CN117788395A - System and method for extracting root phenotype parameters of pinus massoniana seedlings based on images - Google Patents

Abstract

The invention discloses an image-based system and method for extracting phenotype parameters of a pinus massoniana seedling root system, comprising the following steps: collecting an image of a root system of a pinus massoniana seedling and correcting the image; preprocessing the corrected image; performing image refinement operation on the processed image by using an improved ZhangSuen skeleton extraction algorithm; searching endpoints of pixel points in the root system skeleton diagram; performing turning point search on pixel points in the root skeleton diagram; dividing main roots and primary lateral roots of the root systems of the pinus massoniana seedlings by adopting a priority path searching algorithm; extracting root system phenotype parameters of the pinus massoniana seedlings based on the main roots and the primary lateral roots of the segmented pinus massoniana seedling roots; the system and the method can rapidly and accurately measure the root phenotype parameters of the pinus massoniana seedlings, avoid high cost and high error of manual measurement, and do not need to rely on expensive detection equipment.

Description

Image-based extraction system and method for root phenotypic parameters of Masson pine seedlings

Technical field

The invention relates to the field of smart forestry, specifically an image-based extraction system and method for phenotypic parameters of masson pine seedling root system.

Background technique

Masson pine is one of the pioneer tree species for afforestation in my country's subtropical regions. It grows rapidly, has high yield and good quality. It is a good raw material for construction and industrial production. It is very popular in the market and has good economic effects. The root phenotypic data of seedlings is a necessary condition for judging good varieties and strong seedlings. Therefore, it is of great theoretical and practical significance to study the root phenotypic parameter extraction method of masson pine seedlings to meet the rapid and efficient detection requirements for forestry seedling quality evaluation.

The traditional method of manually measuring root system parameters is costly, has large errors and is inefficient. The existing three-dimensional reconstruction equipment in general laboratories cannot meet the evaluation needs for the detection of finer root systems. Therefore, it is difficult to accurately construct a quality evaluation system for masson pine seedlings.

Contents of the invention

The technical problem to be solved by the present invention is to provide a low-cost image-based extraction system and method for root phenotypic parameters of masson pine seedlings in view of the shortcomings of the above-mentioned existing technologies. The system and method for extracting root phenotypic parameters of masson pine seedlings can quickly And it can accurately measure the root phenotypic parameters of masson pine seedlings, avoiding the high cost and high error of manual measurement, and there is no need to rely on expensive detection equipment.

In order to achieve the above technical objectives, the technical solutions adopted by the present invention are:

An image-based method for extracting root phenotypic parameters of masson pine seedlings, including:

Step 1: Collect masson pine seedling root system images and correct the images;

Step 2: Preprocess the corrected image;

Step 3: Use the improved ZhangSuen skeleton extraction algorithm on the processed image to refine the image to extract the root system skeleton of the masson pine seedlings;

Step 4: Based on the root system skeleton diagram obtained in step 3, construct the root system skeleton endpoint judgment conditions, conduct an endpoint search for the pixels in the root system skeleton diagram, and determine the pixel points that meet the judgment conditions as the endpoints of the root system to realize the root system of masson pine seedlings. endpoint positioning;

Step 5: Based on the root system skeleton diagram obtained in step 3, construct the root system skeleton turning point judgment conditions, conduct a turning point search on the pixels in the root system skeleton diagram, and determine the pixel points that meet the judgment conditions as the turning points of the root system to achieve horsetail. Rotation point positioning of pine seedling root system;

Step 6: Based on the end points and turning points of the masson pine seedling root system skeleton located in steps 4 and 5, use a priority path search algorithm with preferential directionality to segment the main root and first-level lateral roots of the masson pine seedling root system;

Step 7: Extract root phenotypic parameters of the Masson pine seedlings based on the segmented taproot and primary lateral roots of the Masson pine seedlings' root system.

As a further improved technical solution of the present invention, the step 1 is specifically:

Step 1.1, dig out the Masson pine seedlings from the soil, clean the soil impurities on the roots of the Masson pine seedlings with water, and lay the cleaned Masson pine seedling roots flat on a white background board that forms a sharp contrast with the black roots;

Step 1.2: Place a 3×4 checkerboard calibration board at a fixed position on the white background. The checkerboard size is 20×20 mm. It is used to calibrate the camera and obtain the camera's internal and external parameters. After collecting the image of the root system of Masson pine seedlings, correct the root system image according to the obtained camera's internal and external parameters.

As a further improved technical solution of the present invention, the step 2 specifically includes:

Step 2.1: The corrected root system image is transformed into a single channel by weighted average grayscale transformation. The weighted average formula is as follows:

Among them, g(x,y) is the pixel value of the output image, R is the pixel value in the R channel in the color image, G is the pixel value in the G channel in the color image, and B is the pixel value in the B channel in the color image. , w _R is the weight value given to the R component, w _G is the weight value given to the G component, w _B is the weight value given to the B component;

Step 2.2. Perform the global threshold method binarization operation on the grayscale image. The global threshold value range is (125,255). The global threshold method formula is as follows:

Where T is the selected threshold size, b(x,y) is the output image pixel;

Step 2.3. Use median filtering to remove noise that has not been eliminated or enhanced after image binarization. The filter kernel size of the median filtering algorithm is 5×5;

Step 2.4: Use the closing operation in morphological operations to fill and eliminate the gaps and holes in the root image and maintain the overall integrity of the root skeleton. The closing operation kernel size is 5×5.

