of the contour of the connected region s are used in a combined constraint to determine sticky seeds. When the formula (1) is satisfied, the connected region is a single seed, otherwise it is a sticky seed;

[0014]

[0015] wherein, S max is the set threshold value of the area of the connected region, and L max is the set threshold value of the perimeter of the contour of the connected region.

[0016] As a preferred solution of the online detection method for seeding performance in the present invention, wherein: the step of performing sticky segmentation on sticky seeds specifically is that,

[0017] All the points of the sticky seeds are surrounded by an irregular convex polygon to establish a convex hull;

[0018] A contour image is drawn closely along the outer edge of the sticky seeds;

[0019] The convex hull defect image is obtained by subtracting the contour image from the established convex hull, and perpendicular lines to the convex hull line segments are made inside the convex hull defect image, so that the other end of the perpendicular line intersects the edge of the contour image. The point where the longest perpendicular line intersects the contour edge is the concave point;

[0020] A circle is established with the concave point as the center and 1 / 3 of the distance between the two nearest concave points as the radius. The established circle forms two intersection points with the contour image;

[0021] The concave point is connected to the two intersection points to form a closed triangle. All the concave points are connected pairwise, and the formed connections must only pass through the seed connected region and the closed triangle;

[0022] The sticky seeds are segmented by connecting the concave points.

[0023] As a preferred solution of the online detection method for seeding performance in the present invention, wherein: the preprocessing of the image includes image graying, image filtering, image binarization and image morphological processing. Among them, the image graying adopts the weighted average method for graying, and its formula is as (2),

[0024] Gray(i, j) = 0.299 * R(i, j) + 0.578 * G(i, j) + 0.114 * B(i, j) (2);

[0025] The image filtering adopts median filtering, and the image binarization adopts the Otsu method, and its formula is as (3),

[0026]

[0027] Use a rectangular (3, 3) convolution kernel to perform 2 erosion operations on the seed image.

[0028] As a preferred solution of the online detection method for seeding performance in the present invention, wherein: analyze the seeding uniformity U, seeding accuracy μ, and dispersion V of the segmented seeds. Specifically,

[0029]

[0030] where X is the number of seeds in the image, is the average value of X, S is the standard deviation of X, m1 is the number of seeds within a specified pixel range in the image, and m2 is the total number of seeds in the image. is the average value of yi, S′ is the standard deviation of yi, and n is the number of seeds within the specified pixel range.

[0031] Compared with the prior art, the present invention has the following technical effects: Through S s and L s The composite constraint is used to determine whether the connected region is a single seed or an adhered seed, and the judgment accuracy is high; it can quickly and accurately segment the adhered seeds to achieve accurate counting of the seeds, and the detection accuracy is high; it can be applied to the detection of seeding conditions in indoor or field environments. BRIEF DESCRIPTION OF THE DRAWINGS

[0032] In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the following will briefly introduce the drawings required for the description of the embodiments. Obviously, the following drawings are only some embodiments of the present invention. For those of ordinary skill in the art, other drawings can be obtained based on these drawings without creative efforts. Among them:

[0033] Figure 1 is a three-dimensional structural schematic diagram of the detection device used in the present invention.

[0034] Figure 2 is the gray-scale effect diagram after image processing.

[0035] Figure 3 is the image output after image filtering and binarization processing.

[0036] Figure 4 is the seed image after 2 erosion processes in the present invention.

[0037] Figure 5 is the seed connected region diagram in the present invention.

[0038] Figure 6 Schematic diagram of seed cutting in the present invention.

[0039] Figure 7 Schematic diagram of cutting seeds when constructing circles with radii of 1 / 2 and 1 / 4 of the distance between the two nearest concave points respectively.

[0040] Figure 8 Schematic diagram of tandem adhesion cutting of seeds.

[0041] Figure 9 Effect diagram of segmentation after using the cutting algorithm in the present invention.

[0042] Figure 10 Schematic diagram of seed counting and coordinate positioning processing in the present invention.

