Automatic point cloud segmentation method based on deep learning algorithm
Through the point cloud automatic segmentation method based on deep learning algorithm, undeveloped saplings in the seedling forest are marked and positioned, which solves the problem of manual search, improves work efficiency and reduces costs.
Patent Information
- Application Number
- CN202311696325.3
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2023-12-12
- Publication Date
- 2025-06-13
AI Technical Summary
Because the seedling forest area is large, it is more troublesome to find undeveloped seedlings manually, which affects work efficiency and requires an automatic segmentation and processing method.
The automatic point cloud segmentation method based on deep learning algorithm is adopted. The measurement area is segmented by the measuring personnel, and the saplings that need to be segmented are marked and positioned from the three-dimensional point cloud data. Various measurement methods and big data are compared, and the range values of the sapling height, thickness and shape are set, and light and heavy marks are performed to facilitate removal.
Accurate and rapid marking and positioning of undeveloped saplings in the seedling forest is achieved, work efficiency is improved, labor cost loss is reduced, and huge benefits are brought to growers.
Smart Images
Figure CN120147625A_ABST
Abstract
Description
Technical Field
[0001] The present invention relates to the technical field of sapling planting, and particularly relates to a method for automatically segmenting point clouds based on a deep learning algorithm. Background Art
[0002] A point cloud is a massive set of points that expresses the spatial distribution of an object and the surface characteristics of the object under the same spatial reference system. The attributes of the point cloud include spatial resolution, point position accuracy, surface normal vector, etc. Sparse point clouds or dense point clouds are the basis for reverse modeling, and there are many specialized reverse software that can perform editing and processing of point clouds; During the process of sapling planting, it is necessary to remove saplings with relatively low vitality or incomplete development in the sapling forest. However, due to the large area of the sapling forest, it is troublesome to search manually, which affects the work efficiency. Therefore, it is necessary to use the point cloud method for automatic segmentation processing. Summary of the Invention
[0003] For this reason, the present invention provides a method for automatically segmenting point clouds based on a deep learning algorithm. Through the measurement personnel, the point cloud area is segmented and processed. From the generated three-dimensional point cloud data, the saplings to be segmented can be marked and located, which is convenient for the user to remove the saplings, so as to solve the problem that due to the large area of the sapling forest, it is troublesome to search manually, which affects the work efficiency. Therefore, it is necessary to use the point cloud method for automatic segmentation processing.
[0004] In order to achieve the above object, the present invention provides the following technical solution: A method for automatically segmenting point clouds based on a deep learning algorithm, including S1: Before automatically segmenting the point cloud of the saplings, it is necessary for the measurement personnel to accurately divide and locate the point cloud area and perform area isolation processing at the same time; S2: After the area isolation is completed, the measurement personnel prepare a variety of measurement methods to perform omnidirectional photogrammetry processing on the point cloud area; S3: The measurement results are compared and combined through big data, and the combined data is summarized; S301: Statistic the summarized results, take a range value of the relative sapling height, thickness and shape, and measure the number of all saplings within the range value through the point cloud; S302: Extract the information of the saplings that do not meet the range value of the sapling height, thickness and shape; S303: Compare the extracted sapling information with big data again, and set a specific threshold outside the sapling range value; S304: Mildly mark the saplings that do not exceed this specific threshold. The marked saplings are amenable to trimming, and can be trimmed by spraying pesticides and pruning to ensure the normal growth of these saplings. The saplings that exceed this specific threshold are subjected to the next segmentation processing; S4: After measurement, the point cloud data is segmented by the point cloud automatic segmentation system, and the saplings to be isolated are heavily marked and located, facilitating the user to perform the next removal process on these saplings.
[0005] Preferably, the measurement methods described in S2 include terrestrial three-dimensional laser scanning acquisition, vehicle-mounted MMS acquisition, handheld laser scanning acquisition, and unmanned aerial vehicle scanning acquisition.
[0006] Preferably, the photogrammetry described in S2 can obtain a high-precision three-dimensional model by a series of calculations on two-dimensional images, and in this process, we can also obtain point cloud data.
[0007] Preferably, the big data system described in S3 includes sapling height data, sapling thickness data, and sapling production shape data.
[0008] Preferably, the content of the marking and positioning described in S4 is displayed by coordinate positioning.
