A method, system and device for extracting a main root point cloud of a taproot system based on a three-dimensional point cloud

Abstract

The application provides a taproot system main root point cloud extraction method, system and device based on a three-dimensional point cloud, which comprises the following steps: slicing an original three-dimensional point cloud of a taproot system root along a root system depth direction to obtain a root slice point cloud; extracting an outermost contour of the root slice projection based on a root projection binary data set obtained by projecting the root slice point cloud to a two-dimensional projection plane and binarizing; if the outermost contour is a ring with holes or a ring with notches, the center of the ring cavity is a main root center position prediction value; selecting a root slice point cloud corresponding to a root projection binary data set closest to the main root center position prediction value set as a main root truncation point cloud; fitting a cylinder through the main root truncation point cloud to extract points in the main root truncation point cloud with a distance to the surface of the cylinder lower than a preset distance threshold, and obtain a taproot system main root point cloud. The method is more accurate in extracting the taproot system main root point cloud.

Description

TECHNICAL FIELD

[0001] The present application relates to the field of root system research, and in particular to a taproot main root point cloud extraction method, system and device based on three-dimensional point cloud. BACKGROUND

[0002] The root system of a plant is divided into taproot system and fibrous root system. The taproot system has a thick main root and thin lateral roots, and the main root and the lateral roots are obviously distinguished. The main root of the fibrous root system is not developed or stops growing early, and the root system is composed of adventitious roots generated from the stem base.

[0003] In the analysis of three-dimensional phenotype parameters of the taproot system, the main root and the lateral root need to be extracted first. However, the surface point cloud of the main root of the taproot system is not closed due to occlusion, insufficient illumination and the like, and is usually in the form of a ring with holes or a ring with notches. Moreover, the main feature of the main root is that the positions of the main root between adjacent slices are continuous. The existing technology does not consider these features of the main root of the taproot system, and thus cannot effectively extract the point cloud of the main root of the taproot system. SUMMARY

[0004] The present application provides a taproot main root point cloud extraction method, system and device based on three-dimensional point cloud, aiming at the shortcomings in the prior art.

[0005] To solve the above technical problems, the present application is solved by the following technical scheme:

[0006] A taproot main root point cloud extraction method based on three-dimensional point cloud, comprising the following steps:

[0007] The original three-dimensional point cloud of the root part of the taproot system is sliced along the root depth direction to obtain a plurality of root part slice point clouds;

[0008] The root part slice point clouds are respectively projected onto a two-dimensional projection plane and subjected to binary processing to obtain a plurality of root part projection binary data sets; wherein the two-dimensional projection plane is perpendicular to the root depth direction;

[0009] Based on the root part projection binary data sets, the outermost contour of the root part slice projection is extracted. If the outermost contour is a ring with holes or a ring with notches, the center of the ring cavity is the main root center position prediction value, and then a main root center position prediction value set is obtained;

[0010] The root part projection binary data sets are subjected to clustering processing, and the root part slice point cloud corresponding to the cluster with the closest center position to the main root center position prediction value set is selected as the main root truncated point cloud;

[0011] The points in the main root truncated point cloud with a distance to the surface of a cylinder below a preset distance threshold are extracted to obtain the main root point cloud of the taproot system; wherein the cylinder is fitted by the main root truncated point cloud.

[0012] As an implementable manner, the method for obtaining the main root center position prediction value based on the root projection binary data set comprises the following steps:

[0013] The outermost contour of the target root projection binary data set is extracted by using a contour detection algorithm to obtain an outermost contour pixel point sequence;

[0014] When the difference between the coordinate value of the first pixel point and the coordinate value of the last pixel point in the outermost contour pixel point sequence is less than or equal to a preset coordinate difference threshold, the projection of the main root is a hole ring.

[0015] When the difference between the coordinate value of the first pixel point and the coordinate value of the last pixel point in the outermost contour pixel point sequence is greater than the preset coordinate difference threshold, the projection of the main root is a notched ring.

[0016] As an implementable manner, when the outermost contour is a hole ring, the main root center position prediction value is obtained by comprising the following steps:

[0017] Based on the root projection binary data set, the holes of the hole ring are identified and filled to obtain a filled binary data set;

[0018] The two-dimensional points with the same coordinate value in the filled binary data set and the root projection binary data set are removed to obtain a hole binary data set;

[0019] The center coordinate of the hole binary data set is the main root center position prediction value.

[0020] As an implementable manner, when the outermost contour is a notched ring, the main root center position prediction value is obtained by comprising the following steps:

[0021] Based on the binary data set of the outermost contour of the notched ring, the recess boundary of the outermost contour is extracted to obtain a plurality of recess boundary projection binary data sets;

[0022] The maximum first distance value in the target recess region projection binary data set is taken as the depth value of the target recess region projection binary data set; wherein the first distance value is the Euclidean distance transformation value of the two-dimensional point to the recess boundary in the target recess region projection binary data set;

[0023] The recess region with the maximum depth value and greater than a preset depth threshold is taken as the main root inner cavity, and the center coordinate of the main root inner cavity is the main root center position prediction value.

[0024] As an implementable manner, the method for obtaining the main root center position prediction value based on the binary data set of the outermost contour of the notched ring comprises the following steps:

[0025] Based on the binarized data set of the outermost contour with a notch, a convex hull of the outermost contour is extracted to obtain an outer contour convex hull projection binarized data set;

[0026] Based on the outer contour convex hull projection binarized data set, a boundary of the outermost contour convex hull is extracted to obtain an outer contour convex hull boundary projection binarized data set;

[0027] The two-dimensional points belonging to the outer contour convex hull boundary projection binarized data set in the projection binarized data set are removed to obtain a total concave region projection binarized data set;

[0028] Based on the total concave region projection binarized data set, a plurality of concave region projection binarized data sets are obtained by combining connected component analysis;

[0029] The two-dimensional points belonging to the outer contour convex hull boundary projection binarized data set in each concave region projection binarized data set are removed to obtain a plurality of concave boundary projection binarized data sets.