[0043] In the figure, 1 is a sowing test stand, 2 is a driving wheel, 3 is a driving motor, 4 is a conveyor belt, 5 is a detection box, 6 is a seed discharging pipe, 7 is an electric seed discharger, 8 is a sowing box, and 9 is a driven wheel. Specific embodiments

[0044] In order to make the above objects, features, and advantages of the present invention more obvious and understandable, the following will describe the specific embodiments of the present invention in detail with reference to the accompanying drawings of the specification.

[0045] In the following description, many specific details are set forth in order to fully understand the present invention. However, the present invention can also be implemented in other ways different from those described herein. Those skilled in the art can make similar extensions without departing from the connotation of the present invention. Therefore, the present invention is not limited by the specific embodiments disclosed below.

[0046] Secondly, the so-called "one embodiment" or "embodiment" herein refers to a specific feature, structure, or characteristic that can be included in at least one implementation manner of the present invention. The "in one embodiment" that appears in different places in this specification does not necessarily refer to the same embodiment, nor is it an individual or alternative embodiment that is mutually exclusive with other embodiments.

[0047] Embodiment 1

[0048] Referring to Figures 1 to 5 , which is the first embodiment of the present invention. This embodiment provides an on-line detection device for sowing performance, which can detect the sowing situation.

[0049] An on-line detection device for seeding performance, which comprises a seeding test stand 1. The two ends of the seeding test stand 1 in the length direction are respectively rotatably connected with a driving wheel 2 and a driven wheel 9. One side of the seeding test stand 1 in the front-back direction is fixedly connected with a driving motor 3. The driving motor 3 is connected with the driving wheel 2. A detection support is fixedly connected to the seeding test stand 1, and a detection box 5 is fixedly connected to the detection support. An image acquisition camera (not shown, which is prior art) is connected in the detection box 5. The driving wheel 2 is connected with the driven wheel 9 through a conveyor belt 4. A seeding box 8 is fixedly connected to the seeding test stand 1 above the conveyor belt 4. An electric seed metering device 7 is fixedly connected to the lower side of the seeding box 8. A seed discharging pipe 6 is fixedly connected to the lower side of the seeding box 8. The seeds discharged by the electric seed metering device 7 fall onto the conveyor belt 4 below through the seed discharging pipe 6. The image acquisition camera can take pictures of the seeds falling onto the conveyor belt 4.

[0050] During seeding experiments, the electric seed metering device 7 operates, and the seeds are discharged downward. The driving motor 3 operates, and the seeds fall onto the moving conveyor belt 4 in sequence. The image acquisition camera works to take pictures of the seeds falling onto the conveyor belt 4, realizing the image acquisition of the seeds. There is a detection analyzer externally. The collected images are sent to the detection analyzer for detection and analysis.

[0051] Embodiment 2

[0052] Referring to Figures 2 to 9 , which is the second embodiment of the present invention. Based on the first embodiment, this embodiment provides an on-line detection method for seeding performance, which can accurately detect the seed dropping situation.

[0053] An on-line detection method for seeding performance includes the following steps.

[0054] S1: The graphic acquisition camera takes pictures of the seed dropping situation on the conveyor belt.

[0055] S2: The detection analyzer receives the seed images, loads the images and performs image preprocessing. The preprocessing of the images includes image graying, image filtering, image binarization and image morphology processing. Among them, the image graying adopts the weighted average method for graying, and its formula is as shown in (2).

[0056] Gray(i, j) = 0.299 * R(i, j) + 0.578 * G(i, j) + 0.114 * B(i, j) (2);

[0057] The image filtering adopts median filtering, and the image binarization adopts the Otsu method, and its formula is as shown in (3).

[0058]

[0059] Use a rectangular (3, 3) convolution kernel to perform 2 erosion operations on the seed images.

