Three-dimensional laser point cloud context feature-based muck pile lumpiness calculation method

Abstract

The invention discloses a muck pile lumpiness calculation method based on three-dimensional laser point cloud context features, and relates to the technical field of smart mines, and the method comprises the steps: carrying out the preprocessing operation of muck pile point cloud data acquired by a three-dimensional laser scanner; based on the context features of the point cloud data in combination with a Euclidean clustering strategy, performing primary segmentation on the preprocessed point cloud data to obtain a first segmentation result; performing secondary segmentation on the first segmentation result by adopting a dynamic parameter optimization algorithm to obtain a second segmentation result; and obtaining a muck pile lumpiness calculation result according to the second segmentation result and the maximum point cloud distance. According to the method, high-precision three-dimensional scene information is obtained by utilizing a laser scanning technology, and accurate calculation of the blasting storage yard lumpiness is realized by combining a context feature analysis method, so that the precision and efficiency of blasting pile lumpiness calculation are improved, and reliable data support is provided for blasting engineering.

Description

Technical Field

[0001] The present invention relates to the field of smart mining technology, and more specifically to a method for calculating the blockiness of explosive piles based on context features of three-dimensional laser point clouds. Background Art

[0002] In open-pit mining production activities, the size of the blast pile not only directly determines the efficiency of blasting, but also has a significant impact on the mining and subsequent loading operations of the mine, providing important data support for adjusting blasting parameters and controlling mining costs, and thus playing a key role in improving the economic benefits of the mine. Traditional methods for calculating the size of blast piles include manual measurement and screening methods. The manual measurement method relies on manual use of tools (such as rulers and measuring tools) to measure the rock blocks in the blast pile one by one, record their sizes, and classify and count them. The screening method is to screen the blast pile rocks through sieves with different apertures, and divide the size levels according to the aperture size of the rock blocks through the sieve. Both the manual measurement method and the screening method have problems such as large workload, low efficiency, and inaccurate statistics. In recent years, with the popularization of 3D laser scanners and the development of related industry technologies. The application of 3D laser scanning technology has improved the measurement efficiency in the field of mine measurement, especially in blasting design and size measurement, improving efficiency, reducing costs, and reducing environmental impact.

[0003] Therefore, how to provide a method for calculating the blast pile block size based on the contextual features of a three-dimensional laser point cloud to efficiently, economically and accurately calculate the blast pile block size of an open-pit mine is an urgent problem that needs to be solved by those skilled in the art. Summary of the invention

[0004] In view of this, the present invention provides a method for calculating the fragmentation of blasting piles based on the contextual features of three-dimensional laser point clouds, which solves the problems of low accuracy and low efficiency in the traditional method of calculating the fragmentation of blasting yards. By using laser scanning technology to obtain high-precision three-dimensional scene information and combining it with the contextual feature analysis method, the fragmentation of blasting yards can be accurately calculated, thereby improving the accuracy and efficiency of the calculation of the fragmentation of blasting piles and providing reliable data support for blasting engineering.

[0005] In order to achieve the above object, the present invention adopts the following technical solution: a method for calculating the burst blockiness of a three-dimensional laser point cloud context feature, comprising:

[0006] Perform preprocessing operations on the point cloud data of the explosive pile acquired by the 3D laser scanner;

[0007] Based on the contextual features of the point cloud data and the Euclidean clustering strategy, the preprocessed point cloud data is segmented for the first time to obtain the first segmentation result;

[0008] A dynamic parameter optimization algorithm is used to perform a second segmentation on the first segmentation result to obtain a second segmentation result;

[0009] The calculation result of the burst blockiness is obtained according to the second segmentation result and the maximum distance of the point cloud.

[0010] Preferably, the contextual features of the point cloud data are based on a set of points in the point cloud data and their adjacent points, taking into account the spatial distribution, geometric shape and local density change attribute information of these points.

[0011] Preferably, the pre-processed point cloud data is initially segmented based on the contextual features of the point cloud data combined with the Euclidean clustering strategy, including:

[0012] Classify the preprocessed point cloud data into two categories: large ore and small ore;

[0013] If the point cloud data is identified as large ore, the small curvature feature is used to perform the initial segmentation processing on the point cloud data;

[0014] If the point cloud data is identified as small ore, the large curvature feature is used to perform initial segmentation processing on the point cloud data.