[0030] As an implementable manner, the extraction of the points in the main root truncated point cloud with a distance to the surface of the cylinder less than a preset distance threshold to obtain the main root point cloud of the straight root system comprises the following steps:

[0031] Random sampling is performed on the main root truncated point cloud based on a preset sampling probability to obtain a main root truncated sampling point cloud;

[0032] A preliminary cylinder of the main root is fitted based on the main root truncated sampling point cloud;

[0033] The points in the main root truncated point cloud with a first main root cylinder distance less than a first preset residual error threshold are extracted to obtain a first internal point cloud; wherein the first main root cylinder distance is the distance from the point in the main root truncated point cloud to the surface of the preliminary cylinder;

[0034] Based on the Euclidean distance of the directional vector of the point in the first internal point cloud to the center line of the cylinder, an intermediate optimal cylinder is obtained by combining a nonlinear least squares method;

[0035] The points in the main root truncated point cloud with a second main root cylinder distance less than a second preset residual error threshold are extracted to obtain a second internal point cloud; wherein the second main root cylinder distance is the distance from the point in the main root truncated point cloud to the surface of the intermediate optimal cylinder;

[0036] When the total number of points of the second internal point cloud is greater than a preset point number threshold, the second point cloud is taken as the main root point cloud, and the preset point number threshold is iterated to the total number of points of the second internal point cloud; wherein the initial preset point number threshold is zero;

[0037] The above processing is repeated until the iteration number reaches a preset maximum iteration number.

[0038] As an implementable manner, the preliminary cylinder fitting the main root based on the main root truncated sampling point cloud comprises the following steps:

[0039] Based on the preset sampling point cloud slice number, the main root truncated sampling point cloud is sliced along the root depth direction to obtain a plurality of main root truncated slice sampling point clouds;

[0040] The main root truncated slice sampling point clouds at the maximum value and the minimum value in the root depth direction are extracted respectively to obtain a maximum value main root truncated sampling point cloud and a minimum value main root truncated sampling point cloud;

[0041] The minimum value main root truncated sampling point cloud and the maximum value main root truncated sampling point cloud are projected onto a two-dimensional projection plane to obtain a minimum value main root truncated projection dataset and a maximum value main root truncated projection dataset;

[0042] The minimum value main root truncated projection dataset and the maximum value main root truncated projection dataset are subjected to two-dimensional plane circle fitting processing to obtain minimum value projection center coordinate values and maximum value projection center coordinate values;

[0043] The minimum value projection center coordinate values and the maximum value projection center coordinate values are expanded to three dimensions to obtain minimum value three-dimensional center coordinate values and maximum value three-dimensional center coordinate values;

[0044] The minimum value three-dimensional center coordinate values and the maximum value three-dimensional center coordinate values are averaged to obtain a cylinder center coordinate of the preliminary cylinder;

[0045] The minimum value three-dimensional center coordinate values and the maximum value three-dimensional center coordinate values are subtracted and converted into unit vector processing to obtain a direction vector of the cylinder center line of the preliminary cylinder;

[0046] The distance of the points in the main root truncated sampling point cloud to the direction vector of the cylinder center line of the preliminary cylinder is calculated, and the distance is averaged to obtain a cylinder radius of the preliminary cylinder.

[0047] As an implementable manner, the intermediate optimal cylinder is obtained based on the Euclidean distance of the points in the first internal point cloud to the direction vector of the cylinder center line, combined with a nonlinear least squares method, comprising the following steps:

[0048] Based on the cylinder center coordinate, the Euclidean distance of each point in the first internal point cloud to the direction vector of the cylinder center line is calculated to obtain an internal cylinder Euclidean distance set;

[0049] The sum of the square of the difference between the Euclidean distance in the internal cylinder Euclidean distance set and the cylinder radius is calculated to obtain a cylinder surface residual value;

[0050] When the residual value of the cylindrical surface takes the minimum value, the center coordinates of the cylinder, the direction vector of the cylinder center line, and the cylinder radius are obtained, that is, the intermediate optimal cylinder parameters.

[0051] A taproot main root point cloud extraction system based on three-dimensional point cloud, characterized by being capable of realizing the method described in any one of the above, comprising a slicing module, a projection binarization module, a main root center prediction module, a main root truncation module, and a main root cylinder fitting module.

[0052] The slicing module is configured to perform slicing processing on the original three-dimensional point cloud of the taproot root along the root depth direction to obtain a plurality of root slice point clouds.

[0053] The projection binarization module is configured to project the root slice point cloud onto a two-dimensional projection plane and perform binarization processing to obtain a plurality of root projection binary data sets; wherein the two-dimensional projection plane is perpendicular to the root depth direction.

[0054] The main root center prediction module is configured to extract the outermost contour of the root slice projection based on the root projection binary data set; if the outermost contour is a ring with holes or a ring with notches, the center of the ring cavity is the main root center position prediction value, and then a main root center position prediction value set is obtained.

[0055] The main root truncation module is configured to perform clustering processing on the root projection binary data set, and select the root slice point cloud corresponding to the cluster closest to the main root center position prediction value set as the main root truncated point cloud.

[0056] The main root cylinder fitting module is configured to extract the points in the main root truncated point cloud whose distance to the surface of the cylinder is lower than a preset distance threshold, to obtain the main root point cloud of the taproot; wherein the cylinder is fitted by the main root truncated point cloud.

[0057] A computer-readable storage medium stores a computer program, and the computer program is executed by a processor to realize the method described above.

[0058] An apparatus includes a memory, a processor, and a computer program stored in the memory and running on the processor, and the processor realizes the method described above when executing the computer program.

[0059] The application has remarkable technical effects due to the above technical scheme: according to the characteristics of the taproot system containing thick main roots and thin lateral roots, the taproot system main root point cloud extraction method based on three-dimensional point cloud is proposed, after the root projection is identified as a hollow annular or a notched annular, the center of the annular cavity is taken as the main root center position prediction value, the root projection is subjected to clustering processing, the root slice point cloud corresponding to the cluster with the closest center position to the main root center position prediction value set is selected as the main root truncation point cloud, and then the cylindrical body is fitted through the main root truncation point cloud, and the cylindrical body parameters are used to extract the main root point cloud. The method of the application is suitable for main root extraction of the taproot system, considers the non-closed condition of the main root surface point cloud caused by occlusion, insufficient illumination and the like, filters out the interference of the inside lateral root truncation in the slice on the main root extraction, and the smoothing filtering of the main root position between adjacent slices reduces the influence of the main root position estimation error in a certain slice on the entire main root center line position. The method of the application is more targeted for the taproot system, and the extracted main root position and main root point cloud are more accurate. BRIEF DESCRIPTION OF DRAWINGS

[0060] In order to more clearly illustrate the technical solutions in the embodiments of the present application or the prior art, the following will briefly introduce the drawings needed to be used in the embodiments or the prior art description. Obviously, the drawings in the following description only some embodiments of the present application, and for those skilled in the art, other drawings can also be obtained without creative labor.

