Moso bamboo joint detection and internode length extraction method based on three-dimensional point cloud

Abstract

The invention discloses a moso bamboo internode length extraction method based on three-dimensional point cloud. The moso bamboo internode length extraction method comprises the steps that 1, bamboo forest three-dimensional point cloud data are collected and preprocessed; 2, bamboo stalk point clouds BS are classified through a reflection intensity threshold value It, and then a single-plant bamboo stalk instance V is obtained through a density clustering algorithm; 3, calculating the main shaft direction D of the bamboo stalks; 4, generating a cross section slice set L at a fixed interval alpha along the main axis direction D, and detecting the bamboo joint area according to the combined feature entropy of the curvature change of the cross section slices and the reflection intensity variance; and 5, calculating the coordinates of the center points of the bamboo joints by using a robust weighted potential function minimization-based bamboo joint center point positioning method, performing third-order B-spline curve fitting on point cloud data between the center points of the adjacent bamboo joints, and calculating the inter-node length of the moso bamboo. The bamboo joints of the moso bamboo can be efficiently and accurately detected, and the internode length is extracted, so that the distribution of the bamboo joints in the growth process of the moso bamboo is accurately mastered, and a good data basis is provided for the evaluation of the growth condition and health level of the moso bamboo and the evaluation of the structural quality.

Description

Technical Field

[0001] The present invention relates to the technical field of intelligent forestry, and more particularly, to a method for detecting bamboo joints and extracting internode lengths of moso bamboo based on three-dimensional point clouds. Background Art

[0002] In the fields of bamboo forest management and bamboo resource utilization, the detection of moso bamboo joints and the accurate calculation of internode lengths are of crucial significance. Accurately grasping the distribution of bamboo joints and their internode lengths during the growth process of moso bamboo can serve as an important indicator for evaluating the growth status, health level, and structural quality of moso bamboo. At the same time, it has practical application value for formulating scientific bamboo forest harvesting, pruning, irrigation, and fertilization plans, as well as guiding bamboo product grading and process optimization. In addition, moso bamboo plays functions such as carbon sequestration and soil and water conservation in the ecosystem, and accurate growth data helps to achieve the sustainable management of bamboo resources.

[0003] Currently, the methods for bamboo joint detection and internode length measurement mainly rely on manual measurement and two-dimensional image processing technology. However, manual measurement not only has low efficiency and strong subjectivity, but also it is difficult to ensure the consistency and accuracy of data in large-scale bamboo forest management. And the two-dimensional image method is limited by planar information, and its detection results are often affected by lighting, occlusion, and image resolution, making it difficult to truly reflect the spatial structure and three-dimensional bending shape of the bamboo stalk, resulting in errors in the calculation of bamboo joint positions and internode lengths.

[0004] The method for extracting bamboo forest phenotypic parameters through three-dimensional point cloud data has been widely used, but there are still many challenges. First, the amount of bamboo forest point cloud data is large, and the processing and calculation complexity are relatively high. Second, the size of moso bamboo joints is relatively small, resulting in great difficulty in bamboo joint detection and precise positioning. In addition, the bamboo stalk part usually has a certain degree of bending, which further increases the difficulty of accurately extracting the internode length of moso bamboo. Summary of the Invention

[0005] The present invention is proposed to solve the above-mentioned deficiencies existing in the prior art, and provides a method for detecting moso bamboo joints and extracting internode lengths based on three-dimensional point clouds, aiming to efficiently and accurately detect moso bamboo joints and extract internode lengths, so as to accurately grasp the distribution of bamboo joints during the growth process of moso bamboo, and provide a good data basis for the evaluation of the growth status, health level, and structural quality of moso bamboo.