[0009] The beneficial effects of the present invention are as follows: 1. Through the measurement personnel's point cloud segmentation processing of the measurement area, the saplings to be segmented can be marked and located from the generated three-dimensional point cloud data, facilitating the user to remove the saplings, so as to solve the problem that due to the large area of the sapling forest, it is troublesome to search manually, which affects the work efficiency. Therefore, it is necessary to use the point cloud method for automatic segmentation processing. This method can accurately and quickly mark and locate the underdeveloped saplings inside the sapling forest, greatly improving the work efficiency, and at the same time reducing the loss of labor costs, bringing great benefits to the growers; 2. In addition, this device uses a three-dimensional modeling system to quickly calculate the scanning results and display them in detail after the calculation, bringing great convenience to the user. At the same time, this device uses a variety of scanning devices to scan the point cloud area, making the detection effect more perfect. Description of the Drawings
[0010] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the following will briefly introduce the drawings required for the description of the embodiments or the prior art. Obviously, the drawings in the following description are only exemplary, and for those of ordinary skill in the art, without creative efforts, other implementation drawings can also be obtained according to the provided drawings.
[0011] The structures, ratios, sizes, etc. illustrated in this specification are only used to cooperate with the content disclosed in the specification for those familiar with this technology to understand and read, and are not used to limit the implementation conditions of the present invention. Therefore, they do not have substantial technical significance. Any modification of the structure, change in the proportional relationship, or adjustment of the size, without affecting the efficacy that the present invention can produce and the purpose that can be achieved, should still fall within the scope covered by the technical content disclosed in the present invention.
[0012] Figure 1 It is a schematic diagram of the overall process provided by the present invention; Figure 2 It is a schematic diagram of the comparison between the sapling point cloud data and big data provided by the present invention; Figure 3 It is a schematic diagram of the overall system structure provided by the present invention; Figure 4 It is a schematic diagram of the information acquisition system provided by the present invention; Specific embodiments
[0013] The following is a description of the preferred embodiments of the present invention with reference to the accompanying drawings. It should be understood that the preferred embodiments described herein are only used to illustrate and explain the present invention and are not used to limit the present invention.
[0014] Referring to the attached Figure 1 - attached Figure 4 , the point cloud automatic segmentation method based on deep learning algorithm provided by the present invention includes S1: Before the point cloud of the saplings is automatically segmented, the surveyor needs to accurately divide the point cloud area and at the same time perform area isolation processing; S2: After the area isolation is completed, the surveyor prepares various measurement methods to perform omnidirectional photogrammetry processing on the point cloud area; S3: The measurement results are compared and combined with big data, and the combined data is summarized; S301: Statistically process the summarized results, take a range value of the relative sapling height, thickness and shape, and measure the number of all saplings within the compound range value through point cloud; S302: Extract the information of the saplings that do not meet the range value of the sapling height, thickness and shape; S303: Compare the extracted sapling information with big data again, and set a specific threshold outside the sapling range value; S304: Mildly mark the saplings that do not exceed this specific threshold. The saplings marked this time are amenable to trimming. They can be trimmed by spraying pesticides and pruning to ensure the normal growth of these saplings. The saplings that exceed this specific threshold are subjected to the next segmentation process; S4: The point cloud data after measurement is segmented by the point cloud automatic segmentation system, and the saplings to be isolated are heavily marked and located, facilitating the user to perform the next removal process on these saplings. Among them, to achieve the purpose of data acquisition, the present device is implemented by the following technical solutions: The measurement methods described in S2 include terrestrial three-dimensional laser scanning acquisition, vehicle-mounted MMS acquisition, handheld laser scanning acquisition, and unmanned aerial vehicle scanning acquisition. Among them, to achieve the purpose of measurement and calculation, the present device is implemented by the following technical solutions: The photogrammetry described in S2 can obtain a high-precision three-dimensional model through a series of calculations on two-dimensional images, and in this process, we can also obtain point cloud data. Among them, to achieve the purpose of marking and positioning, the present device is implemented by the following technical solutions: The big data system described in S3 includes sapling height data, sapling thickness data, and sapling production shape data, and the content of marking and positioning described in S4 is displayed by coordinate positioning.