[0061] Figure 1 The flowchart of the method embodiment of the present application;

[0062] Figure 2 The slice diagram of the original three-dimensional point cloud in the embodiment of the present application is divided along the direction perpendicular to the root system depth;

[0063] Figure 3 The diagram of the main root (green part) in the embodiment of the present application;

[0064] Figure 4 The diagram of the main root in the embodiment of the present application is projected as a ring with holes in the two-dimensional projection plane;

[0065] Figure 5 The diagram of the main root in the embodiment of the present application is projected as a ring with notches in the two-dimensional projection plane;

[0066] Figure 6 The diagram of the filtered main root center prediction line in the continuous slice in the embodiment of the present application;

[0067] Figure 7 The point cloud diagram of the main root truncation sampling point cloud in the embodiment of the present application;

[0068] Figure 8 A side view schematic diagram of the main root truncated sampling point cloud in an embodiment of the present application;

[0069] Figure 9 A schematic diagram of the projection of the maximum value main root truncated sampling point cloud and the minimum value main root truncated sampling point cloud on a two-dimensional projection plane and the two two-dimensional plane circles fitted in an embodiment of the present application;

[0070] Figure 10 A schematic diagram of the first internal point cloud and the fitted intermediate optimal cylinder in an embodiment of the present application;

[0071] Figure 11 A point cloud schematic diagram of the main root point cloud (green part) in an embodiment of the present application;

[0072] Figure 12 A whole schematic diagram of the system embodiment of the present application. DETAILED DESCRIPTION

[0073] The present application will be further described in conjunction with embodiments. The following embodiments are explanations of the present application and the present application is not limited to the following embodiments. The features in the following embodiments can be combined with each other without conflict.

[0074] Embodiment 1:

[0075] In one embodiment, a taproot point cloud extraction method based on three-dimensional point cloud, as shown in FIG. 1, includes the following steps: Figure 1

[0076] S100: Slice the original three-dimensional point cloud of the taproot along the root depth direction to obtain a plurality of root slice point clouds;

[0077] S200: Project the root slice point cloud onto a two-dimensional projection plane and perform binaryzation processing to obtain a plurality of root projection binary data sets; wherein the two-dimensional projection plane is perpendicular to the root depth direction;

[0078] S300: Based on the root projection binary data set, extract the outermost contour of the root slice projection; if the outermost contour is a hole ring or a notched ring, the center of the ring cavity is the taproot center position prediction value, and then the taproot center position prediction value set is obtained;

[0079] S400: Cluster the root projection binary data set, and select the root slice point cloud corresponding to the cluster closest to the taproot center position prediction value set as the main root truncated point cloud;

[0080] ​S500: Extract the points in the taproot truncation point cloud that are less than a preset distance threshold from the surface of the cylinder, to obtain the taproot point cloud of the straight root system; wherein the cylinder is fitted by the taproot truncation point cloud.

[0081] The root depth direction refers to the main growth direction of the taproot, and the two-dimensional projection plane is a plane perpendicular to the root depth direction. In the specific embodiments, the root depth direction of the root point cloud is assumed to be the Y direction (first direction) of the existing three-dimensional coordinates, and the two-dimensional projection plane is the XZ two-dimensional plane of the existing three-dimensional coordinates for ease of illustration. After filtering the spatial range of the root point cloud (for example, BoxFilter), the root is divided into N (=40) slices along the Y direction, each slice has a taproot truncation, and each slice contains a number of taproots and lateral roots.

[0082] In another embodiment, in S100, the original three-dimensional point cloud of the straight root system root is sliced along the root depth direction to obtain a plurality of layers of root slice point clouds, including the following steps:

[0083] (1) Perform spatial range-based filtering on the original three-dimensional point cloud to obtain a spatially filtered root three-dimensional point cloud. Spatial range-based filtering of the original three-dimensional point cloud is a key step in point cloud preprocessing, which is mainly used to remove noise, outliers or extract specific regions, thereby reducing the amount of calculation and providing cleaner data for subsequent operations. Spatial range-based filtering methods include but are not limited to: direct spatial range filtering methods such as box filter, spherical / cylindrical filter, polygon / convex hull filter, which directly extract or delete specific regions in the point cloud by setting geometric boundaries; removing outliers by calculating the distance distribution of each point with neighboring points; voxel grid filtering; radius filtering that deletes points with fewer than a threshold number of neighbors; straight-through filtering that sets a range threshold along a specified axis (X / Y / Z) and retains points within the interval; spatial density clustering filtering that retains point clusters that meet spatial distribution conditions based on density clustering and removes isolated points, etc. The specific spatial filtering method can be selected according to the point cloud density, the actual situation of the taproot and the root system, and the calculation efficiency requirement.

[0084] (2) Slice the spatially filtered root three-dimensional point cloud along the root depth direction to obtain a plurality of root slice point clouds. As shown in FIG. 2, the original three-dimensional point cloud is divided into a plurality of slices along the direction perpendicular to the root depth direction, and the spatially filtered root three-dimensional point cloud is divided into a plurality of slice point clouds along the Y axis of the root depth direction, each slice containing a number of taproots and lateral roots. Figure 2 and the maximum value of the Y coordinates of all points in the spatially filtered root three-dimensional point cloud are found, and the spatially filtered root three-dimensional point cloud is divided into a plurality of slice point clouds along the Y axis of the root depth direction, each slice containing a number of taproots and lateral roots.​​ total span of Y axis dividing the value range of Y axis into intervals, then the height of each slice is , and the Y axis interval of the first, the , the slice is , , , Then each point in the spatial filtered root three-dimensional point cloud is assigned to the corresponding interval slice according to its Y coordinate value. The sliced point cloud is a local section of the original point cloud in the Y direction, similar to the part cut out from the point cloud with a "thin knife" parallel to the XZ plane, and the coordinate value of the sliced point cloud in the Y direction is within a very small range, which can be regarded as 2.5D data, a point set on the XZ plane, but with a slight difference in the Y direction.

[0085] If the spatial filtered root three-dimensional point cloud is not evenly distributed on the Y axis, some slices may contain very few points, or even some slices may have no points. In implementation, empty slices can be allowed to exist. Partial overlap can also be set for adjacent slices. Regardless of how it is set, ultimately all root slice point clouds cover the spatial filtered root three-dimensional point cloud.

[0086] Each slice contains a number of truncated main roots and truncated lateral roots, and due to the inclination and intersection of numerous lateral roots, the projection shape of the lateral root truncation on the XZ two-dimensional plane is indefinite. The main root truncation in each slice is approximately a hollow cylinder, so the projection of the main root truncation on the XZ two-dimensional plane is approximately a ring with a hole. However, due to the occlusion of lateral roots and other reasons, the main root surface point cloud obtained by three-dimensional reconstruction will have defects and the main root surface point cloud will not be closed, at which time the projection of the main root truncation on the XZ two-dimensional plane is similar to a ring with a notch. Therefore, the projection of each root slice on the XZ two-dimensional plane is performed, and the center position of the ring with a hole or the ring with a notch is extracted on the projected binary image, and the center position of the ring with a hole or the ring with a notch is recorded as the estimated center position of the main root truncation in the slice . As shown in Figures 3-5 , the green part in Figure 3 is a partial main root truncation, the Figure 4 is a schematic diagram of the main root truncation as a ring with a hole on the two-dimensional projection plane, and the Figure 5 is a schematic diagram of the main root truncation as a ring with a notch on the two-dimensional projection plane.