As a further improved technical solution of the present invention, the step 3 specifically includes:

Step 3.1, based on the preprocessed root system image of Masson pine seedlings, set the pixel points with a pixel value of 255 in the root system image to 1, use the improved ZhangSuen skeleton extraction algorithm on the root system image, reduce the image skeleton from a multi-pixel width to a unit pixel width, and keep the basic shape and trend of the skeleton;

Step 3.2. Among them, the improved ZhangSuen algorithm mainly adjusts the calculation area of the root system image to the minimum bounding box range of the masson pine seedling root system. After each iterative calculation, the minimum bounding box range of the seedling root system will be updated; adjust the image calculation Judgment condition priority. The judgment condition for image calculation is that if the currently traversed point is a boundary point, delete the point from the root system point set and set the condition as a priority judgment; when there are no points that can be deleted further, the algorithm ends and generates Compute the skeleton of the region.

As a further improved technical solution of the present invention, the step 4 specifically includes:

Step 4.1, based on the root system skeleton image of Masson pine seedlings obtained in step 3, all pixel points with a pixel value of 255 in the image are set to 1, and the pixel value of the point on the root system skeleton line in the image is 1, and the pixel value of the non-skeleton point is 0;

Step 4.2, construct the judgment condition of the root endpoint of Masson pine seedlings, traverse each point with non-zero pixel value on the root skeleton map, and mark and search its 3×3 neighborhood range with the current point as the center, mark the pixel points in the neighborhood that meet the judgment condition as the root endpoint, and append them to the endpoint set End_Points; the point with non-zero pixel value is referred to as non-zero pixel point;

The conditions for judging the end points of the root system of masson pine seedlings are as follows:

Judgment condition (1): There is one and only one non-zero pixel in the 3×3 neighborhood of the current point;

Judgment condition (2): There are two adjacent non-zero pixels in the 3×3 neighborhood of the current point;

If the judgment condition (1) or the judgment condition (2) is met, the current point is judged to be the endpoint.

As a further improved technical solution of the present invention, the step 5 specifically includes:

Step 5.1. Based on the refined skeleton diagram of the masson pine seedling root system obtained in step 3, set all pixels with a pixel value of 255 in the image to 1;

Step 5.2: Construct the judgment conditions for the root system turning point of masson pine seedlings, traverse all non-zero pixels in the root system skeleton diagram, and perform a mark search on the 3×3 neighborhood range of the current point, and select the pixels in the neighborhood that meet the judgment conditions Mark it as a root turning point and append it to the turning point collection Turning_Points;

The conditions for judging the root system turning point of masson pine seedlings are as follows:

Judgment condition (1): In the 3×3 neighborhood of the current point, at least three pixels among P2, P4, P6, and P8 are non-zero pixels;

Judgment condition (2): Within the 3×3 neighborhood of the current point, at least three pixels among P3, P5, P7, and P9 are non-zero pixels;

Judgment condition (3): There are three non-zero pixels in the 3×3 neighborhood of the current point, and two of them are not adjacent;

Judgment condition (4): There are four non-zero pixel points in the 3×3 neighborhood of the current point, of which at most two points are adjacent;

If the judgment condition (1), judgment condition (2), judgment condition (3) or judgment condition (4) is met, the current point is determined to be a turning point;

Among them, P2-P9 represents the 3×3 neighborhood of the current point, P2 is located directly above the current point, P3 is located at the upper right of the current point, P4 is located to the right of the current point, P5 is located to the lower right of the current point, P6 is located directly below the current point, P7 is located at the lower left of the current point, P8 is located to the left of the current point, and P9 is located to the upper left of the current point.

As a further improved technical solution of the present invention, the step 6 specifically includes:

Step 6.1. Set the pixel value of 255 in the masson pine seedling root system skeleton map to 1. Based on the root system endpoints and transition points located in steps 4 and 5, first set the top endpoint A in the root system skeleton map. That is, the first endpoint A in the endpoint set End_Points is used as the current starting point, the pixel value of the current point is set to 0, and the current point is deleted from the endpoint set End_Points;

Step 6.2, determine whether there is a non-zero pixel in the 3×3 neighborhood of the current point. If there is a non-zero pixel in P2, P4, P6, and P8, the non-zero pixel with the smallest index value in the 3×3 neighborhood is preferentially selected as the next judgment traversal point. If there is no non-zero pixel in P2, P4, P6, and P8, the non-zero pixel with the smallest index value in the 3×3 neighborhood of P3, P5, P7, and P9 is selected as the next judgment traversal point;

Step 6.3: If the current traversal point is neither an endpoint nor a turning point, set the pixel value of the point to 0 and jump back to step 6.2 to continue the loop;

Step 6.4. If the current traversed point is a turning point, set the pixel value of the point to 0, append it to the set of turning points Pathway_Turning_Points, and jump back to step 6.2 to continue executing the loop;

Step 6.5:

6.5.1. If the current traversal point is an endpoint, set the pixel value of the endpoint to 0, delete it from the endpoint set End_Points, and save the root path from the starting point to the endpoint;

6.5.2. Then, jump the current traversal point back to the last turning point in the passing turning point set Pathway_Turning_Points;

6.5.3. Determine whether there are non-zero pixels in the 3×3 neighborhood of the jumping point;

6.5.4. If there are non-zero pixels in the 3×3 neighborhood of the jumped pivot point, jump back to step 6.2 and continue executing the loop;

6.5.5. If there are no non-zero pixels in the 3×3 neighborhood of the jumped-back turning point, the jumped-back turning point will be deleted from Pathway_Turning_Points, and the current traversal point will be jumped back to the last point in the updated Pathway_Turning_Points set. A turning point, returning to the 6.5.3 execution loop;

Step 6.6, when the endpoint set End_Points is an empty set, the algorithm ends;

Step 6.7: Calculate the Euclidean distance d of all saved root paths. The root path corresponding to the longest Euclidean distance is the main root of the masson pine seedling root system. The Euclidean distance of the main root is recorded as d _M ;

Among them, n is the total number of points on the root system path; x _i and y _i are the coordinates of the points;

Step 6.8. The turning point on the main root of the root system is the first-level turning point, and the longest root extending out from the first-level turning point is the first-level lateral root; therefore, delete the point on the main root from the original skeleton diagram, and then Taking each first-level turning point on the main root as the current starting point, repeat steps 6.2 to 6.6, and then save the root path information below the first-level lateral roots;

Step 6.9, respectively calculate the Euclidean distance d of the root path starting from each primary turning point, wherein the root path corresponding to the longest Euclidean distance starting from each primary turning point is the primary lateral root of the root system of the Masson pine seedling, and the Euclidean distance of the primary lateral root is recorded as d _L .