[0060] Among them, R(i, j), G(i, j), and B(i, j) represent the red, green, and blue pixel values of the (i, j)th pixel respectively. Grayscaling by the image weighted average method can reduce the data volume of the image and facilitate further processing of the image;

[0061] S3: Determine the adhered seeds. Specifically, for the connected regions in the image, use the connected region area S s and the perimeter L of the connected region contour s combined constraints to determine the adhered seeds. When the formula (1) is satisfied, the connected region is a single seed, otherwise it is an adhered seed;

[0062]

[0063] Among them, S max is the set threshold of the connected region area, and L max is the set threshold of the connected region contour perimeter. In this embodiment, S max is preferably 112 (unit: pixel), and L max is preferably 45.3 (unit: pixel);

[0064] S4: Perform adhesion segmentation on the adhered seeds, specifically including,

[0065] S401, surround all the points of the adhered seeds with an irregular convex polygon to establish a convex hull;

[0066] S402, draw a contour image closely along the outer edge of the adhered seeds;

[0067] S403, subtract the contour image from the established convex hull to obtain a convex hull defect image, and draw a perpendicular line to the convex hull line segment inside the convex hull defect image, so that the other end of the perpendicular line intersects the contour image edge. The point where the longest perpendicular line intersects the contour edge is the concave point;

[0068] S404, use the concave point as the center and 1 / 3 of the distance between the two nearest concave points as the radius to establish a circle. The established circle forms two intersection points with the contour image;

[0069] S405, connect the concave point with the two intersection points to form a closed triangle. Connect all the concave points pairwise, and the formed connection lines must only pass through the seed connected region and the closed triangle;

[0070] S406, realize the segmentation of the adhered seeds through the connection lines of the concave points;

[0071] Analysis of the seeding uniformity U, seeding accuracy μ, and dispersion V, specifically,

[0072]

[0073] Among them, X is the number of seed grains in the image, is the average value of X, S is the standard deviation of X, m1 is the number of seeds within the specified pixel range in the image, and m2 is the total number of seeds in the image. is the average value of yi, S′ is the standard deviation of yi, and n is the number of seeds within the specified pixel range.

[0074] Figure 2 is the output result of graying the seed image by the weighted average method. Its gray histogram has an obvious bimodal shape, and there is an obvious segmentation threshold between the two peaks. The resulting grayed seed image has a good effect, can achieve an ideal segmentation of the pixels of the seed image and the background image, and can meet the requirements of subsequent binarization.

[0075] The median filtering method is used to process the image. It can effectively suppress the noise in the image, remove unnecessary impurities and meaningless noise points in the image, enhance the image quality, and the output result is as shown in Figure 3 (a) in. Binarization is to convert the pixel gray value of the image to 0 or 255 on the basis of the gray scale image, presenting the entire image in black and white. In this application, the Otsu method is used for image binarization processing. It is a gray-scale image adaptive threshold segmentation algorithm that can automatically calculate the threshold according to the gray histogram, distinguish the seed image and the background image to the greatest extent, and the output result is as shown in Figure 3 (b) in.

[0076] In this application, a convolution kernel of rectangle (3,3) is used. The seed image after 2 times of erosion processing on the seed image is as shown in Figure 4 shown, effectively eliminating meaningless noise points, and at the same time can eliminate the image boundary points and segment some slightly adhered seed images.

[0077] A connected region refers to an image region composed of foreground pixels with the same pixel value and adjacent positions. The area S of the connected region s refers to the number of seed pixels contained in the connected region, and the contour perimeter L of the connected region s refers to the number of pixels required to surround the edge of the connected region. Whether the seeds in the connected region are adhered determines the subsequent processing of the connected region. The seed connected region map is as shown in Figure 5 shown.