[0087] In another embodiment, in S200, the root slice point cloud is projected onto a two-dimensional projection plane and binarized to obtain a plurality of root projection binary data sets, including the following steps:

[0088] S210: Will Each root slice point cloud is projected onto a two-dimensional projection plane to obtain... A root slice projection dataset. Slice The projection of the point cloud onto the XZ two-dimensional plane is to... 3D points in a point cloud Convert to a 2D point that retains only the X and Z coordinates .

[0089] S220: Will The coordinates of two-dimensional points projected onto the XZ plane from the root slice projection dataset. Discretize into image coordinates to obtain A discrete dataset of root projection coordinates is provided for subsequent image processing. The discrete dataset of root projection coordinates is then represented as: ,in, Indicates the first The two-dimensional coordinates of a two-dimensional point in a discrete dataset of root projection coordinates. This represents the floor function. Indicates the first Two-dimensional coordinate values ​​of the root slice projection dataset. Indicates the first The minimum coordinate values ​​of the root slice point cloud in two perpendicular directions on the two-dimensional projection plane (XZ plane). This represents the side length of each voxel / mesh during spatial filtering, for example... .

[0090] S230: Set the pixel values ​​of the two-dimensional points in the root slice projection dataset to the first preset binary values ​​to obtain the root projection binary dataset, and set the pixel values ​​of other two-dimensional points in the two-dimensional projection plane to the second preset binary values.

[0091] The size is ,in, If sliced The point cloud coordinate range is: , , If gridSize=0.005, then The size is .

[0092] The binarization method involves projecting the XZ two-dimensional points... Corresponding image coordinates The pixel value of the first preset binary value (1, pure white) is assigned, and the pixel values ​​of other points are assigned the second preset binary value (0, pure black). Generally, the first preset binary value is set to 1 (pure white), and the second preset binary value is set to 0 (pure black), resulting in the final image. the projection binary image of each root section point cloud in the XZ plane As shown in Figure 3 the projection binary image corresponds to the root section point cloud one by one, and the number is also . The second preset binary can also be set to 1 (pure white), and the first preset binary can be set to 0 (pure black), as long as the projected points and non-projected points can be distinguished.

[0093] In the projection binary image , the main root truncation appears as a ring with a hole or a ring with a notch, and the outermost contour of the projection binary image is extracted. If the contour is closed (the starting point and the ending point of the contour coincide), the ring is not notched. If the contour has discontinuous endpoints (two obvious endpoints), the contour is not closed, and the ring has a notch.

[0094] In another embodiment, in S300, based on the root projection binary data set, the outermost contour of the root section projection is extracted; if the outermost contour is a ring with a hole or a ring with a notch, the center of the ring cavity is the main root center position prediction value, and then the main root center position prediction value set is obtained, including the following steps:

[0095] S310: Use a contour detection algorithm (such as a boundary tracking algorithm) to extract the outermost contour of the root projection binary data set, and obtain the outermost contour pixel point sequence.

[0096] S320: If the difference between the coordinate values of the first pixel point and the last pixel point in the outermost contour pixel point sequence is less than or equal to a preset coordinate difference threshold, the projection of the main root is a ring with a hole.

[0097] If the difference between the coordinate values of the first pixel point and the last pixel point in the outermost contour pixel point sequence is greater than the preset coordinate difference threshold, the projection of the main root is a ring with a notch.

[0098] S330: The center of the hole ring or the ring with a notch obtained by calculation is the main root center position prediction value, and then the main root center position prediction value set is obtained.

[0099] In another embodiment, when the projection of the main root is a ring with a hole, as shown in Figure 4 , it is a schematic diagram of the main root truncation projected as a ring with a hole in the two-dimensional projection plane, and the main root center position prediction value is obtained, as follows:

[0100] (1) Adopting hole recognition algorithm (such as Flood Fill edge seed method) and hole filling algorithm (such as Flood Fill internal filling, imfill function of matlab), based on root projection binary data set, the holes in the projection binary graph are recognized and filled, and the filled binary data set is obtained, which corresponds to the binary graph after filling the holes .

[0101] (2) The two-dimensional points with the same coordinate values in the filled binary data set and the root projection binary data set are removed, and the hole binary data set is obtained.

[0102] (3) The center coordinates of the hole binary data set are calculated, which are the estimated center coordinates of the main root lumen The center coordinates of the hole are the center position of the binary graph after filling the holes , which is the main root center position prediction value of the slice. Geometric moment centroid method, bounding box center method (Bounding Box center), maximum value point method based on distance transformation, Hough circle detection method, and pixel coordinate average value method can be used to calculate the center coordinates of the hole binary data set.

[0103] (4) The main root center position prediction values of other slices are obtained by the same method, and the main root center position prediction value set is obtained.

[0104] In another embodiment, when the projection of the main root is a notched ring, as shown in Figure 5 , it is a schematic diagram of the main root truncated in the two-dimensional projection plane and projected as a notched ring. For a notched ring, the recessed area in the notch should be determined first, and then the recessed area with the maximum depth and greater than the threshold value is taken as the main root lumen, and finally the main root center position prediction value is obtained. The specific steps are as follows:

[0105] (1) Using edge detection or contour tracking algorithm (such as Moore-Neighbor tracking), the outer contour of the projection binary graph is extracted, and the root outer contour projection binary data set is obtained. For a notched ring, there are two contours: outer contour and inner contour (notch part). The outer contour can be extracted by convex hull algorithm (such as Graham scan or Andrew algorithm). The convex hull convexHullMask is calculated by Andrew's Monotone Chain Algorithm or Graham Scan; the convex hull convexHullMask will ignore all recessed parts, forming a completely convex closed region;

[0106] (2) Based on the outer contour convex hull, get the convex hull mask, when the coordinates of a point is located in the closed region of the outer contour convex hull, then the point is the convex hull mask, which can be generated by the polygon filling algorithm (such as Scanline Fill). Extract the contour boundary from the convex hull mask to get the convex hull boundary imgHullPerim, which can be realized by edge detection or contour tracking algorithm.

