Point cloud semantic segmentation method and system based on voxel cohesion guided active learning

Abstract

The invention provides a point cloud semantic segmentation method and system based on voxel cohesion guided active learning. The method comprises the steps of obtaining target three-dimensional point cloud data corresponding to a scene to be detected; inputting the target three-dimensional point cloud data into a point cloud semantic segmentation model, and obtaining a semantic segmentation category result corresponding to the target three-dimensional point cloud data output by the point cloud semantic segmentation model, the point cloud semantic segmentation model is obtained by training sample three-dimensional point cloud data marked with semantic segmentation category labels, and the sample three-dimensional point cloud data is point cloud data in a sample voxel grid determined based on voxel condensation degree; the voxel condensation degree is determined according to the number of data points corresponding to the point cloud data points of different semantic segmentation category labels in the sample voxel grid. According to the invention, the accuracy and generalization ability of a point cloud semantic segmentation result are improved.

Description

Technical Field

[0001] The present invention relates to the field of computer vision technology, and in particular to a point cloud semantic segmentation method and system based on voxel cohesion-guided active learning. Background Art

[0002] Understanding three-dimensional scenes is an important research direction in modern computer vision and intelligent systems, especially in accurate recognition and decision-making tasks in complex environments. Comprehensive analysis of the structure, shape and position of objects in three-dimensional scenes can provide basic support for research in autonomous driving, robot navigation and virtual reality.

[0003] Point cloud semantic segmentation is one of the core technologies for achieving 3D scene understanding. It aims to assign specific semantic labels to each point in the point cloud data, thereby achieving comprehensive perception and classification of the scene. Existing point cloud semantic segmentation methods mainly rely on data-driven deep learning methods to train high-precision segmentation models through a large number of labeled data sets. However, the sparsity, irregularity and high labeling cost of point cloud data have greatly limited the performance of the segmentation model, resulting in limited promotion of the model to practical applications.

[0004] Therefore, there is an urgent need for a point cloud semantic segmentation method and system based on voxel cohesion guided active learning to solve the above problems. Summary of the invention

[0005] In view of the problems existing in the prior art, the present invention provides a point cloud semantic segmentation method and system based on voxel cohesion guided active learning.

[0006] The present invention provides a point cloud semantic segmentation method based on voxel cohesion guided active learning, comprising: Obtain target three-dimensional point cloud data corresponding to the scene to be detected; The target three-dimensional point cloud data is input into a point cloud semantic segmentation model to obtain a semantic segmentation category result corresponding to the target three-dimensional point cloud data output by the point cloud semantic segmentation model, wherein the point cloud semantic segmentation model is trained by sample three-dimensional point cloud data marked with semantic segmentation category labels, and the sample three-dimensional point cloud data is point cloud data in a sample voxel grid determined based on voxel cohesion, and the voxel cohesion is determined according to the number of data points corresponding to point cloud data points with different semantic segmentation category labels in the sample voxel grid.

[0007] According to a point cloud semantic segmentation method based on voxel cohesion guided active learning provided by the present invention, the point cloud semantic segmentation model is trained by the following steps: Determine, according to the feature map corresponding to the sample laser radar point cloud data, the semantic category label of each point cloud data point in the sample laser radar point cloud data; voxelize the sample laser radar point cloud data to obtain a plurality of pending voxel grids, wherein the data points in the pending voxel grids are determined based on the coordinate information of the point cloud data points in the sample laser radar point data and a preset voxel size, and the voxel point category labels of the data points in the pending voxel grids correspond to the semantic category labels of the point cloud data points in the sample laser radar point data; Calculating the voxel cohesion of the undetermined voxel grid according to the uncertainty score of the undetermined voxel grid, wherein the uncertainty score is calculated based on the number of data points of each different voxel point category label in the voxel grid and the number of all data points in the voxel grid; Based on the voxel cohesion of each of the to-be-determined voxel grids, determining the to-be-determined voxel grid with the lowest voxel cohesion as the sample voxel grid; Marking the point cloud data points corresponding to each data point in the sample voxel grid in the sample lidar point data to obtain the sample three-dimensional point cloud data marked with the semantic segmentation category label; Based on the sample three-dimensional point cloud data, a preset three-dimensional neural network model is trained to obtain the point cloud semantic segmentation model.