As a further improved technical solution of the present invention, the step 7 specifically includes:

Step 7.1. Use the calibration block to calculate the scaling coefficient μ of the root system. Use the scaling coefficient to correct the main root length and primary lateral root length calculated in steps 6.7 and 6.9 to obtain their respective theoretical lengths;

d _MR = d _M × μ (4);

d _LR = d _L ×μ (5);

Among them, d _MR is the theoretical length of the main root, d _LR is the theoretical length of the primary lateral root;

Step 7.2. Calculate the number of endpoints after excluding the main root endpoint, that is, the number of lateral roots of the masson pine seedling root system;

Num _{Lateral_Roots} =len(End_Points)-2 (6);

Among them, Num _{Lateral_Roots} is the number of lateral roots of the seedling root system; len() represents the total number of points in the set, and End_Points represents the endpoint set established in step 4.

In order to achieve the above technical objectives, another technical solution adopted by the present invention is:

An image-based extraction system for root phenotypic parameters of masson pine seedlings, including:

The image acquisition module is used to collect root system images of masson pine seedlings and correct the root system images;

Image preprocessing module, used to preprocess the corrected root system image;

The root system skeleton refinement module is used to refine the root system image and realize the root system skeleton extraction of masson pine seedlings;

The root endpoint positioning module is used to search for the endpoints of the pixels in the extracted root skeleton map, determine the pixels that meet the conditions as the endpoints of the root system, and realize the root endpoint positioning of the Masson pine seedlings;

The root system turning point positioning module is used to search for turning points in the pixels in the extracted root system skeleton diagram, and determine the pixels that meet the conditions as the turning points of the root system to realize the turning point positioning of the root system of masson pine seedlings;

The root system segmentation module is used to segment the main root and first-level lateral roots of the root system based on the located end points and turning points of the root system of masson pine seedlings, using a priority path search algorithm with preferential directionality;

The root phenotypic parameter extraction module is used to extract the root phenotypic parameters of Masson pine seedlings. The root phenotypic parameters include the theoretical length of the main root, the theoretical length of the first-level lateral roots, and the number of lateral roots.

The beneficial effects of the present invention are:

(1) The present invention optimizes the running speed of the ZhangSuen skeleton extraction algorithm and adjusts the calculation area of the root system image to the minimum bounding box range of the masson pine seedling root system. After each iterative calculation, the minimum bounding box range of the seedling root system will be updated. . At the same time, the priority of the judgment conditions for image calculation is adjusted, and the condition (if the current calculation point is a boundary point, mark it for deletion) is set as the priority judgment to reduce unnecessary calculations.

(2) The present invention constructs the judgment conditions for the end points and turning points of the root system skeleton of masson pine seedlings, searches for the end points and turning points of the pixels in the root system skeleton diagram, and determines the pixel points that meet the conditions as the end points or turning points of the root system. Realizing the positioning of the end points and turning points of the root system of masson pine seedlings provides the basis for the separation of the main root and lateral roots of the seedling root system;

(3) The preferred path search algorithm (PPSA) of the present invention with preferential directionality can accurately find the main root and first-level lateral roots while ensuring the correct trend and branches of the root system skeleton, and is superior in identifying multi-level lateral roots. , there is no recognition error, and the continuity and integrity of the path are guaranteed. Therefore, the PPSA algorithm shows excellent accuracy and reliability in root skeleton segmentation and multi-level lateral root identification.

(4) The present invention aims to solve the problems of large errors and low efficiency in traditional manual root measurement, and the inability of existing laboratory three-dimensional reconstruction equipment to detect finer roots to meet the evaluation requirements. The present invention designs a low-cost and high-efficiency image-based method for detecting root phenotypic information of Masson pine seedlings. The method saves measurement costs and improves the measurement accuracy of root phenotypic parameters of Masson pine seedlings. At the same time, the method and process are also applicable to other forest seedlings, providing an effective and accurate method for quickly and accurately measuring seedling root phenotypic parameters.

(5) The present invention can quickly and accurately measure the root phenotypic parameters of masson pine seedlings, avoiding the high cost and high error of manual measurement, and does not need to rely on expensive detection equipment; the root system phenotypic parameters measured by the present invention Type parameters can meet the needs of forestry seedling quality evaluation.

BRIEF DESCRIPTION OF THE DRAWINGS

Figure 1 is a flow chart of the image-based extraction method of root phenotypic parameters of masson pine seedlings.

Figure 2 is a numbering map of points in the 3×3 neighborhood of a pixel.

Figure 3 is an explanation diagram of the judgment conditions for the end points of the root system skeleton of masson pine seedlings.

(a) in Figure 3 is an explanation diagram of the judgment condition (1) for the end points of the root system skeleton of masson pine seedlings.

(b) in Figure 3 is an explanation diagram of the judgment condition (2) for the end points of the root system skeleton of masson pine seedlings.

Figure 4 is an explanation diagram of the judgment conditions for the turning point of the root system skeleton of masson pine seedlings.

Figure 4 (a) is an explanatory diagram of the judgment conditions (1) for the turning point of the root skeleton of Masson pine seedlings.

Figure 4 (b) is an explanatory diagram of the judgment conditions (2) for the turning point of the root skeleton of Masson pine seedlings.