[0107] (3) Difference operation between the convex hull mask convexHullMask and the projection binary graph The difference part is the total concave area (root total concave area projection binary data set, the sum of all concave areas); the total concave area is analyzed by connected component analysis to get multiple independent concave regions concaveRegions containing the main root inner cavity (multiple root concave area projection binary data set);

[0108] (4) Extract the coordinates common to the concave region concaveRegions and the convex hull boundary to get multiple concave region boundaries concaveBorder.

[0109] (5) Calculate the Euclidean distance transform value distBorder of each point in the concave region concaveRegions to the concave region boundary concaveBorder (i.e. the straight line distance of each point to the nearest concave region boundary).

[0110] (6) Take the maximum value of the Euclidean distance transform value distBorder as the depth of the concave region.

[0111] (7) Take the concave region with the maximum depth and greater than the depth threshold as the main root inner cavity, and the center coordinates of the main root inner cavity are the main root center position prediction value.

[0112] The following is a matlab code example for calculating the depth of the concave region:

[0113] concaveRegions = convexHullMask&~Ixz;

[0114] imgHullPerim = bwperim(convexHullMask);

[0115] concaveBorder = (concaveRegion&imgHullPerim);

[0116] distBorder = bwdist(concaveBorder);

[0117] concaveDepth = concaveRegion.*distBorder;

[0118] (8) Using the same method, the predicted values ​​of the main root center position of other slices are obtained, and then the set of predicted values ​​of the main root center position is obtained.

[0119] In another embodiment, in S400, clustering is performed on the binary dataset of root projections, and the root slice point cloud corresponding to the cluster whose center position is closest to the predicted value set of the principal root center position is selected as the principal root truncation point cloud, including the following steps:

[0120] S410: Perform median filtering on the predicted value set of the principal root center position to obtain the filtered predicted value set of the principal root center position.

[0121] Based on the continuity of the principal root center positions in adjacent slices, the predicted center position sequence of the principal roots in continuous slices is truncated. Window size is Median filtering, Where the number of slice layers containing the principal root is assumed to be... , This represents the total number of point cloud slices. This represents the total number of principal root point cloud slices. The filtered positions... Connect them sequentially to form the central line of the main root of the entire root system.

[0122] like Figure 6 The diagram shows the filtered principal root center prediction line. The red dots represent the principal root center positions corresponding to the predicted principal root positions, the green dots represent the filtered principal root center positions corresponding to the predicted principal root positions, and the blue line represents the filtered principal root center prediction line drawn based on the predicted principal root positions.

[0123] S420: Cluster all root slice projection datasets (the root truncation point clouds within the slice are projected onto the XZ plane) to obtain several clusters. It's important to note that this step first clusters a single root slice projection dataset, obtaining one or more clusters corresponding to that dataset; then, each of the other root slice projection datasets is clustered, resulting in one or more clusters for each dataset. All these clusters constitute the "several clusters" mentioned in this paragraph. In other words, a single cluster originates from the same slice.

[0124] S430: Calculate the center position of each cluster to obtain several cluster center positions.

[0125] S440: Calculate the center position corresponding to the predicted value set of the filtered principal root center position for each cluster center position. The distances are used to obtain several cluster principal root distances; the root slice point cloud corresponding to the minimum cluster principal root distance is the principal root truncation point cloud.

[0126] In another embodiment, in S500, points in the truncated point cloud of the primary root that are at a distance less than a preset distance threshold from the surface of the cylinder are extracted to obtain the primary root point cloud of the straight root system; wherein, the cylinder is fitted to the primary root using the truncated point cloud of the primary root, including the following steps:

[0127] S510: Based on the preset sampling probability in the principal root truncated point cloud Random sampling was performed to obtain the principal root truncated sampling point cloud. ,like Figure 7 As shown, Figure 7 This is a schematic diagram of a point cloud with a truncated sampled point cloud based on the principal root. The sampling probability determines the probability of randomly selecting the smallest sample set in each iteration, and is related to the proportion of inliers in the data. The higher the sampling probability (higher proportion of inliers), the fewer iterations are required. If the proportion of inliers is unknown, it can be estimated using an adaptive method. The residual threshold is an error threshold used to determine whether a data point is an inlier. When the error from a data point to the fitted model (e.g., the distance from a point to a line) is less than the residual threshold, the point is considered an inlier. The residual threshold is usually determined experimentally or based on data characteristics, depending on the specific problem. A residual threshold that is too small will miss valid inliers, leading to underfitting of the model. A residual threshold that is too large will mistakenly treat outliers as inliers, reducing the robustness of the model. The maximum number of iterations is used to limit the maximum number of iterations to avoid infinite loops.

[0128] Before fitting the initial cylinder of the principal root, the parameters need to be initialized. The truncated point cloud of the principal root is known. Initialize the maximum number of points (preset point threshold). The initial parameters are set to zero, i.e., the optimal cylinder parameters are initialized. Empty, that is Initialize the principal root point cloud Empty, that is The initial iteration count is zero, that is... Set sampling probability First preset residual threshold Second preset residual threshold and the maximum number of iterations For example, setting .

[0129] S520: Based on the sampling point cloud with truncated principal root Estimate the initial values ​​of the cylinder parameters Cylinder parameters Including the center coordinates of the cylinder cylinder radius The direction vector of the cylinder's centerline .

[0130] In another embodiment, the preliminary cylinder parameters are obtained, specifically including:

[0131] (1) Based on the preset number of sampling point cloud slices, the main root truncated sampling point cloud is sliced along the root depth direction (Y direction), and the main root truncated sampling point cloud at the maximum value and the minimum value in the root depth direction is extracted respectively, to obtain the maximum value main root truncated sampling point cloud and the minimum value main root truncated sampling point cloud, as shown in FIG. 1, which is a side view schematic diagram of the main root truncated sampling point cloud, wherein the blue point cloud is the maximum value main root truncated sampling point cloud, and the red point cloud is the minimum value main root truncated sampling point cloud. Figure 8

[0132] (2) The maximum value main root truncated sampling point cloud and the minimum value main root truncated sampling point cloud are projected onto a two-dimensional projection plane to obtain a maximum value main root truncated projection data set and a minimum value main root truncated projection data set.

[0133] (3) Two-dimensional plane circles of the minimum value main root truncated projection data set and the maximum value main root truncated projection data set are fitted to obtain the centers of the two fitted plane circles: the minimum value projection center coordinates and the maximum value projection center coordinates The two-dimensional plane circle can be fitted by using geometric fitting (such as Taubin method), algebraic fitting (such as Kasa method), least squares method, Hough transform method, etc. As shown in FIG. 2, which is a schematic diagram of the projection of the maximum value main root truncated sampling point cloud and the minimum value main root truncated sampling point cloud onto the two-dimensional projection plane and the fitting of the two two-dimensional plane circles, wherein the blue points are the points of the maximum value main root truncated sampling point cloud projected onto the two-dimensional projection plane, and the red points are the points of the minimum value main root truncated sampling point cloud projected onto the two-dimensional projection plane. Figure 9

[0134] (4) The minimum value projection center coordinates and the maximum value projection center coordinates are expanded to three dimensions to obtain the minimum value three-dimensional center coordinates and the maximum value three-dimensional center coordinates The minimum value three-dimensional center coordinates are represented as: , and the maximum value three-dimensional center coordinates are represented as: .