[0078] Specifically, when dividing seeds, an irregular convex polygon is used to enclose all the points of the adhered seeds to establish a convex hull, as shown in Figure 6 (a) in; draw a contour image closely along the outer edge of the adhered seeds, as shown in Figure 6 (b) in; subtract the contour image from the established convex hull to obtain a convex hull defect image, and draw a perpendicular line to the convex hull line segment within the convex hull defect image, so that the other end of the perpendicular line intersects the edge of the contour image, as shown in Figure 6as shown in (c) therein; the point where the longest perpendicular line intersects the contour edge is the concave point, such as Figure 6 as shown in (d) therein; a circle is constructed with the concave point as the center and 1 / 3 of the distance between the two nearest concave points as the radius, such as Figure 6 as shown in (e) therein; the constructed circle forms two intersection points with the contour image, such as Figure 6 as shown in (f) therein; the concave point is connected to the two intersection points to form a closed triangle, such as Figure 6 as shown in (g) therein; all concave points are connected pairwise, and the formed connecting lines must pass through the seed connected region and the closed triangle, such as Figure 6 as shown in (h) therein; finally, the adhesion of seeds is achieved by connecting the concave points.

[0079] When dividing seeds, a circle is constructed with the concave point as the center and 1 / 2 of the distance between the two nearest concave points as the radius, such as Figure 7 as shown in (a) therein, over-segmentation occurs; a circle is constructed with 1 / 4 of the distance between the two nearest concave points as the radius, such as Figure 7 as shown in (b) therein, under-segmentation occurs, while in this embodiment, preferably, a circle is constructed with 1 / 3 of the distance between the two nearest concave points as the radius, which can achieve accurate segmentation of the adhered seeds.

[0080] The adhesion methods of seeds in the connected region are different, which can be roughly divided into a mixed adhesion method and a series adhesion method. For the mixed adhesion seeds, the segmentation effect is as Figure 6 shown, and for the series adhesion seeds, the segmentation effect is as Figure 8 shown. Through the segmentation method in the present invention, accurate segmentation of seeds can be achieved.

[0081] Figure 9 is Figure 5 the segmentation effect based on the segmentation algorithm of the present invention. Seeds that are still adhered after 2 erosion operations can be effectively separated by improving the convex hull algorithm. The connected regions formed by the adhered seeds form separate connected regions under the segmentation of the segmentation line, achieving the expected segmentation.

[0082] The present invention uses S s and L s compound constraints to judge whether the connected region is a single seed or an adhered seed, with a high accuracy of judgment; it can quickly and accurately segment the adhered seeds to achieve accurate counting of seeds, with a high accuracy of detection; it can be applied to the detection of sowing situations in indoor or field environments.

[0083] Embodiment 3

[0084] Referring to the figure, it is the third embodiment of the present invention. Based on the second embodiment, experimental means are used to verify the technical effect that the use of the present invention can achieve accurate detection of sowing situations.

[0085] A seeding experiment is carried out in the laboratory. In this embodiment, the detection analyzer is a computer. The image acquisition camera transmits data to the computer, and the computer controls the photographing time of the image acquisition camera. The functional parameters of the detection device are shown in Table 1, and the parameters of the image acquisition camera are shown in Table 2.

[0086] Table 1 Functional parameters of the detection device

[0087]

[0088] Table 2 Parameter table of the image acquisition camera

[0089]

[0090] When performing seeding performance detection, first start the drive motor. After the conveyor belt rotates stably, start the electric seed metering device. The electric seed metering device guides the seeds from the seed box through the external groove wheel into the seed discharging pipe and then onto the conveyor belt. The conveyor belt carries the seeds to the image acquisition camera for photographing, and then the image is transmitted into the recognition algorithm of the computer. After the recognition method of the present invention, the adhesion of seeds is segmented, the area, perimeter and coordinate information of each seed are counted, and the accurate counting of seeds is achieved. Finally, the calculation and evaluation are completed according to the seeding performance index.

[0091] The seeding dispersion in this application refers to the degree of dispersion of seeds within the width of the seed groove. Five consecutive images of the seed belt are taken on the conveyor belt, with a pixel of 800×450. The seeds falling within the pixel range of 168 - 281 in the y direction are located through the algorithm, and the distance y formed by each seed and the straight line with the pixel value of 168 in the y direction is recorded. i , The corresponding relationship between the seeding strip width and the image pixels is as Figure 10 shown.