(c) in Figure 4 is an explanation diagram of the judgment condition (3) for the turning point of the root system skeleton of masson pine seedlings.

(d) in Figure 4 is an explanation diagram of the judgment condition (4) for the turning point of the root system skeleton of masson pine seedlings.

Figure 5 is the algorithm flow chart of the priority path search algorithm (PPSA).

Figure 6 is a rendering of using the improved ZhangSuen algorithm to extract the root system skeleton of masson pine seedlings.

Figure 7 is a rendering of the positioning of the end points and turning points of the root system skeleton of masson pine seedlings.

Figure 7(a) shows the positioning effect of the end points of the root system skeleton of masson pine seedlings.

(b) in Figure 7 is the positioning effect of the turning point of the root system skeleton of masson pine seedlings.

Figure 8 is a rendering of the priority path search algorithm (PPSA) for segmenting the root system of masson pine seedlings.

Detailed ways

The specific embodiments of the present invention will be further described below based on the accompanying drawings:

An image-based method for extracting root phenotypic parameters of masson pine seedlings. The process is shown in Figure 1, including the following steps:

Step 1: Collect masson pine seedling root system images and correct the root system images;

Step 2: preprocess the corrected Masson pine root system image;

Step 3: Use the improved ZhangSuen skeleton extraction algorithm on the processed root image to refine the image to extract the root skeleton of the masson pine seedlings;

Step 4: Based on the root system skeleton diagram obtained in step 3, construct the root system skeleton endpoint judgment conditions, conduct an endpoint search on the pixels in the root system skeleton diagram, and determine the pixels that meet the conditions as the endpoints of the root system to realize the root system endpoints of masson pine seedlings. position;

Step 5: Based on the root system skeleton diagram obtained in step 3, construct the root system skeleton turning point judgment conditions, conduct a turning point search on the pixels in the root system skeleton diagram, and determine the pixels that meet the conditions as the turning points of the root system to realize the masson pine Rotation point positioning of seedling root system;

Step 6: Based on the end points and turning points of the masson pine seedling root system skeleton located in steps 4 and 5, use a Priority Path Search Algorithm (hereinafter referred to as PPSA) with priority directionality to segment the masson pine seedling root system. Taproot and primary lateral roots;

Step 7: Based on the root skeleton segmented by the PPSA algorithm, the root phenotypic parameters of Masson pine seedlings are extracted.

The step 1 specifically includes:

Step 1.1. Dig out the masson pine seedlings from the soil, and use water to clean the soil impurities on the roots; lay the cleaned masson pine roots flat on a white background plate that contrasts with the black roots to facilitate the follow-up. Perform image processing on root system images;

Step 1.2. Place a 3×4 checkerboard calibration board at a fixed position on the white background board. The checkerboard size is 20×20mm. It is used to calibrate the camera and obtain the internal and external parameters of the camera. After collecting images of the root system of Masson pine seedlings, the root system image was corrected based on the acquired internal and external parameters of the camera.

The step 2 specifically includes:

Step 2.1. The corrected root system image needs to undergo a weighted average method grayscale transformation operation to convert the three channels into a single channel to reduce the amount of calculation during image processing. The weighted average method formula is as follows:

Among them, g(x,y) is the pixel value of the output image, R, G, and B are the pixel values in the three channels of the color image respectively, w _R , w _G and w _B are the differences given to the three components of RGB respectively. weight value. In order to achieve accurate segmentation, assign values to w _R , w _G and w _B respectively, w _R = 0.299, w _G = 0.587, w _B = 0.114;

Step 2.2. In view of the obvious difference between the brightness of the target root system and the background in the grayscale image, perform a global threshold method binarization operation on the grayscale image. The global threshold value range is (125,255). The formula of the global threshold method is as follows:

Where T is the selected threshold size, b(x,y) is the output image pixel;

Step 2.3: Use median filtering to remove the noise that is not eliminated or enhanced after the image is binarized. The filter kernel size of the median filtering algorithm is 5×5.

Step 2.4. After median filtering, the masson pine seedling root system image will produce some gaps and some small holes. Use the closing operation in the morphological operation to fill and eliminate the gaps and small holes on the masson pine seedling root system skeleton diagram, and To maintain the overall integrity of the root system skeleton, the closed computing kernel size is 5×5.

The step 3 specifically includes:

Step 3.1. Based on the preprocessed masson pine seedling root system image, set the pixel value of 255 in the root system image to 1, use the improved ZhangSuen skeleton extraction algorithm on the root system image, and reduce the image skeleton from multi-pixel width to unit pixel width, and maintain the basic shape and trend of the skeleton;

Step 3.2. This embodiment adopts the improved ZhangSuen skeleton extraction algorithm. The improvement point of the improved ZhangSuen skeleton extraction algorithm is mainly to adjust the calculation area of the root system image to the minimum bounding box range of the root system of the masson pine seedlings. After each iteration calculation, The minimum bounding box range of the seedling root system will be updated once. At the same time, the priority of the judgment conditions for image calculation is adjusted, and the condition (if the current calculation point is a boundary point, mark it for deletion, that is, delete the calculation point from the entire root system point set) is set as a priority judgment to reduce unnecessary calculations. When there are no more points that can be deleted, the algorithm ends and the skeleton of the calculation area is generated. The rendering of using the improved ZhangSuen algorithm to extract the root system skeleton of masson pine seedlings is shown in Figure 6.