[0015] Preferably, the pre-processed point cloud data is classified into two categories: large ore and small ore, including:

[0016] Select multiple point cloud data of different scales to calculate normal vectors and local curvature features of point cloud data of different scales respectively, and perform feature aggregation to form a multi-scale feature descriptor;

[0017] Representative areas are selected from the point cloud data as training samples, and the multi-scale features of the training samples are mapped to the feature space. The classification model is trained in the feature space using a support vector machine. According to the preset blockiness requirements, the ores are divided into two categories: large ores and small ores.

[0018] Preferably, a dynamic parameter optimization algorithm is used to perform a second segmentation on the first segmentation result to obtain a second segmentation result, including:

[0019] Determine the volume of each ore in the point cloud data and record the volume of the largest ore;

[0020] Identify and determine the ores that require local optimization;

[0021] The ore that needs to be locally optimized is subjected to secondary segmentation processing.

[0022] Preferably, dynamic parameter optimization algorithm segmentation is performed by dynamically adjusting curvature features and density features in context features.

[0023] Preferably, the context features of the point cloud data are combined with a Euclidean clustering strategy, including:

[0024] The CANUPO algorithm is used to extract the semantic features of point cloud data;

[0025] The semantic features are combined with the curvature features in the context features to improve the Euclidean clustering segmentation algorithm.

[0026] It can be seen from the above technical scheme that, compared with the prior art, the present invention discloses a method for calculating the blockiness of a burst pile based on the contextual features of a three-dimensional laser point cloud, including: preprocessing the point cloud data of the burst pile acquired by a three-dimensional laser scanner; performing an initial segmentation on the preprocessed point cloud data based on the contextual features of the point cloud data combined with a Euclidean clustering strategy to obtain a first segmentation result; performing a secondary segmentation on the first segmentation result using a dynamic parameter optimization algorithm to obtain a second segmentation result; and obtaining a blockiness calculation result of the burst pile based on the second segmentation result and the maximum distance of the point cloud.

[0027] The present invention has the following beneficial effects:

[0028] (1) The present invention realizes the application of three-dimensional laser scanning technology in the analysis of blast pile fragmentation after mine blasting, and optimizes the segmentation of blast pile point cloud by combining the improved Euclidean clustering strategy, thereby improving the calculation accuracy of blast pile fragmentation and overcoming the shortcomings of traditional blast pile fragmentation calculation methods such as low accuracy and large workload;

[0029] (2) Compared with the traditional method for calculating the size of explosive piles, the method based on 3D laser scanning technology has the advantages of high safety, fast data acquisition, and high efficiency in calculating the size of explosive piles. At the same time, it solves the problem that the traditional 3D laser point cloud segmentation algorithm has low accuracy in identifying explosive pile ores, which leads to inaccurate calculation results of explosive pile size;

[0030] (3) The local point cloud is segmented again by using a segmentation algorithm that optimizes the dynamic parameters of the local point cloud of the blast pile, which improves the recognition accuracy of the blast pile ore and the calculation accuracy of the blast pile blockiness. BRIEF DESCRIPTION OF THE DRAWINGS

[0031] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, the drawings described below are only embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on the provided drawings without paying creative work.

[0032] Figure 1 A schematic flow chart of a method for calculating the burst blockiness of a three-dimensional laser point cloud context feature provided by the present invention.

[0033] Figure 2A schematic flow chart of a method for calculating the blockiness of an open-pit copper mine blast pile based on three-dimensional laser point cloud context features is provided in an embodiment of the present invention.

[0034] Figure 3 A schematic diagram of a method for segmenting a pile point cloud based on context features combined with a Euclidean clustering strategy provided in an embodiment of the present invention. DETAILED DESCRIPTION

[0035] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0036] The embodiment of the present invention discloses a method for calculating the blasting pile fragmentation based on the contextual features of a three-dimensional laser point cloud. By analyzing the fragmentation of ore after blasting through three-dimensional laser scanning technology, it can not only help improve the blasting efficiency, but also help reduce the cost of ore crushing and processing, thereby improving the ore utilization efficiency and making a positive contribution to environmental protection and resource utilization.