[0135] (5) The minimum value three-dimensional center coordinates and the maximum value three-dimensional center coordinates are averaged to obtain the median three-dimensional center coordinates which is the initial center coordinates of the cylinder. ​​

[0136] (6) The coordinates of the three-dimensional center of the circle for the minimum value and the maximum three-dimensional center coordinates Perform a subtraction operation to obtain the difference coordinates of the center of the circle. ; the difference in coordinates between the center of the circle Converting to a unit vector yields the initial direction vector of the cylinder's centerline. .

[0137] (7) Calculate the initial direction vector from the point in the principal root truncated point cloud to the center line of the cylinder. The distance is calculated, and the average distance is taken as the initial radius of the cylinder. The first point in the principal root truncation point cloud Points Initial direction vector to the center line of the cylinder distance Represented as: .

[0138] S530: Calculate the principal root truncation point cloud based on the initial parameters of the cylinder. From the point in the initial cylinder The distance to the surface is used to obtain the distance to the first principal root column. If the distance to the first principal root column is less than the first preset residual threshold, then the point is an interior point, and the first interior point cloud is finally obtained. Main root truncation point cloud The Middle The distance from each point to the surface of the initial cylinder (distance from the first principal root cylinder). Represented as: ,in, Represents the point cloud of the truncated principal root. The Middle The coordinates of each point This represents the initial center coordinates of the cylinder. This represents the initial direction vector of the cylinder's centerline. This represents the initial radius of the cylinder. The norm of a vector This represents the cross product of vectors.

[0139] S540: Based on the first internal point cloud The Euclidean distance from the midpoint to the axis of the cylinder is used to obtain the optimal parameters of the intermediate cylinder by combining the nonlinear least squares method.

[0140] In another embodiment, the optimal intermediate cylinder parameters are obtained as follows:

[0141] (1) For the first internal point cloud any point in ,calculate Euclidean distance of the direction vector to the cylinder center line, get the internal cylinder Euclidean distance , the first internal point cloud The Euclidean distance of the direction vector to the cylinder center line of the th point in the th point cloud is expressed as: , where represents the reference point on the cylinder axis (usually the center coordinates of the cylinder), represents the direction vector of the cylinder center line, represents the norm of the vector, represents the dot product (inner product) of the vector. represents the distance square (total length square) of point to point represents the square of the projection length of vector in the direction of the cylinder axis, represents the perpendicular distance of point to the cylinder axis, that is, the radial distance of the cylinder, that is, the estimated cylinder radius.

[0142] Therefore, the internal cylinder Euclidean distance can also be expressed as: , where represents the coordinate value of point , and represents the coordinate value of the center point of the cylinder , and represents the direction vector of the cylinder center line.

[0143] (2) Calculate the sum of the square of the difference between the internal cylinder Euclidean distance and the cylinder radius , get the cylinder surface residual value .

[0144] Where the cylinder center coordinates , the direction vector of the cylinder center line , and the cylinder radius are all undetermined variables, and the purpose of the nonlinear least squares method is to find the cylinder center coordinates , the direction vector of the cylinder center line , and the cylinder radius that minimize the cylinder surface residual value.

[0145] (3) When the residual value on the cylinder surface reaches its minimum, the center coordinates of the cylinder, the direction vector of the cylinder's centerline, and the cylinder's radius are obtained, which are the intermediate optimal cylinder parameters (intermediate optimal cylinder center coordinates). , the direction vector of the center line of the optimal cylinder and the optimal radius of the intermediate cylinder ).like Figure 10 The diagram shows a schematic of the first internal point cloud and the fitted intermediate optimal cylinder, where red represents the point cloud of the first internal point cloud and blue represents the fitted intermediate optimal cylinder.

[0146] The following is a MATLAB code example for fitting the intermediate optimal parameters of a cylinder using the nonlinear least squares method:

[0147] options = optimoptions('lsqnonlin', 'Algorithm', 'levenberg-marquardt');

[0148] CylinderParai = lsqnonlin(@(params)cylinderResiduals(params,pts1),CylinderPara0, [], [], options);

[0149] The `cylinderResiduals` function calculates the distance between the point cloud `points` and the surface of the cylinder `params`. The MATLAB code is as follows:

[0150] function residuals = cylinderResiduals(params, points)

[0151] % Vectorized computation

[0152] vecs = points - params.center;% N×3

[0153] proj_lengths = vecs * params.axisDir';% N×1 (dot product)

[0154] proj_points = params.center + proj_lengths * params.axisDir;% N×3

[0155] distances = sqrt(sum((points - proj_points).^2, 2)); % N×1

[0156] residuals = distances - params.radius;% N×1

[0157] end

[0158] S550: Based on the intermediate optimal cylinder parameters, the distance from the main root truncated point cloud to the surface of the intermediate optimal cylinder is calculated to obtain the second main root cylinder distance; the points in the main root truncated point cloud with the second main root cylinder distance less than the second preset residual threshold are extracted to obtain the second internal point cloud. The first preset residual threshold and the second preset residual threshold can be set to be the same or different, and can be set according to actual needs.

[0159] S560: When the total number of points in the second internal point cloud is greater than the preset point number threshold, the second point cloud is selected as the main root point cloud, and the preset point number threshold is iterated to the total number of points in the second internal point cloud, and the intermediate optimal cylinder (parameters) is the optimal cylinder (parameters) fitted by the main root point cloud. That is, the main root point cloud is the second point cloud with the largest total number of points.

[0160] S570: Steps S520-S560 are repeated until the iteration number reaches the preset maximum iteration number, to obtain the main root point cloud and the corresponding cylinder, as shown in FIG. 6B, which is a point cloud diagram of the main root point cloud (green part). Figure 11

[0161] Embodiment 2:

[0162] A straight root system main root point cloud extraction system based on three-dimensional point cloud, as shown in FIG. 1, can implement the method described in any one of the embodiments, and includes a slicing module 100, a projection binarization module 200, a main root center prediction module 300, a main root truncation module 400, and a main root cylinder fitting module 500. Figure 12 The slicing module 100 is configured to perform slicing processing on the original three-dimensional point cloud of the root part of the straight root system along the root depth direction to obtain a plurality of layers of root part slice point clouds.