The step 4 specifically includes:

Step 4.1, based on the root system skeleton image of Masson pine seedlings obtained in step 3, all pixel points with a pixel value of 255 in the image are set to 1; in the image, the pixel value of the point on the root system skeleton line is 1 (skeleton point), and the pixel value of the non-skeleton point is 0 (background point);

Step 4.2, traverse each non-zero pixel value on the root system skeleton map, and mark and search the 3×3 neighborhood range with it as the center, mark the pixel points in the neighborhood that meet the judgment conditions as root system endpoints, and append them to the endpoint set End_Points; as shown in Figure 2, P2-P9 represents the 3×3 neighborhood range centered on P1;

The endpoint is the end point of the root system, so the judgment conditions for constructing the endpoint of the root system of masson pine seedlings are as follows:

Judgment condition (1): There is one and only one non-zero pixel in the 3×3 neighborhood of the current point;

Judgment condition (2): There are two adjacent non-zero pixels in the 3×3 neighborhood of the current point.

Sum(n)＝1 Judgment condition(1);

Among them, n represents the 3×3 neighborhood range of the current point; Sum represents the total number of non-zero pixels in the 3×3 neighborhood of the current point; Index(i) represents the index of the i-th non-zero pixel point in the 3×3 neighborhood. ; Judgment condition (1) and judgment condition (2) satisfy one of them to determine that the current point is the endpoint; in the condition explanation diagram shown in (a) and (b) in Figure 3, the black square represents non- Zero pixel point, the white square represents the pixel value is zero point.

The step 5 specifically includes:

Step 5.1. Based on the refined skeleton diagram of the masson pine seedling root system obtained in step 3, set all pixels with a pixel value of 255 in the image to 1;

Step 5.2: Traverse all non-zero pixels in the root skeleton diagram, and perform a mark search on the 3×3 neighborhood of the current point. Mark the pixels in the neighborhood that meet the judgment conditions as root turning points, and append them to In the turning point collection Turning_Points;

The turning point is the bifurcation point in the root system of Masson pine seedlings, so the judgment conditions for constructing the turning point of the root system of Masson pine seedlings are as follows:

Judgment condition (1): Within the 3×3 neighborhood of the current point, at least three pixels among P2, P4, P6, and P8 are non-zero points;

Judgment condition (2): Within the 3×3 neighborhood of the current point, at least three pixels among P3, P5, P7, and P9 are non-zero points;

Judgment condition (3): There are three non-zero pixels in the 3×3 neighborhood of the current point, and two of them are not adjacent;

Judgment condition (4): There are four non-zero pixel points in the 3×3 neighborhood of the current point, of which at most two points are adjacent;.

Sum(P2,P4,P6,P8)≥3 Judgment condition (1);

Sum(P3,P5,P7,P9)≥3 Judgment condition (2);

Among them, n represents the 3×3 neighborhood range of the current point; Sum represents the total number of non-zero pixels in the 3×3 neighborhood of the current point; Index(i) represents the number of non-zero pixels in the 3×3 neighborhood of the current point. Index; sum(x=m) represents the total number of elements equal to m in the set is the turning point; in the condition explanation diagram of (a)-(d) in Figure 4, the black square represents the non-zero pixel point, and the white square represents the zero pixel value.

As shown in Figure 7, (a) in Figure 7 is a positioning effect diagram of the end point of the root system skeleton of Masson pine seedlings, and (b) in Figure 7 is a positioning effect diagram of the turning point of the root system skeleton of Masson pine seedlings.

The step 6 is specifically shown in Figure 5, including:

Step 6.1. Set the pixel value of 255 in the masson pine seedling root system skeleton diagram to 1; based on the root system endpoints and transition points located in steps 4 and 5, first set the top endpoint A in the root system skeleton diagram as Starting point, set the pixel value of the current point to 0, and delete the current point from the end point set End_Points;

Step 6.2. Determine whether there are non-zero points in the 3×3 neighborhood of the current point. If there are non-zero pixels in P2, P4, P6, and P8, give priority to the non-zero point p _i with the smallest index value in the 3×3 neighborhood ( i=2,4,6,8) as the next judgment point, if P2, P4, P6, and P8 are all zero pixels, then select P3, P5, P7, and P9 with the smallest index value in the 3×3 neighborhood. Zero point p _i (i=3,5,7,9) is used as the next judgment point;

Step 6.3: If the current traversal point is neither an endpoint nor a turning point, set the pixel value of the point to 0 and jump back to step 6.2 to continue the loop;

Step 6.4, if the current traversal point is a turning point, set the pixel value of the point to 0, add it to the turning point set Pathway_Turning_Points, and jump back to step 6.2 to continue the loop;

Step 6.5. If the current traversal point is an endpoint, set the pixel value of the point to 0, delete it from the endpoint set End_Points, and save the root path from the starting point to the endpoint; then jump back to the current traversal point. The last turning point in the turning point set Pathway_Turning_Points; if there are no non-zero points around the jumping turning point, the jumping turning point will be deleted from Pathway_Turning_Points, and then the current traversal point will be jumped back to the last point in the updated Pathway_Turning_Points set A turning point; then jump back to step 6.2 to continue executing the loop;

Step 6.6, when the endpoint set End_Points is an empty set, the algorithm ends, and finally all the seedling root path information is saved;

Step 6.7: Calculate the Euclidean distance d of all saved root paths. The root path with the longest Euclidean distance is the main root of the masson pine seedling root system. The Euclidean distance of the main root is recorded as d _M ;

Among them, n is the total number of points on the root system path; x _i and y _i are the coordinates of the points;

Step 6.8. The turning point on the main root of the root system is recorded as the first-level turning point, and the longest root extending out from it is the first-level lateral root; therefore, the points on the main root are deleted from the original skeleton diagram, and then the points on the main root are Each first-level turning point is used as the starting point, and steps 6.2 to 6.6 are executed repeatedly to save the root path information below all first-level lateral roots;

Step 6.9: Calculate the Euclidean distance d of the root path starting from each first-level turning point respectively. Among them, the root corresponding to the longest Euclidean distance starting from each first-level turning point is the first-level lateral root of the masson pine seedling root system. , the Euclidean distance of the first-level lateral root is recorded as d _L .