[0008] According to a point cloud semantic segmentation method based on voxel cohesion guided active learning provided by the present invention, the voxel cohesion of the undetermined voxel grid is calculated according to the uncertainty score of the undetermined voxel grid, including: Based on the voxel cohesion calculation formula, the voxel cohesion of the to-be-determined voxel grid is calculated according to the uncertainty score of the to-be-determined voxel grid. The voxel cohesion calculation formula is:

[0009] in, Indicates The first The number of data points with voxel class labels, Indicates A grid of undetermined voxels, represents a collection of voxel grids, Indicates The first voxel in the undetermined voxel grid The predicted category corresponding to the data point, Represents the total number of voxel point category labels, Indicates The voxel cohesion corresponding to the undetermined voxel grid is Indicates The number of all data points in the undetermined voxel grid, Indicates The uncertainty score corresponding to the undetermined voxel grid, Represents the total number of data point categories within the pending voxel grid.

[0010] According to a point cloud semantic segmentation method based on voxel cohesion guided active learning provided by the present invention, the point cloud data points corresponding to each data point in the sample voxel grid in the sample lidar point data are marked to obtain the sample three-dimensional point cloud data marked with the semantic segmentation category label, including: According to the preset perception distance, each data point in the sample voxel grid is judged. If there is a key data point within the preset perception distance with any data point to be marked as the center, a feature difference value is calculated according to the feature vectors corresponding to the data point to be marked and the key data point, wherein the key data point is a data point in the sample voxel grid that has been determined to be a point cloud data point to be marked; When it is determined that the feature difference value is less than or equal to a preset feature difference threshold, the data points to be marked are removed to obtain a sample voxel grid after redundant data points are removed; Marking the point cloud data points corresponding to each data point of the sample voxel grid after the redundant data points are removed in the sample lidar point data to obtain the sample three-dimensional point cloud data marked with the semantic segmentation category label; When it is determined that the feature difference value is greater than a preset feature difference threshold, or when the key data point does not exist within the preset perception distance centered on the data point to be marked, the data point to be marked is retained as the key data point.

[0011] According to a point cloud semantic segmentation method based on voxel cohesion guided active learning provided by the present invention, the method further includes: Sort the uncertainty scores in descending order, and sort the top scores in the sorting results. The undetermined voxel grids corresponding to the uncertain scores are used as the category coordination voxel grids; Acquire pseudo label data of each data point in the category coordination voxel grid and a preset voxel point category weight corresponding to the pseudo label data; Obtaining a data point uncertainty score corresponding to each data point in the category-coordinated voxel grid according to a product result between the uncertainty score of the category-coordinated voxel grid and the preset voxel point category weight; Sort the uncertainty scores of the data points in descending order, and put the sorting results first The data points corresponding to the uncertain scores of the data points are taken as the data points to be marked; Mark the point cloud data points corresponding to the data points to be marked in the sample laser radar point data.

[0012] According to a point cloud semantic segmentation method based on voxel cohesion guided active learning provided by the present invention, the preset three-dimensional neural network model is trained based on the sample three-dimensional point cloud data to obtain the point cloud semantic segmentation model, including: Based on the sample three-dimensional point cloud data and the cross entropy loss function, the preset three-dimensional neural network model is trained to obtain the point cloud semantic segmentation model.