[0163] The projection binarization module 200 is configured to project the root part slice point cloud onto a two-dimensional projection plane and perform binarization processing to obtain a plurality of root part projection binary data sets; wherein the two-dimensional projection plane is perpendicular to the root depth direction.

[0164] The main root center prediction module 300 is configured to extract the outermost contour of the root part slice projection based on the root part projection binary data set; if the outermost contour is a ring with holes or a ring with notches, the center of the ring cavity is a main root center position prediction value, and then a main root center position prediction value set is obtained.

[0165] ​​

[0166] The main root truncation module 400 is configured to perform clustering processing on the root projection binary data set, and select a root section point cloud corresponding to a cluster closest to a main root center position prediction value set as a main root truncation point cloud.

[0167] The main root column fitting module 500 is configured to extract points in the main root truncation point cloud having a distance to a cylindrical surface lower than a preset distance threshold value, to obtain the main root point cloud of the straight root system, wherein the cylindrical surface is fitted by the main root truncation point cloud.

[0168] Various changes and modifications can be made within the spirit and scope of the application without departing from it, and all equivalent technical solutions also belong to the scope of the application.

[0169] Each of the embodiments in the specification is described in a progressive manner, and each embodiment focuses on the differences from other embodiments. The same or similar parts between the embodiments can be referred to each other.

[0170] Those skilled in the art should understand that the embodiments of the present application can be provided as a method, device, or computer program product. Therefore, the present application can take the form of a complete hardware embodiment, a complete software embodiment, or an embodiment combining software and hardware aspects. Moreover, the present application can take the form of a computer program product implemented on one or more computer usable storage media (including but not limited to magnetic disk storage, CD-ROM, optical storage, etc.) containing computer usable program code.

[0171] The present application is described with reference to flowcharts and / or block diagrams according to the method, terminal device (system), and computer program product of the present application. It should be understood that each flow and / or block in the flowchart and / or block diagram, and the combination of flows and / or blocks in the flowchart and / or block diagram can be realized by computer program instructions. These computer program instructions can be provided to a general-purpose computer, a special-purpose computer, an embedded processor, or other programmable data processing terminal device to produce a machine, so that the instructions executed by the processor of the computer or other programmable data processing terminal device produce a means for implementing the functions specified in the flowchart Figure 1 The means for implementing the functions specified in the flowchart Figure 1 The means for implementing the functions specified in the flowchart

[0172] These computer program instructions can also be stored in a computer readable storage medium that can guide the computer or other programmable data processing terminal device to work in a specific way, so that the instructions stored in the computer readable storage medium produce a product including instruction means, which implements the functions specified in the flowchart Figure 1 The means for implementing the functions specified in the flowchart Figure 1the function specified in the one or more blocks.

[0173] These computer program instructions can also be loaded into computer or other programmable data processing terminal devices, so that a series of operation steps are performed on the computer or other programmable terminal devices to generate computer-implemented processes, so that the instructions executed on the computer or other programmable terminal devices provide processes for implementing the flowcharts Figure 1 one or more flows and / or blocks Figure 1 the function specified in the one or more blocks.

[0174] It should be noted that:

[0175] The phrase "one embodiment" or "an embodiment" appearing in the specification does not necessarily refer to the same embodiment, although it can. Any implementation of the techniques described in this specification that are included in at least one embodiment of the present invention are intended to fall within the scope of the present invention.

[0176] In addition, it should be noted that the specific embodiments described in this specification, the shape of the zero, component, the name taken, etc. can be different. Any equivalent or simple change made in accordance with the configuration, features and principles described in the patent concept of the present invention is included in the protection scope of the present invention. Those skilled in the art can make various modifications or supplements to the described specific embodiments or use similar ways to replace them, as long as they do not deviate from the structure of the present invention or exceed the scope defined by the present claims.

Claims

1. A method for taproot point cloud extraction based on three-dimensional point cloud, characterized in that, The method comprises the following steps: The original three-dimensional point cloud of the taproot root is sliced along the root depth direction to obtain a plurality of root slice point clouds; The root slice point clouds are projected onto a two-dimensional projection plane and binarized to obtain a plurality of root projection binary data sets; wherein the two-dimensional projection plane is perpendicular to the root depth direction; Based on the root projection binary data set, the outermost contour of the root slice projection is extracted; if the outermost contour is a ring with holes or a ring with notches, the center of the ring inner cavity is the main root center position prediction value, and then a main root center position prediction value set is obtained; The root projection binary data set is clustered, and the root slice point cloud corresponding to the cluster with the closest center position to the main root center position prediction value set is selected as the main root truncation point cloud; The points in the main root truncation point cloud with a distance to the surface of the cylinder below a preset distance threshold are extracted to obtain the main root point cloud of the taproot; wherein the cylinder is fitted by the main root truncation point cloud.

2. The three-dimensional point cloud based taproot hill point cloud extraction method according to claim 1, wherein, The method for extracting the outermost contour of the root slice projection based on the root projection binary data set comprises the following steps: The outermost contour of the target root projection binary data set is extracted using a contour detection algorithm to obtain an outermost contour pixel point sequence; When the difference between the coordinate values of the first pixel point and the last pixel point in the outermost contour pixel point sequence is less than or equal to a preset coordinate difference threshold, the projection of the main root is a ring with holes; When the difference between the coordinate values of the first pixel point and the last pixel point in the outermost contour pixel point sequence is greater than the preset coordinate difference threshold, the projection of the main root is a ring with notches.

3. The three-dimensional point cloud based taproot hill point cloud extraction method according to claim 1, wherein, When the outermost contour is a ring with holes, the main root center position prediction value is obtained by the following steps: Based on the root projection binary data set, the holes of the ring with holes are identified and filled to obtain a filled binary data set; The two-dimensional points in the filled binary data set with the same coordinate values as those in the root projection binary data set are removed to obtain a hole binary data set; The center coordinates of the hole binary data set are calculated as the main root center position prediction value.

4. The three-dimensional point cloud based taproot hill point cloud extraction method according to claim 1, wherein, When the outermost contour is a ring with notches, the main root center position prediction value is obtained by the following steps: Based on the binary data set of the outermost contour of the ring with notches, the recess boundary of the outermost contour is extracted to obtain a plurality of recess boundary projection binary data sets; The maximum first distance value in the target recess region projection binary data set is taken as the depth value of the target recess region projection binary data set; wherein the first distance value is the Euclidean distance transform value of the two-dimensional points in the target recess region projection binary data set to the recess boundary; The recess region with the maximum depth value and greater than a preset depth threshold is taken as the main root inner cavity, and the center coordinates of the main root inner cavity are taken as the main root center position prediction value.