Similarly, if you want to save the secondary lateral roots, record the turning point on the primary lateral root of the root system as the secondary turning point, and the longest root extending from it as the starting point is the secondary lateral root; similarly, steps 6.8-6.9 can be used to save the root path information of all secondary lateral roots and calculate the Euclidean distance of the secondary lateral roots.

Figure 8 is a rendering of the root system segmentation of masson pine seedlings through the priority path search algorithm (PPSA) in step 6.

The step 7 specifically includes:

Step 7.1, use the calibration block to calculate the scaling factor μ of the root system, and use the scaling factor to correct the main root and primary lateral root lengths calculated in steps 6.7 and 6.9 to obtain their respective theoretical lengths:

_dMR = _dM ×μ (4);

d _LR = d _L ×μ (5);

Among them, d _MR and d _LR are the theoretical length of the main root and the theoretical length of the primary lateral root respectively;

Step 7.2. Calculate the number of endpoints after excluding the main root endpoint, that is, the number of lateral roots of the masson pine seedling root system:

Num _{Lateral_Roots} =len(End_Points)-2 (6);

Among them, Num _{Lateral_Roots} is the number of lateral roots of the seedling root system; len() represents the total number of points in the set, and End_Points represents the endpoint set established in step 4.

This embodiment also provides an image-based method for extracting root phenotypic parameters of masson pine seedlings, including:

An image acquisition module is used to acquire images of the root system of Masson pine seedlings and correct the images;

Image preprocessing module, used to enhance root system features, remove irrelevant noise, and fill gaps and holes in the root system;

The root system skeleton refinement module is used to refine the root system image and realize the root system skeleton extraction of masson pine seedlings;

The root system endpoint positioning module is used to search for endpoints of pixels in the root system skeleton diagram, and determine the pixels that meet the conditions as the endpoints of the root system to realize the positioning of the root system endpoints of masson pine seedlings;

The root system turning point positioning module is used to search for turning points in the pixels in the root system skeleton diagram, and determine the pixels that meet the conditions as the turning points of the root system to realize the turning point positioning of the root system of masson pine seedlings;

The root system segmentation module is used to use a priority path search algorithm with preferential directionality for the root system of masson pine seedlings to segment the main root and first-level lateral roots of the root system;

The root phenotypic parameter extraction module is used to extract the root phenotypic parameters of Masson pine seedlings. The root phenotypic parameters include main root length, first-level lateral root length and number of lateral roots.

This embodiment optimizes the running speed of the ZhangSuen skeleton extraction algorithm, adjusts the calculation area of the root system image to the minimum bounding box range of the Masson pine seedling root system, and updates the minimum bounding box range of the seedling root system after each iterative calculation. At the same time, the judgment condition priority of the image calculation is adjusted, and the condition (if the current calculation point is a boundary point, then mark it for deletion) is set as the priority judgment to reduce unnecessary calculations.

This embodiment constructs the judgment conditions of the endpoints and turning points of the root system skeleton of Masson pine seedlings, searches for endpoints and turning points for the pixels in the root system skeleton map, and determines the pixels that meet the conditions as the endpoints or turning points of the root system, so as to realize the positioning of the endpoints and turning points of the root system of Masson pine seedlings, and provide a basis for the segmentation of the main root and lateral root of the seedling root system;

A priority path search algorithm (PPSA) with priority directionality in this embodiment can accurately find the main root and first-level lateral roots while ensuring that the trend and branches of the root system skeleton are correct, and in multi-level lateral root identification, there is no In the event of recognition errors, the continuity and integrity of the path are guaranteed. Therefore, the PPSA algorithm shows excellent accuracy and reliability in root skeleton segmentation and multi-level lateral root identification.

In this embodiment, a low-cost, high-efficiency image-based masson pine sapling is designed to address the problems of traditional manual measurement of root systems with large errors and low efficiency, and the existing three-dimensional reconstruction equipment in the laboratory does not meet the evaluation requirements for the detection of finer root systems. Root phenotypic information detection method. This method saves measurement costs and improves the measurement accuracy of root phenotypic parameters of masson pine seedlings. At the same time, this method and process are also applicable to other forest tree seedlings, providing an effective and accurate method for quickly and accurately measuring seedling root phenotypic parameters.

The protection scope of the present invention includes but is not limited to the above embodiments. The protection scope of the present invention shall be based on the claims. Any replacement, deformation, and improvement of the technology that can be easily thought of by technicians in this field shall fall within the protection scope of the present invention.

Claims (9)

1. An image-based method for extracting phenotype parameters of a pinus massoniana seedling root system is characterized by comprising the following steps:

step 1: collecting an image of a root system of a pinus massoniana seedling and correcting the image;

step 2: preprocessing the corrected image;

step 3: carrying out image refinement operation on the processed image by using an improved ZhangSuen skeleton extraction algorithm to realize root skeleton extraction of pinus massoniana seedlings;

step 4: constructing root skeleton end point judging conditions based on the root skeleton diagram obtained in the step 3, searching end points of pixel points in the root skeleton diagram, judging the pixel points meeting the judging conditions as end points of a root system, and realizing the end point positioning of the root system of the pinus massoniana seedlings;

step 5: constructing a root system skeleton turning point judgment condition based on the root system skeleton diagram obtained in the step 3, carrying out turning point search on pixel points in the root system skeleton diagram, judging the pixel points meeting the judgment condition as turning points of a root system, and realizing the positioning of the turning points of the root system of the pinus massoniana seedlings;

step 6: based on the end points and the turning points of the root system skeleton of the pinus massoniana seedlings positioned in the step 4 and the step 5, dividing the main root and the primary lateral root of the root system of the pinus massoniana seedlings by adopting a priority path searching algorithm with priority directivity;

step 7: and extracting root system phenotype parameters of the pinus massoniana seedlings based on the main roots and the primary lateral roots of the segmented pinus massoniana seedling roots.