[0037] In a specific embodiment of the present invention, a method for calculating the burst blockiness of a three-dimensional laser point cloud context feature is provided. Figure 1 As shown, including:

[0038] Perform preprocessing operations on the point cloud data of the explosive pile acquired by the 3D laser scanner;

[0039] Based on the contextual features of point cloud data combined with the Euclidean clustering strategy, the preprocessed point cloud data is segmented for the first time to obtain the first segmentation result; the three-dimensional laser point cloud data of the explosion pile after the initial segmentation is used to improve the subsequent point cloud segmentation algorithm;

[0040] The dynamic parameter optimization algorithm is used to perform secondary segmentation on the first segmentation result to obtain the second segmentation result; based on the initially segmented explosive pile point cloud data, the recognition sensitivity of the adhesion part of the ore point cloud is enhanced to improve the segmentation effect of the explosive pile point cloud; the three-dimensional laser point cloud data of the explosive pile processed by the dynamic parameter optimization segmentation algorithm is used for the subsequent explosive pile block calculation;

[0041] The calculation result of the blast pile fragmentation is obtained according to the second segmentation result and the maximum distance of the point cloud. According to the second segmentation result, the fragmentation of the blast pile ore is calculated using the point cloud maximum distance algorithm, and the identification of the ore is judged, and the local ore identification error is optimized, and finally the blast pile point cloud with high recognition accuracy and the calculation result of the blast pile fragmentation are output.

[0042] Specifically, the preprocessing operations include processing methods such as registration and denoising; based on context features combined with the Euclidean clustering strategy, the preprocessed three-dimensional laser point cloud data of the explosion pile is effectively segmented, and the effectiveness and superiority of this segmentation method are verified.

[0043] Specifically, the blast pile block size is the blast pile block size of an open-pit copper mine.

[0044] Specifically, the contextual features of the point cloud data are based on a set of points in the point cloud data and their adjacent points, taking into account the spatial distribution, geometric shape and local density change attribute information of these points.

[0045] By analyzing the relative position, shape difference and closeness of each point in the point cloud with its surrounding points, the local and global structural information of the point cloud is revealed. On this basis, curvature features and density features are important components of context features, providing key information about the local shape changes of the point cloud surface and the degree of concentration of point clusters.

[0046] Specifically, based on the contextual features of the point cloud data and the Euclidean clustering strategy, the preprocessed point cloud data is initially segmented, including:

[0047] Classify the preprocessed point cloud data into two categories: large ore and small ore;

[0048] If the point cloud data is identified as a large ore, considering that its curvature difference is small, the small curvature feature is used to perform the initial segmentation processing on the point cloud data;

[0049] If the point cloud data is identified as small ore, considering its large curvature difference, the large curvature feature is used to perform initial segmentation processing on the point cloud data.

[0050] Specifically, the pre-processed point cloud data is classified into two categories: large ore and small ore, including:

[0051] Selecting point cloud data of multiple different scales (the point cloud data of different scales are determined according to the point cloud resolution), respectively calculating normal vectors and local curvature features of the point cloud data of different scales, and performing feature aggregation to form a multi-scale feature descriptor;

[0052] Representative areas are selected from the point cloud data as training samples, and the multi-scale features of the training samples are mapped to the feature space. The classification model is trained in the feature space using a support vector machine. According to the preset blockiness requirements, the ores are divided into two categories: large ores and small ores.

[0053] Based on the contextual features of point cloud data and the Euclidean clustering strategy, the preprocessed point cloud data is classified into two categories: large and small. Appropriate curvature segmentation conditions are selected according to different categories to improve the accuracy of point cloud segmentation.

[0054] Specifically, the preset block size requirement is the block size requirement allowed by the excavator bucket.

[0055] Specifically, a dynamic parameter optimization algorithm is used to perform a second segmentation on the first segmentation result to obtain a second segmentation result, including:

[0056] Before local segmentation optimization, the volume of each ore in the point cloud data is measured, and the volume of the largest ore in the point cloud data is manually recorded; this is used as the basis for automatic processing;

[0057] Identify and determine ores that require local optimization due to sticking;

[0058] The ore that needs to be locally optimized is subjected to secondary segmentation processing.

[0059] Specifically, dynamic parameter optimization algorithm segmentation is performed by dynamically adjusting the curvature features and density features in the context features.

[0060] For areas where the local point cloud recognition effect of the explosion pile is poor, a dynamic parameter optimization algorithm is used to perform secondary segmentation of the point cloud. By dynamically adjusting the curvature features and density features in the context features, effective segmentation of the local point cloud of the explosion pile is achieved.

[0061] In a specific embodiment of the present invention, the size of the ore segmentation result is calculated before local improvement. Then, the size of the largest ore in the blast pile point cloud area is manually measured and compared with the size of each ore in the point cloud to determine whether there is ore with adhesion. Finally, for the ore point cloud that needs to optimize the recognition result, this part of the point cloud is segmented again by adjusting a finer curvature threshold and density threshold to optimize the accuracy of ore recognition.