[0006] To achieve the above object of the present invention, the following technical solutions are adopted:

[0007] A method for extracting the internode length of moso bamboo based on three-dimensional point clouds according to the present invention is characterized by including the following steps:

[0008] S1. Collect three-dimensional point cloud data of the bamboo forest and perform preprocessing to obtain the preprocessed three-dimensional point cloud data of the bamboo forest;

[0009] S2. Set the reflection intensity threshold to , and classify the points in the pre - processed 3D point cloud data of the bamboo forest with a reflection intensity higher than as the bamboo stalk point cloud data . Then, use the density clustering algorithm to separate the bamboo stalk point cloud data to obtain a single - plant bamboo stalk instance V;

[0010] S3. Use the PCA (Principal Component Analysis) method to process the single - plant bamboo stalk instance V to obtain the main axis direction D of the bamboo stalk;

[0011] S4. Along the main axis direction at a fixed interval generate a set L of cross - sectional slices of the single - plant bamboo stalk instance V, and perform bamboo node detection on each cross - sectional slice according to the joint feature entropy of the cross - sectional slice curvature change and the reflection intensity variance to obtain all bamboo node regions;

[0012] S5. Use the bamboo node center point positioning method based on the minimization of the robust weighted potential function to process the detected bamboo node regions to obtain the bamboo node center point coordinates of each bamboo node region. After sorting all the bamboo node center point coordinates in ascending order, perform a third - order B - spline curve fitting on the point cloud data along the main axis direction between adjacent bamboo node center points, thereby calculating the Phyllostachys pubescens internode length.

[0013] The feature of the Phyllostachys pubescens internode length extraction method based on 3D point cloud of the present invention also lies in that S3 includes the following steps:

[0014] S3.1. Calculate the centralized mean of the single - plant bamboo stalk instance using Equation (1):

[0015] (1)

[0016] In Equation (1), represents the number of points in the single - plant bamboo stalk instance , is the coordinate of the i - th point in the single - plant bamboo stalk instance ;

[0017] S3.2. Obtain the covariance matrix of the single - plant bamboo stalk instance using Equation (2):

[0018] (2)

[0019] In Equation (2), T represents the transpose;

[0020] S3.3. For the covariance matrix Perform eigen decomposition to obtain a number of eigenvalues and a number of eigenvectors, and use the first eigenvector D as the principal axis direction of the single bamboo culm instance V.

[0021] Furthermore, S4 includes the following steps:

[0022] S4.1. Generate a set of cross-sectional slices of the single bamboo culm instance along the principal axis direction D at a fixed interval ;

[0023] S4.2. Calculate the curvature and the intensity variance of the j-th cross-sectional slice of the single bamboo culm instance V respectively using equations (3) and (4) ;

[0024] (3)

[0025] (4)

[0026] In equation (3), is the number of point clouds in the j-th cross-sectional slice , is the normal vector of the -th point in , is the gradient value of ;

[0027] In equation (4), represents the reflection intensity value of the -th point in the j-th cross-sectional slice , represents the average value of the reflection intensities of all points within the j-th cross-sectional slice ;

[0028] S4.3. Calculate the joint feature entropy of the j-th cross-sectional slice using equation (5):

[0029] (5)

[0030] In equation (5), M represents the total number of slices in the set of cross-sectional slices ;

[0031] When the joint feature entropy of the j-th cross-sectional slice is greater than the entropy threshold, it indicates that the j-th cross-sectional slice is a bamboo node region. ​

[0032] Further, S5 includes the following steps:

[0033] S5.1. If the j-th cross-sectional slice is a bamboo node region, calculate the curvature of the j-th point in the bamboo node region using Equation (6) in ; ;

[0034] S5.2. Construct a robust weighted potential function using Equation (7) :

[0035] (7)

[0036] In Equation (7), represents the three-dimensional coordinates of the j-th point in the bamboo node region in ; represents the Huber loss function represents the three-dimensional coordinates of the preset initial bamboo node center point in the bamboo node region ; represents the reflection intensity scale parameter represents the curvature scale parameter;

[0037] S5.3. Define and initialize the iteration step size as = 1, and initialize the three-dimensional coordinates of the bamboo node center point at the (j - 1)-th iteration in the bamboo node region in as ;

[0038] S5.4. Perform iterative update of gradient descent on using Equation (8) to obtain the bamboo node center point at the j-th iteration in the bamboo node region : :

[0039] (8)

[0040] In Equation (8), represents the derivative of the Huber loss function;

[0041] S5.5. When is less than the termination threshold, it means that is the bamboo node center point of the bamboo node region , denoted as ; otherwise, assign + 1 to and then return to S5.4 to execute sequentially;