2. The method for extracting the phenotype parameters of the root system of the pinus massoniana seedlings based on the images according to claim l, wherein the step 1 is specifically as follows:

step 1.1, digging pinus massoniana seedlings from soil, cleaning soil impurities on a pinus massoniana seedling root system by water flow, and spreading the cleaned pinus massoniana seedling root system on a white background plate which is in clear contrast with a black root system;

and 1.2, placing a 3X 4 checkerboard calibration plate at a fixed position on a white background plate, wherein the size of the checkerboard is 20X 20mm, and the checkerboard calibration plate is used for calibrating a camera and acquiring internal and external parameters of the camera, and correcting root system images according to the acquired internal and external parameters of the camera after acquiring images of the root systems of pinus massoniana seedlings.

3. The method for extracting the phenotype parameters of the root system of the pinus massoniana seedlings based on the images according to claim 1, wherein the step 2 specifically comprises the following steps:

step 2.1, performing gray level transformation operation on the corrected root system image by a weighted average method, and converting the three channels into single channels, wherein the weighted average method has the formula as follows:

where G (x, y) is the pixel value of the output image, R is the pixel value of the R channel in the color image, G is the pixel value of the G channel in the color image, B is the pixel value of the B channel in the color image, w _R To give weight to the R component, w _G To give weight to the G component, w _B To give a weight value to the B component;

step 2.2, executing a global thresholding binarization operation on the gray level image, wherein the global threshold range is (125, 255), and the global thresholding formula is as follows:

wherein T is the selected threshold value, and b (x, y) is the pixel point of the output image;

step 2.3, removing noise which is not eliminated or enhanced after image binarization by using median filtering, wherein the size of a filtering kernel of a median filtering algorithm is 5 multiplied by 5;

and 2.4, filling and eliminating gaps and small holes on the root system image by using a closed operation in morphological operation, and keeping the integral integrity of the root system skeleton, wherein the size of a closed operation kernel is 5 multiplied by 5.

4. The method for extracting the phenotype parameters of the root system of the pinus massoniana seedlings based on the images according to claim 1, wherein the step 3 specifically comprises the following steps:

step 3.1, based on the preprocessed masson pine seedling root system image, setting a pixel point with a pixel value of 255 in the root system image as 1, using an improved ZhangSuen skeleton extraction algorithm for the root system image, reducing the image skeleton from a multi-pixel width to a unit pixel width, and keeping the basic shape and trend of the skeleton;

step 3.2, wherein the improved zhangsuin algorithm mainly adjusts the calculation area of the root system image to the minimum bounding box range of the pinus massoniana seedling root system, and the minimum bounding box range of the seedling root system is updated once after each iteration calculation; adjusting the priority of a judgment condition of image calculation, wherein the judgment condition of image calculation is that if the point currently traversed is a boundary point, the point is deleted from the root point set, and the condition is set as priority judgment; when there are no points that can be deleted continuously, the algorithm ends, generating a skeleton of the calculation region.

5. The method for extracting the phenotype parameters of the root system of the pinus massoniana seedlings based on the images according to claim 1, wherein the step 4 specifically comprises the following steps:

step 4.1, setting all pixel points with the pixel value of 255 in the image as 1 based on the pinus massoniana seedling root system thinning skeleton diagram obtained in the step 3, wherein the pixel value of a point on a root system skeleton line in the image is 1, and the pixel value of a non-skeleton point is 0;

step 4.2, constructing judging conditions of root system endpoints of pinus massoniana seedlings, traversing each pixel value non-zero point on a root system skeleton diagram, carrying out marker search on a 3X 3 neighborhood range of the pinus massoniana seedling by taking a current point as a center, marking the pixel Points meeting the judging conditions in the neighborhood as the root system endpoints, and adding the pixel Points into an endpoint set end_points; wherein the non-zero point of the pixel value is called as non-zero pixel point for short;

judging conditions of the root system end points of the pinus massoniana seedlings are as follows:

judgment condition (1): within the 3 x 3 neighborhood of the current point, there is and only one non-zero pixel point;

judgment condition (2): within the 3×3 neighborhood of the current point, there are two adjacent non-zero pixel points;

if the judgment condition (1) or the judgment condition (2) is satisfied, the current point is judged as an endpoint.

6. The method for extracting the phenotype parameters of the root system of the pinus massoniana seedlings based on the images according to claim 5, wherein the step 5 specifically comprises the following steps:

step 5.1, setting all pixel points with the pixel value of 255 in the image as 1 based on the pinus massoniana seedling root system thinning skeleton diagram obtained in the step 3;

step 5.2, constructing judging conditions of root system Turning Points of pinus massoniana seedlings, traversing all non-zero pixel Points in a root system skeleton diagram, performing marker search on a 3X 3 neighborhood range of a current point, marking the pixel Points in the neighborhood which meet the judging conditions as the root system Turning Points, and adding the pixel Points into a turning_points set;

the judging conditions of the root system turning points of the pinus massoniana seedlings are as follows:

judgment condition (1): in the 3X 3 neighborhood of the current point, at least three pixel points in P2, P4, P6 and P8 are non-zero pixel points;

judgment condition (2): in the 3X 3 neighborhood of the current point, at least three pixel points in P3, P5, P7 and P9 are non-zero pixel points;

judgment condition (3): three non-zero pixel points are arranged in the 3X 3 neighborhood of the current point, and are not adjacent to each other;

judgment condition (4): within the 3 x 3 neighborhood of the current point, there are four non-zero pixel points, of which at most two are adjacent;

if the judging condition (1), the judging condition (2), the judging condition (3) or the judging condition (4) is met, judging that the current point is a turning point;

wherein, P2-P9 represent 3X 3 neighborhood of the current point, P2 is located right above the current point, P3 is located right above the current point, P4 is located right side of the current point, P5 is located right below the current point, P6 is located right below the current point, P7 is located left below the current point, P8 is located left side of the current point, and P9 is located left above the current point.