5. The three-dimensional point cloud based taproot hill point cloud extraction method according to claim 4, characterized in that, The method for extracting the recess boundary of the outermost contour based on the binary data set of the outermost contour of the ring with notches to obtain a plurality of recess boundary projection binary data sets comprises the following steps: Based on the binary data set of the outermost contour of the ring with notches, the convex hull of the outermost contour is extracted to obtain an outer contour convex hull projection binary data set; Based on the outer contour convex hull projection binary data set, the boundary of the outermost contour convex hull is extracted, and an outer contour convex hull boundary projection binary data set is obtained; The two-dimensional points belonging to the outer contour convex hull projection binary data set in the projection binary data set are removed, and a total concave region projection binary data set is obtained; Based on the total concave region projection binary data set, combined with connected component analysis, a plurality of concave region projection binary data sets are obtained; The two-dimensional points belonging to the outer contour convex hull boundary projection binary data set in each concave region projection binary data set are removed, and a plurality of concave boundary projection binary data sets are obtained.

6. The three-dimensional point cloud based taproot hill point cloud extraction method according to claim 1, wherein, The distance of the point in the main root truncated point cloud to the surface of the cylinder is less than the preset distance threshold value, and the main root point cloud of the straight root system is obtained, including the following steps: Random sampling is performed on the main root truncated point cloud based on a preset sampling probability, and a main root truncated sampling point cloud is obtained; A preliminary cylinder of the main root is fitted based on the main root truncated sampling point cloud; The points in the main root truncated point cloud with a first main root cylinder distance less than a first preset residual threshold value are extracted, and a first internal point cloud is obtained; wherein the first main root cylinder distance is the distance from the point in the main root truncated point cloud to the surface of the preliminary cylinder; Based on the Euclidean distance of the direction vector of the point in the first internal point cloud to the cylinder center line, combined with a nonlinear least squares method, an intermediate optimal cylinder is obtained; The points in the main root truncated point cloud with a second main root cylinder distance less than a second preset residual threshold value are extracted, and a second internal point cloud is obtained; wherein the second main root cylinder distance is the distance from the point in the main root truncated point cloud to the surface of the intermediate optimal cylinder; When the total number of points in the second internal point cloud is greater than a preset point number threshold value, the second point cloud is taken as the main root point cloud, and the preset point number threshold value is iterated to the total number of points in the second internal point cloud; wherein the initial preset point number threshold value is zero; The above processing is repeated until the iteration number reaches a preset maximum iteration number.

7. The three-dimensional point cloud based taproot hill point cloud extraction method according to claim 6, wherein, The preliminary cylinder of the main root is fitted based on the main root truncated sampling point cloud, including the following steps: Based on the preset sampling point cloud slice number, the main root truncated sampling point cloud is sliced along the root depth direction to obtain a plurality of main root truncated slice sampling point clouds; The maximum value main root truncated sampling point cloud and the minimum value main root truncated sampling point cloud are obtained by respectively extracting the main root truncated slice sampling point clouds at the maximum value and the minimum value in the root depth direction; The minimum value main root truncated sampling point cloud and the maximum value main root truncated sampling point cloud are projected onto a two-dimensional projection plane to obtain a minimum value main root truncated projection data set and a maximum value main root truncated projection data set; The minimum value main root truncated projection data set and the maximum value main root truncated projection data set are subjected to two-dimensional plane circle fitting processing to obtain minimum value projection center coordinate values and maximum value projection center coordinate values; The minimum value projection center coordinate values and the maximum value projection center coordinate values are extended to three dimensions to obtain minimum value three-dimensional center coordinate values and maximum value three-dimensional center coordinate values; The minimum value three-dimensional center coordinate values and the maximum value three-dimensional center coordinate values are averaged to obtain the cylinder center coordinate of the preliminary cylinder; The minimum value three-dimensional center coordinate values and the maximum value three-dimensional center coordinate values are subtracted and converted to unit vector processing to obtain the direction vector of the cylinder center line of the preliminary cylinder; The distance of each point in the main root truncated point cloud to the direction vector of the cylinder center line of the preliminary cylinder is calculated, and the distances are averaged to obtain the cylinder radius of the preliminary cylinder.

8. The three-dimensional point cloud based taproot hill point cloud extraction method according to claim 6, wherein, The intermediate optimal cylinder is obtained based on the Euclidean distance of each point in the first internal point cloud to the direction vector of the cylinder center line, and a nonlinear least squares method. Based on the cylinder center coordinates, the Euclidean distance of each point in the first internal point cloud to the direction vector of the cylinder center line is calculated to obtain an internal cylinder Euclidean distance set. The sum of the square of the difference between the Euclidean distance in the internal cylinder Euclidean distance set and the cylinder radius is calculated to obtain a cylinder surface residual value. When the cylinder surface residual value takes the minimum value, the cylinder center coordinates, the direction vector of the cylinder center line, and the cylinder radius are obtained as the intermediate optimal cylinder parameters.

9. A taproot system based on three-dimensional point cloud of taproot point cloud extraction system, characterized in that, The method of any one of claims 1 to 8 can be implemented, including a slicing module, a projection binarization module, a main root center prediction module, a main root truncation module, and a main root cylinder fitting module. The slicing module is configured to perform slicing processing on the original three-dimensional point cloud of the taproot root along the root depth direction to obtain a plurality of root slice point clouds. The projection binarization module is configured to project the root slice point cloud onto a two-dimensional projection plane and perform binarization processing to obtain a plurality of root projection binary data sets; wherein the two-dimensional projection plane is perpendicular to the root depth direction. The main root center prediction module is configured to extract the outermost layer contour of the root slice projection based on the root projection binary data set; if the outermost layer contour is a ring with holes or a ring with notches, the center of the ring cavity is the main root center position prediction value, and a main root center position prediction value set is obtained. The main root truncation module is configured to perform clustering processing on the root projection binary data set, and select the root slice point cloud corresponding to the cluster closest to the main root center position prediction value set as the main root truncated point cloud. The main root cylinder fitting module is configured to extract the points in the main root truncated point cloud that are below a preset distance threshold from the surface of the cylinder to obtain the main root point cloud of the taproot; wherein the cylinder is fitted by the main root truncated point cloud.

10. A computer-readable storage medium storing a computer program, the computer-readable storage medium comprising: The computer program is executed by the processor to implement the method of any one of claims 1 to 8.

11. An apparatus comprising a memory, a processor, and a computer program stored in the memory and running on the processor, characterized in that, The processor executes the computer program to implement the method of any one of claims 1 to 8.

Images