[0042] S5.6. Obtain a single bamboo culm instance according to the process of S5.1 - S5.5 The central point coordinates of each bamboo joint area on are arranged in ascending order according to the main axis direction D, and then the central axes of two adjacent bamboo joints are calculated ;

[0043] S5.7. Calculate the distances from all points between and to the central axis respectively. The points with distances less than the threshold are used as fitting points, and then the third - order B - spline curve fitting method is used to fit all fitting points to obtain the actual growth path curve of the single bamboo culm instance V;

[0044] S5.8. By integrating the arc length of the growth path curve, the length of the Phyllostachys pubescens internodes between and is obtained.

[0045] An electronic device according to the present invention includes a memory and a processor, characterized in that the memory is used to store a program for supporting the processor to execute the method for detecting bamboo joints and extracting internode lengths of Phyllostachys pubescens, and the processor is configured to execute the program stored in the memory.

[0046] A computer - readable storage medium according to the present invention, characterized in that a computer program stored on the computer - readable storage medium executes the steps of the method for detecting bamboo joints and extracting internode lengths of Phyllostachys pubescens when run by a processor.

[0047] Compared with the prior art, the beneficial effects of the present invention are as follows:

[0048] 1. By collecting and processing high - precision three - dimensional point cloud data of bamboo forests, the present invention overcomes the disadvantages of insufficient three - dimensional information and susceptibility to environmental factors in the two - dimensional image method, and is more efficient than the traditional manual method at the same time.

[0049] 2. The present invention detects bamboo joints according to the joint feature entropy of slice curvature change and reflection intensity variance. Compared with the planar limitation of two - dimensional image detection and the detection method with a single threshold, it has strong anti - interference ability, and improves the accuracy and robustness of bamboo joint detection.

[0050] 3. The present invention uses the bamboo joint central point method based on the minimization of robust weighted potential function to calculate the bamboo joint central point. Compared with the direct calculation or simple geometric fitting method, it can more effectively reduce the interference of point cloud noise and outliers, and improve the accuracy of bamboo joint central point calculation.

[0051] ​4. The present invention uses a third-order B-spline curve to fit the actual growth path of bamboo internodes and calculates the internode length. Compared with directly calculating the internode length through coordinates, it takes into account the natural growth state of the bamboo stalk and improves the accuracy of internode length extraction. BRIEF DESCRIPTION OF THE DRAWINGS

[0052] Figure 1 It is a flowchart of the method for detecting bamboo nodes and extracting internode lengths of Moso bamboo based on 3D point cloud in the embodiment of the present invention;

[0053] Figure 2 It is an example diagram of separating a single bamboo stalk by fusing the reflection intensity threshold and the density clustering algorithm in the embodiment of the present invention. Among them, part (a) is a schematic diagram of preliminary separation by the reflection intensity threshold, and part (b) is a schematic diagram of separating a single bamboo stalk by the density clustering algorithm on the basis of preliminary separation;

[0054] Figure 3 It is a schematic diagram of calculating the main axis direction of the bamboo stalk by PCA and generating cross-sectional slices at fixed intervals in the embodiment of the present invention;

[0055] Figure 4 It is a schematic diagram of detecting the bamboo node position by the joint feature entropy of the slice curvature change and the reflection intensity variance in the embodiment of the present invention;

[0056] Figure 5 It is a schematic diagram of calculating the bamboo node center point and extracting the internode length in the embodiment of the present invention. Among them, part (a) is a schematic diagram of sorting the coordinate points in ascending order after calculating the bamboo node center point by the bamboo node center point method based on the minimization of the robust weighted potential function, and part (b) is a schematic diagram of using a third-order B-spline curve to fit the adjacent internode point cloud data. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0057] To make the above objects, features and advantages of the present invention more obvious and understandable, the following detailed description of the specific embodiments of the present invention is provided with reference to the accompanying drawings.