7. The method for extracting the phenotype parameters of the root system of the pinus massoniana seedlings based on the images according to claim 1, wherein the step 6 specifically comprises the following steps:

step 6.1, setting a pixel point with a pixel value of 255 in a masson pine seedling root system skeleton diagram as 1, firstly setting an endpoint A at the top of the root system skeleton diagram, namely a first endpoint A in an endpoint set end_points as a current starting point based on the root system endpoints and the turning Points positioned in the step 4 and the step 5, setting the pixel value of the current point as 0, and deleting the current point from the endpoint set end_points;

step 6.2, judging whether non-zero pixel points exist in a 3×3 neighborhood of the current point, if the non-zero pixel points exist in P2, P4, P6 and P8, preferentially selecting the non-zero pixel point with the minimum index value in the 3×3 neighborhood as a next judging traversal point, and if the non-zero pixel points do not exist in P2, P4, P6 and P8, selecting the non-zero pixel points with the minimum index values in the 3×3 neighborhood of P3, P5, P7 and P9 as the next judging traversal points;

step 6.3, if the current traversing point is neither an endpoint nor a turning point, setting the pixel value of the point to 0, and jumping back to the step 6.2 to continue to execute the loop;

step 6.4, if the current traversing point is a Turning point, setting the pixel value of the point to 0, adding the pixel value to a passing Turning point set path_turning_points, and jumping back to the step 6.2 to continue to execute the loop;

step 6.5:

6.5.1, if the current traversing point is an endpoint, setting the endpoint pixel value to 0, deleting the endpoint pixel value from the endpoint set end_points, and storing a root system path from the starting point to the endpoint;

6.5.2 then, jumping the current traversal point back to the last point in the set of Points through path_turning_points;

6.5.3, judging whether non-zero pixel points exist in the 3 multiplied by 3 neighborhood of the jumping-back point;

6.5.4, if the non-zero pixel points exist in the 3 multiplied by 3 neighborhood of the jumping-back point, continuing to execute the loop in the jumping-back step 6.2;

6.5.5 if there is no non-zero pixel point in the 3×3 neighborhood of the skipped point, deleting the skipped point from path_turning_points, and skipping the current traversal point back to the last point in the updated path_turning_points set, returning to 6.5.3 for executing the loop;

step 6.6, ending the algorithm when the endpoint set end_points is an empty set;

step 6.7, calculating the Euclidean distance d of all the stored root system paths, wherein the root system path corresponding to the longest Euclidean distance is the main root of the root system of the pinus massoniana seedlings, and the Euclidean distance of the main root is recorded as d _M ；

Where n is the total number of points on the root path; x is x _i ，y _i Coordinates of points;

step 6.8, the turning point on the root system main root is a first-stage turning point, and the longest root stretched out by taking the first-stage turning point as a starting point is a first-stage lateral root; therefore, deleting the points on the main root from the original skeleton diagram, and repeatedly and circularly executing the steps 6.2 to 6.6 by taking each primary turning point on the main root as a current starting point, so as to further store root system path information below all primary lateral roots;

step 6.9, respectively calculating Euclidean distance d of root system paths starting from each primary turning point, wherein the root system path corresponding to the longest Euclidean distance starting from each primary turning point is the primary lateral root of the pinus massoniana seedling root system, and the Euclidean distance of the primary lateral root is recorded as d _L 。

8. The method for extracting the phenotype parameters of the root system of the pinus massoniana seedlings based on the images according to claim 7, wherein the step 7 specifically comprises the following steps:

step 7.1, calculating a scaling factor mu of a root system by using a calibration block, and correcting the main root length and the primary side root length calculated in the steps 6.7 and 6.9 by using the scaling factor to obtain respective theoretical lengths;

d _MR ＝d _M ×μ (4)；

d _LR ＝d _L ×μ (5)；

wherein d _MR Theoretical length of principal root, d _LR Is the theoretical length of the primary lateral root;

step 7.2, calculating the number of endpoints after the endpoints of the main root are removed, namely the number of lateral roots of the root system of the pinus massoniana seedlings;

Num _{Lateral_Roots} ＝len(End_Points)-2 (6)；

wherein Num is _{Lateral_Roots} The number of lateral roots of the root system of the seedling; len () represents the total number of Points in the set and End Points represents the set of endpoints established in step 4.

9. An image-based pinus massoniana seedling root phenotype parameter extraction system is characterized by comprising:

the image acquisition module is used for acquiring root system images of the pinus massoniana seedlings and correcting the root system images;

the image preprocessing module is used for preprocessing the corrected root system image;

the root system skeleton refining module is used for refining the root system image to realize extraction of the root system skeleton of the pinus massoniana seedlings;

the root system end point positioning module is used for searching the end points of the pixel points in the extracted root system skeleton diagram, judging the pixel points meeting the conditions as the end points of the root system, and realizing the end point positioning of the root system of the pinus massoniana seedlings;

the root system turning point positioning module is used for carrying out turning point search on pixel points in the extracted root system skeleton diagram, judging the pixel points meeting the conditions as turning points of root systems, and realizing the turning point positioning of the root systems of the pinus massoniana seedlings;

the root system segmentation module is used for segmenting main roots and primary lateral roots of the root system by adopting a priority path search algorithm with priority directivity according to the positioned end points and turning points of the root system of the pinus massoniana seedlings;

the root system phenotype parameter extraction module is used for extracting root system phenotype parameters of the pinus massoniana seedlings, wherein the root system phenotype parameters comprise the theoretical length of main roots, the theoretical length of primary lateral roots and the number of lateral roots.