[0062] The embodiment of the present invention performs dynamic parameter optimization by setting a smaller curvature feature threshold and adding a density feature threshold. The main purpose of the dynamic parameter optimization method is to solve the following problem: in different blast piles, the different stacking conditions of the ore will lead to the diversity of the normal vector angles of the ore contact position. The single discriminant segmentation based solely on the normal vector angle cannot be well adapted to different blast pile point cloud data, and the segmentation algorithm will cause the local ore point cloud to stick together in some blast pile areas.

[0063] Specifically, the contextual features of point cloud data are combined with the Euclidean clustering strategy, including:

[0064] The CANUPO algorithm is used to extract the semantic features of point cloud data; thereby obtaining the extracted semantic feature results; the extracted semantic features can well classify the blast pile into two categories: large ore and small ore.

[0065] The semantic features are combined with the curvature features in the context features to improve the Euclidean clustering segmentation algorithm. The accuracy of the point cloud recognition of the explosive pile is improved. The extraction of the curvature features in the context features aims to understand the geometric shape of the surface of the explosive pile, which can help distinguish different areas in the explosive pile and provide data support for the subsequent calculation of the explosive pile blockiness.

[0066] The embodiments of the present invention use point cloud context information to accurately understand the global structure, shape, and interrelationships between objects in point cloud data, and are the key to improving point cloud processing and analysis performance. Therefore, the calculation of blast pile block size based on point cloud context features is crucial for optimizing open-pit mining.

[0067] Specifically, the calculation result of the burst blockiness is obtained according to the second segmentation result and the maximum distance of the point cloud, including:

[0068] First, the 3D laser point cloud of the blast pile is divided into several separate ore point clouds through point cloud segmentation technology. Then, the principle of Euclidean distance is applied to measure the size of these ore point clouds. The final stage of block size analysis is based on the results of ore identification and size measurement in the first two steps, and statistical analysis is performed, such as calculating the size information of the ore and its distribution and other statistical data.

[0069] The embodiment of the present invention calculates the fragmentation of the blast pile ore through the point cloud maximum distance algorithm based on the locally improved parameter adaptive method, judges the identification status of the ore, optimizes the situation of local ore identification errors, and finally outputs the blast pile point cloud with high identification accuracy and the calculation result of the blast pile fragmentation.

[0070] In a specific embodiment of the present invention, Figure 2 As shown, the three-dimensional laser scanner selects RIEGL_VZ1000 three-dimensional laser scanner to collect blast pile data, and the collection area is the copper factory and Fujiawu mining areas of Dexing Copper Mine. Then, the point cloud data preprocessing operation is performed on the collected blast pile data, and finally the preprocessed blast pile point cloud data is obtained.

[0071] like Figure 3As shown in the figure, the explosive pile point cloud segmentation based on context features combined with Euclidean clustering strategy is performed according to the preprocessed point cloud data. The steps are as follows: first, the explosive pile point cloud is manually cropped to obtain large ore point cloud samples and small ore point cloud samples to obtain training samples; then, the point cloud features of the training samples at different scales are analyzed, and a training classifier is constructed on this basis. The best classification scale combination is found according to the features of the training samples and the visual interpretation results, and the optimal classification hyperplane is generated; finally, a classification label is assigned to each point in the point cloud by voting, so as to classify the explosive pile point cloud based on the optimal classification hyperplane and extract the classification information of the explosive pile point cloud data.

[0072] According to the classification results of the CANUPO algorithm, the ores in the blast area are roughly classified into small ores and large ores. In the two types of ore point clouds, the curvature difference of small ores is large, and the curvature difference of large ores is relatively small. Therefore, under the traversal search condition of the original Euclidean clustering (searching neighborhood points according to the Euclidean distance threshold), for the points to be classified, the classification judgment basis is added: the difference in multi-scale-dimensional features is judged according to the Euclidean distance threshold. If the multi-scale-dimensional feature difference of the point is relatively small, the original Euclidean segmentation is performed for classification; if the multi-scale-dimensional feature difference of the point is large, the curvature condition is added to the original Euclidean segmentation method for classification.