[0058] In this embodiment, a method for detecting bamboo nodes and extracting internode lengths of Moso bamboo based on 3D point cloud, as Figure 1 shown, includes the following steps:

[0059] S1. Collect 3D point cloud data of the bamboo forest and perform preprocessing to obtain the preprocessed 3D point cloud data of the bamboo forest;

[0060] In specific implementation, the format of each acquisition point is the three-dimensional spatial coordinates of X, Y, and Z plus the reflection intensity Intensity of the point. Since the data acquisition in the bamboo forest is vulnerable to environmental factors, it is necessary to normalize the reflection intensity to [0, 1] to reduce the error in subsequent analysis. Then, it is necessary to perform statistical filtering on the data to remove abnormal data caused by sensor errors or environmental factors. Finally, it is necessary to reduce the point cloud data through voxel grid downsampling to reduce the computational complexity.

[0061] S2. Set the reflection intensity threshold to , and classify the points with a reflection intensity higher than in the preprocessed three-dimensional point cloud data of the bamboo forest as bamboo stalk point cloud data , and then use the density clustering algorithm to separate the bamboo stalk point cloud data to obtain a single bamboo stalk instance V;

[0062] S2.1. First, set the reflection intensity threshold . According to the different reflection characteristics of the ground, bamboo canopy, and bamboo stalks in the bamboo forest, initially extract the bamboo stalk point cloud through the intensity threshold. Classify the points with a reflection intensity higher than this threshold in the three-dimensional point cloud data of the bamboo forest as bamboo stalk point cloud data , as shown in part (a) of Figure 2 ;

[0063] S2.2. Then, for the bamboo stalk point cloud data , calculate its neighborhood density, and use the DBSCAN density clustering algorithm to further separate the bamboo stalk point cloud to obtain a single bamboo stalk instance , as shown in part (b) of Figure 2 ;

[0064] S3. Use the PCA principal component analysis method to process the single bamboo stalk instance V to obtain the main axis direction D of the bamboo stalk;

[0065] S3.1. Use Equation (1) to calculate the centered mean of the single bamboo stalk instance :

[0066] (1)

[0067] In Equation (1), represents the number of point clouds of the single bamboo stalk instance , is the coordinate of any i-th point in the single bamboo stalk instance ;

[0068] S3.2. Use Equation (2) to obtain the covariance matrix C of the single bamboo stalk instance :

[0069] (2)

[0070] In formula (2), T represents transposition;

[0071] S3.3. Perform eigendecomposition on the covariance matrix C to obtain a number of eigenvalues ​​and a number of eigenvectors, and use the first eigenvector D as the main axis direction of the single bamboo stalk instance V.

[0072] S4, such as Figure 3 As shown, along the main axis At fixed intervals Generate a set L of cross-sectional slices of a single bamboo stalk instance V, and perform bamboo node detection on each cross-sectional slice according to the joint feature entropy of the curvature change of the cross-sectional slice and the variance of the reflection intensity to obtain all bamboo node areas;

[0073] S4.1, along the main axis direction D at fixed intervals Generate a single bamboo stalk example A collection of cross-sectional slices ;

[0074] S4.2. Calculate the j-th cross-sectional slice of a single bamboo stalk instance V using equations (3) and (4) respectively: Curvature and intensity variance ;

[0075] (3)

[0076] (4)

[0077] In formula (3), is the jth cross-sectional slice The number of point clouds in for Middle The normal vector of a point, for The gradient value of

[0078] In formula (4), represents the jth cross-sectional slice Middle The reflection intensity value of each point, represents the jth cross-sectional slice The average reflection intensity of all points within.

[0079] S4.3. Calculate the j-th cross-sectional slice using equation (5) The joint feature entropy :

[0080] (5)

[0081] In formula (5), M represents the total number of slices in the set of cross-sectional slices ;

[0082] When the joint feature entropy of the j-th cross-sectional slice is greater than the entropy threshold, it indicates that the j-th cross-sectional slice is a bamboo node area. As shown, the entropy threshold determination is performed on each cross-sectional slice, and the regions that meet the conditions are regarded as the regions containing bamboo nodes (marked with yellow frames in the figure), reducing the false detection and missed detection caused by a single condition determination, and accurately determining the specific regions where the bamboo nodes are located on the bamboo stalk. Figure 4

[0083] S5. Use the bamboo node center point positioning method based on the minimization of the robust weighted potential function to process the detected bamboo node regions, obtain the bamboo node center point coordinates of each bamboo node region, sort all the bamboo node center point coordinates in ascending order, and then perform a third-order B-spline curve fitting on the point cloud data along the main axis direction between adjacent bamboo node center points, so as to calculate the length of the bamboo internodes of the moso bamboo.