[0092] In this experiment, the seeding rate used is 150 - 225 kg / hm 2 , and after measurement, the thousand - grain weight of wheat used is 40 g. The seeding rate of wheat converted according to the planting row spacing of 25 cm is about 15 - 36 g / m 2 , During the experiment, a moderate seeding speed of 0.22 m / s is selected. The system detection results are shown in Table 8. It can be found from Table 3 that the algorithm accuracy rate is above 93% under different seeding rates, and the overall average algorithm accuracy rate is 95.15%, indicating that the recognition accuracy of the total number of seed particles using the present invention is relatively high.

[0093] Table 3 Detection results under different seeding rates

[0094]

[0095] When conducting the system detection effect test under different seeding speeds, a moderate seeding rate of 24 g / m is selected. 2For the seeding rate, the system detection results are shown in Table 9. It can be found from Table 9 that the algorithm accuracy is above 93.19% at different seeding traveling speeds, and the overall average algorithm accuracy is 95.13%, which is comparable to the test results of different seeding rates.

[0096] Table 4 Detection Results under Different Seeding Traveling Speeds

[0097]

[0098] From the analysis of the above two groups of test results, within a certain range of seeding rate and seeding traveling speed conditions, the influence on seeding uniformity, accuracy, and dispersion is not significant, indicating the feasibility of using the present invention to achieve seeding detection.

Claims

1. An on-line detection method for seeding performance, comprising the following steps: Loading an image and image preprocessing; Judging adhered seeds; Performing adhesion segmentation on the adhered seeds, specifically: Enclosing all points of the adhered seeds with an irregular convex polygon to establish a convex hull; Drawing a contour image closely along the outer edge of the adhered seeds; Subtracting the contour image from the established convex hull to obtain a convex hull defect image, and drawing perpendicular lines to the convex hull line segments within the convex hull defect image, so that the other end of the perpendicular line intersects the edge of the contour image. The point where the longest perpendicular line intersects the contour edge is the concave point; Taking the concave point as the center and 1 / 3 of the distance between the two nearest concave points as the radius to establish a circle. The established circle forms two intersection points with the contour image; Connecting the concave point with the two intersection points to form a closed triangle. Connecting all concave points pairwise, and the formed connections must only pass through the seed connected area and the closed triangle; Realizing the segmentation of the adhered seeds through the connection of the concave points; Analyzing the seeding uniformity, accuracy and dispersion of the segmented seeds, so as to realize the on-line detection of the seeding performance of the seeder.

2. The online detection method for seeding performance according to claim 1, characterized in that: The step of judging the adhered seeds specifically is: Use the connected region area S of the connected regions in the image s and the perimeter L of the contour of the connected region s A composite constraint is used to determine the adhered seeds. When the formula (1) is satisfied, the connected region is a single seed, otherwise it is an adhered seed; Among them, S max is the set area threshold of the connected region, and L max is the set perimeter threshold of the connected region contour.

3. The online detection method for seeding performance according to claim 1 or 2, characterized in that: The preprocessing of the image includes image graying, image filtering, image binarization and image morphological processing. Among them, the image graying adopts the weighted average method for graying, and its formula is as (2), Gray(i, j) = 0.299 * R(i, j) + 0.578 * G(i, j) + 0.114 * B(i, j) (2); The image filtering adopts median filtering, and the image binarization adopts the Otsu method, and its formula is as (3), Using a rectangular (3, 3) convolution kernel to perform 2 erosion operations on the seed image.

4. The on-line detection method for seeding performance according to claim 3, characterized in that: Analyzing the seeding uniformity U, seeding accuracy μ and dispersion V of the segmented seeds, specifically: where X is the number of seed grains in the image, is the average value of X, S is the standard deviation of X, m1 is the number of seeds within the specified pixel range in the image, and m2 is the total number of seeds in the image, is the average value of yi, and S′ is the standard deviation of y i and n is the number of seeds within the specified pixel range.

Images