[0015] The usage process of the present invention is as follows: First, the staff first locates the sapling area to be measured to achieve an isolation effect, making the area fixed during point cloud detection, thereby making the positioning effect more accurate. After isolation, information collection methods including terrestrial three-dimensional laser scanning acquisition, vehicle-mounted MMS acquisition, handheld laser scanning acquisition, and unmanned aerial vehicle scanning acquisition are used to collect information on the saplings in the isolated area. The collected information is compared and screened with big data through a computer. The data on the regular length characteristics, thickness characteristics, and shape characteristics of saplings in the big data are compared and screened with the data of the saplings in the detected area. Then, the saplings after screening are summarized, the results after summarization are statistically analyzed, a range value of the relative sapling height, thickness, and shape is taken, and the number of saplings within the composite range value of all saplings is measured by point cloud. The information of the saplings that do not meet the range values of sapling height, thickness, and shape is extracted, and the extracted sapling information is compared with big data again. A specific threshold outside the sapling range value is set, and the saplings that do not exceed this specific threshold are lightly marked. The saplings marked this time are amenable to trimming and can be trimmed by spraying pesticides and pruning to ensure the normal growth of these saplings. The saplings that exceed this specific threshold are subjected to the next segmentation process. The number of saplings to be processed is summarized, and then calculated by photogrammetry. Photogrammetry can obtain a high-precision three-dimensional model through a series of calculations on two-dimensional images, thereby obtaining their specific positions. Multiple underdeveloped saplings are individually and heavily marked and located, and the marking process is carried out by polar coordinate positioning to facilitate the user to transplant specific saplings. This method can accurately and quickly mark and locate the incompletely developed saplings inside the sapling forest, greatly improving the work efficiency and reducing the loss of labor costs, bringing great benefits to growers. In addition, this device uses a three-dimensional modeling system to quickly calculate the scanning results and displays them in detail after the calculation, bringing great convenience to users. At the same time, this device uses a variety of scanning devices to scan and process the point cloud area, making the detection effect more perfect.
[0016] The above are only the preferred embodiments of the present invention. Any person skilled in the art may modify the present invention by using the technical solutions described above or modify it into an equivalent technical solution. Therefore, any simple modification or equivalent replacement made according to the technical solutions of the present invention falls within the scope of the present invention's claims.
Claims
1. A point cloud automatic segmentation method based on deep learning algorithms, comprising S1: Before the automatic segmentation of the point cloud of saplings, it is necessary for the surveyors to accurately divide the point cloud area and at the same time perform area isolation processing; S2: After the area isolation is completed, the surveyors prepare various measurement methods to perform all-round photogrammetric measurement on the point cloud area; S3: The measurement results are compared and combined through big data, and the combined data is summarized; S301: Statistically process the summarized results, obtain a range value for the relative height, thickness and shape of the saplings, and measure the number of all saplings within the composite range value through the point cloud; S302: Extract the information of the saplings that do not meet the range values of the height, thickness and shape of the saplings; S303: Compare the extracted sapling information with the big data again and set a specific threshold outside the sapling range value; S304: Mildly mark the saplings that do not exceed this specific threshold. The marked saplings are amenable to trimming and can be trimmed by spraying pesticides and pruning to ensure the normal growth of these saplings. The saplings that exceed this specific threshold are subjected to the next segmentation process; S4: The measured point cloud data is segmented by the point cloud automatic segmentation system, and the saplings to be isolated are heavily marked and located, facilitating the user to perform the next removal process on these saplings.
2. The point cloud automatic segmentation method based on deep learning algorithms according to claim 1, characterized in that: The measurement methods described in S2 include ground three-dimensional laser scanning acquisition, vehicle-mounted MMS acquisition, hand-held laser scanning acquisition and UAV scanning acquisition.
3. The point cloud automatic segmentation method based on deep learning algorithms according to claim 1, characterized in that: The photogrammetry described in S2 can obtain a high-precision three-dimensional model by a series of calculations on the two-dimensional image, and we can also obtain point cloud data in this process.
4. The point cloud automatic segmentation method based on deep learning algorithms according to claim 1, characterized in that: The big data system described in S3 includes sapling height data, sapling thickness data and sapling production shape data.
5. The point cloud automatic segmentation method based on deep learning algorithms according to claim 1, characterized in that: The content of the marking and positioning described in S4 is displayed by polar coordinate positioning.