[0084] S5.1. If the j-th cross-sectional slice is a bamboo node region, then use formula (6) to calculate the curvature of the -th point in the bamboo node region :

[0085] (6)

[0086] In formula (6), , , respectively represent the eigenvalues of the -th point.

[0087] S5.2. Use formula (7) to construct a robust weighted potential function :

[0088] (7)

[0089] In formula (7), represents the three-dimensional coordinates of the -th point in the bamboo node region , represents the Huber loss function, represents the three-dimensional coordinates of the preset initial bamboo node center point in the bamboo node region , represents the reflection intensity scale parameter, represents the curvature scale parameter.​

[0090] S5.3. Define and initialize the iteration step size as = 1, and initialize the three-dimensional coordinates of the center point of the bamboo node in the bamboo node area in the -1th iteration as ;

[0091] S5.4. Use Equation (8) to perform iterative update of gradient descent on to obtain the center point of the bamboo node in the bamboo node area in the th iteration: :

[0092] (8)

[0093] In Equation (8), represents the derivative of the Huber loss function;

[0094] S5.5. When is less than the termination threshold, it means that is the center point of the bamboo node in the bamboo node area , denoted as ; otherwise, assign + 1 to and then return to S5.4 to execute sequentially.

[0095] Due to the different materials and geometric shapes of bamboo nodes, there is usually a curvature peak, that is, the position where the bamboo node is located. However, the characteristics of three-dimensional point cloud data are disordered and discrete. Therefore, if the curvature peak point is directly used as the bamboo node, or simple center point calculation is performed as the actual coordinates of the bamboo node center point, errors will inevitably occur. Construct a robust weighted potential function F and perform iterative update of gradient descent to approach the actual center point of the bamboo node, and finally obtain the accurate coordinates of the bamboo node center point.

[0096] S5.6. Obtain the center point coordinates of each bamboo node area on a single bamboo stalk instance according to the process of S5.1 - S5.5, and after arranging them in ascending order along the main axis direction D, as shown in part (a) of Figure 5 , calculate the central axis between two adjacent bamboo node center points and ;

[0097] S5.7. Calculate the distances from all points between and to the central axis the distance, and taking the points with a distance less than the threshold as the fitting points, so as to use the cubic B-spline curve fitting method to fit all the fitting points, and obtain the actual growth path curve of the single bamboo culm instance V, as shown in Figure 5 part (b) of

[0098] S5.8. By integrating the arc length of the growth path curve, the between the internode length of the moso bamboo.

[0099] In this embodiment, an electronic device includes a memory and a processor. The memory is used to store a program that supports the processor to execute the above method, and the processor is configured to execute the program stored in the memory.

[0100] In this embodiment, a computer-readable storage medium stores a computer program, and when the computer program is run by a processor, it executes the steps of the above method.

Claims

1. A method for extracting the internode length of moso bamboo based on 3D point cloud, characterized in that, Including the following steps: S1. Collect three-dimensional point cloud data of bamboo forests and perform preprocessing to obtain preprocessed three-dimensional point cloud data of bamboo forests; S2. Set the reflection intensity threshold to , and classify the points in the preprocessed three-dimensional point cloud data of the bamboo forest with a reflection intensity higher than as the point cloud data of bamboo culms , and then use the density clustering algorithm to separate the point cloud data of bamboo culms to obtain a single bamboo culm instance V; S3. Use the PCA principal component analysis method to process the single bamboo culm instance V to obtain the main axis direction D of the bamboo culm; S4. Along the main axis direction At fixed intervals Generate a set L of cross-sectional slices of a single bamboo culm instance V, and perform bamboo node detection on each cross-sectional slice according to the joint feature entropy of the cross-sectional slice curvature change and the reflection intensity variance to obtain all bamboo node regions; S5. Use the bamboo node center point positioning method based on the minimization of the robust weighted potential function to process the detected bamboo node regions to obtain the bamboo node center point coordinates of each bamboo node region. After sorting all the bamboo node center point coordinates in ascending order, perform a cubic B-spline curve fitting on the point cloud data along the main axis direction between adjacent bamboo node center points, thereby calculating the length of the bamboo internodes.