[0073] In view of the situation that some blast pile areas have local segmentation errors, it is necessary to add a secondary segmentation algorithm. First, the ore segmentation results are calculated before local improvement, and the ore size is calculated according to the Euclidean space distance. Then, the size of the largest ore in the blast pile point cloud area is manually measured. By comparing the size of each ore and the size of the largest ore in the blast pile area, it is possible to determine the ores with adhesion. This is because the connection points of small ores are densely present and are mistakenly identified as part of the same ore during the nearest neighbor search process. Therefore, it is necessary to perform local optimization of the ore identification results. Finally, for the blast pile point cloud whose identification results need to be optimized, the normal vector angle threshold in the algorithm is re-determined and the segmentation algorithm with dynamic parameter optimization is used to perform secondary segmentation on the local point cloud of the blast pile to obtain the final identification result of the blast pile ore.

[0074] A new blast pile point cloud segmentation method provided in an embodiment of the present invention can effectively identify, segment and estimate the size of ore after blasting, so as to improve the accuracy of blast pile block calculation in open-pit copper mines.

[0075] In this specification, each embodiment 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. For the device disclosed in the embodiment, since it corresponds to the method disclosed in the embodiment, the description is relatively simple, and the relevant parts can be referred to the method part.

[0076] The above description of the disclosed embodiments enables one skilled in the art to implement or use the present invention. Various modifications to these embodiments will be apparent to one skilled in the art, and the general principles defined herein may be implemented in other embodiments without departing from the spirit or scope of the present invention. Therefore, the present invention will not be limited to the embodiments shown herein, but rather to the widest scope consistent with the principles and novel features disclosed herein.

Claims

1. A method for calculating the burst blockiness of a three-dimensional laser point cloud context feature, characterized in that: include: Perform preprocessing operations on the point cloud data of the explosive pile acquired by the 3D laser scanner; Based on the contextual features of the point cloud data and the Euclidean clustering strategy, the preprocessed point cloud data is segmented for the first time to obtain the first segmentation result; A dynamic parameter optimization algorithm is used to perform a second segmentation on the first segmentation result to obtain a second segmentation result; The calculation result of the burst blockiness is obtained according to the second segmentation result and the maximum distance of the point cloud.

2. The method for calculating the burst blockiness of a three-dimensional laser point cloud context feature according to claim 1, characterized in that: The contextual features of the point cloud data are based on a set of points in the point cloud data and their adjacent points, taking into account the spatial distribution, geometric shape and local density change attribute information of these points.

3. The method for calculating the burst blockiness of a 3D laser point cloud context feature according to claim 1, characterized in that: Based on the contextual features of the point cloud data and the Euclidean clustering strategy, the preprocessed point cloud data is initially segmented, including: Classify the preprocessed point cloud data into two categories: large ore and small ore; If the point cloud data is identified as large ore, the small curvature feature is used to perform the initial segmentation processing on the point cloud data; If the point cloud data is identified as small ore, the large curvature feature is used to perform initial segmentation processing on the point cloud data.

4. The method for calculating the burst blockiness of a 3D laser point cloud context feature according to claim 2, characterized in that: The pre-processed point cloud data is classified into two categories: large ore and small ore, including: Select multiple point cloud data of different scales, calculate the normal vector and curvature local features of the point cloud data of different scales respectively, and perform feature aggregation to form a multi-scale feature descriptor; Representative areas are selected from the point cloud data as training samples, and the multi-scale features of the training samples are mapped to the feature space. The classification model is trained in the feature space using a support vector machine. According to the preset blockiness requirements, the ores are divided into two categories: large ores and small ores.

5. The method for calculating the burst blockiness of a 3D laser point cloud context feature according to claim 1, characterized in that: The first segmentation result is segmented again using a dynamic parameter optimization algorithm to obtain a second segmentation result, including: Determine the volume of each ore in the point cloud data and record the volume of the largest ore; Identify and determine the ores that require local optimization; The ore that needs to be locally optimized is subjected to secondary segmentation processing.

6. The method for calculating the burst blockiness of a 3D laser point cloud context feature according to claim 1, characterized in that: Dynamic parameter optimization algorithm segmentation is performed by dynamically adjusting the curvature features and density features in the context features.

7. The method for calculating the burst blockiness of a 3D laser point cloud context feature according to claim 1, characterized in that: The contextual features of point cloud data are combined with the Euclidean clustering strategy, including: The CANUPO algorithm is used to extract the semantic features of point cloud data; The semantic features are combined with the curvature features in the context features to improve the Euclidean clustering segmentation algorithm.

Images