2. The method for extracting the internode length of moso bamboo based on 3D point cloud according to claim 1, wherein S3 includes the following steps: S3.

1. Calculate the centralized mean of a single bamboo culm instance using Equation (1) :[[]]​ (1) In formula (1), represents the number of point clouds of a single bamboo culm instance , is the coordinate of any i-th point in the single bamboo culm instance ; S3.

2. Obtain the covariance matrix of the single bamboo culm instance using Equation (2) :​ (2) In formula (2), T represents the transpose; S3.

3. Perform eigen decomposition on the covariance matrix to obtain a number of eigenvalues and a number of eigenvectors, and use the first eigenvector D as the principal axis direction of the single bamboo culm instance V.

3. The method for extracting the internode length of moso bamboo based on 3D point cloud according to claim 2, wherein S4 includes the following steps: S4.

1. Generate a set of cross-sectional slices of a single bamboo culm instance at fixed intervals along the main axis direction D ;​​ S4.

2. Calculate the curvature and the strength variance of the j-th cross-sectional slice of the single bamboo culm instance V using equations (3) and (4) respectively ; (3) (4) In Equation (3), is the number of point clouds in the j-th cross-sectional slice , is the normal vector of the -th point in , and is the gradient value of In formula (4), represents the reflection intensity value of the nth point in the jth cross-sectional slice, and represents the average value of the reflection intensities of all points within the jth cross-sectional slice; S4.

3. Calculate the joint feature entropy of the j-th cross-sectional slice using Equation (5) :[[]]END]]​ (5) In formula (5), M represents the total number of slices in the set of cross-sectional slices ; When the joint feature entropy of the j-th cross-sectional slice is greater than the entropy threshold, it indicates that the j-th cross-sectional slice is a bamboo node region. ​ 4. The method for extracting the internode length of moso bamboo based on 3D point cloud according to claim 3, characterized in that S5 includes the following steps: S5.

1. If the j-th cross-sectional slice is a bamboo node region, then use Equation (6) to calculate the curvature of the n-th point in ; S5.

2. Constructing a robust weighted potential function using Equation (7) :[[]]END]] (7) In formula (7), represents the three-dimensional coordinates of the nth point in the bamboo node region, represents the Huber loss function, represents the three-dimensional coordinates of the preset initial bamboo node center point in the bamboo node region, represents the reflection intensity scale parameter, represents the curvature scale parameter; S5.

3. Define and initialize the iteration step size as = 1, and initialize the three-dimensional coordinates of the center point of the bamboo joint in the bamboo joint area at the -1-th iteration to be ; S5.

4. Use formula (8) to perform iterative update by gradient descent to obtain the bamboo node center point at the th iteration in the bamboo node area : (8) In Equation (8), represents the derivative of the Huber loss function; S5.

5. When is less than the suspension threshold, it indicates that is the center point of the bamboo node area , denoted as ; otherwise, after assigning +1 to , return to S5.4 and execute sequentially; S5.

6. Obtain a single bamboo culm instance according to the process of S5.1 - S5.5 The central point coordinates of each bamboo joint area on And The central axis of ; S5.

7. Calculation All the points between and the central axis are respectively measured for their distances. The points with distances less than the threshold are taken as the fitting points, and then the cubic B-spline curve fitting method is used to fit all the fitting points to obtain the actual growth path curve of the single bamboo culm instance V; S5.

8. By integrating the arc length of the growth path curve, the length between of the Phyllostachys pubescens internodes is obtained.

5. An electronic device, comprising a memory and a processor, characterized in that, The memory is used to store a program that supports the processor to execute the method for detecting bamboo nodes and extracting internode lengths described in any one of claims 1-4, and the processor is configured to execute the program stored in the memory.

6. A computer-readable storage medium having a computer program stored thereon, characterized in that, When the computer program is run by the processor, it executes the steps of the method for detecting bamboo nodes and extracting internode lengths described in any one of claims 1-4.

Images