[0013] The present invention also provides a point cloud semantic segmentation system based on voxel cohesion guided active learning, comprising: Point cloud data acquisition module, used to obtain target three-dimensional point cloud data corresponding to the scene to be detected; A point cloud semantic segmentation module is used to input the target three-dimensional point cloud data into a point cloud semantic segmentation model to obtain a semantic segmentation category result corresponding to the target three-dimensional point cloud data output by the point cloud semantic segmentation model, wherein the point cloud semantic segmentation model is trained by sample three-dimensional point cloud data marked with semantic segmentation category labels, and the sample three-dimensional point cloud data is point cloud data in a sample voxel grid determined based on voxel cohesion, and the voxel cohesion is determined according to the number of data points corresponding to point cloud data points with different semantic segmentation category labels in the sample voxel grid.

[0014] The present invention also provides an electronic device, comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein when the processor executes the program, a point cloud semantic segmentation method based on voxel cohesion guided active learning as described above is implemented.

[0015] The present invention also provides a non-transitory computer-readable storage medium having a computer program stored thereon. When the computer program is executed by a processor, the method for point cloud semantic segmentation based on voxel cohesion-guided active learning as described in any one of the above is implemented.

[0016] The present invention also provides a computer program product, comprising a computer program, which, when executed by a processor, implements any of the above-mentioned point cloud semantic segmentation methods based on voxel cohesion-guided active learning.

[0017] The point cloud semantic segmentation method and system based on voxel cohesion guided active learning provided by the present invention determine the point cloud data in the sample voxel grid through voxel cohesion, and then perform point cloud semantic segmentation processing based on the point cloud semantic segmentation model trained with these point cloud data, thereby improving the accuracy and generalization ability of the point cloud semantic segmentation results. BRIEF DESCRIPTION OF THE DRAWINGS

[0018] In order to more clearly illustrate the technical solutions in the present invention or 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 some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.

[0019] Figure 1 A schematic diagram of the process flow of the point cloud semantic segmentation method based on voxel cohesion guided active learning provided by the present invention; Figure 2 A schematic diagram of a point cloud semantic segmentation framework based on active learning provided by the present invention; Figure 3 A schematic diagram of the structure of a point cloud semantic segmentation system based on voxel cohesion guided active learning provided by the present invention; Figure 4 This is a schematic structural diagram of an electronic device provided by the present invention. DETAILED DESCRIPTION

[0020] In order to make the purpose, technical solution and advantages of the present invention clearer, the technical solution of the present invention will be clearly and completely described below in conjunction with the drawings of the present invention. Obviously, the described embodiments are 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.

[0021] Point cloud semantic segmentation technology plays a key role in many fields, such as environmental perception in autonomous driving, path planning in robot navigation, scene reconstruction in virtual reality, building information modeling and material classification. However, the sparsity, irregularity and high labeling cost of point cloud data greatly limit the performance of segmentation models, resulting in limited promotion of models to practical applications.

[0022] With the development of deep learning, many point cloud semantic segmentation methods have been proposed, which can be roughly divided into the following categories: 1. Method based on 2D projection: Project the 3D point cloud into a 2D image through multi-view, bird's-eye view and spherical projection, and then use the designed 2D convolutional neural network to analyze the 3D scene. Although this method can give full play to the advantages of 2D convolutional networks, the 3D geometric information will inevitably be lost during the projection process, which limits the segmentation performance.

[0023] 2. Voxel-based methods: By voxelizing point clouds, irregular and unevenly dense point clouds are converted into regular voxel grids, so that two-dimensional convolutional networks can be extended to three-dimensional data analysis. Early voxel-based methods often faced problems of high computational complexity and information loss. With the introduction of sparse convolution, the computational cost has been greatly reduced, improving the efficiency of voxel methods.

[0024] 3. Point-based methods: This type of method directly processes the original, uneven point cloud data, and uses flexible methods such as multilayer perceptron (MLP) and 3D point convolution to process the points, which maximizes the original 3D information of the point cloud. However, although this method can retain rich geometric information, its performance depends on a large amount of labeled data, which poses a challenge to practical applications.

[0025] Compared with indoor scenes, outdoor LiDAR point clouds have higher requirements on the efficiency of segmentation networks due to their large scale, uneven density and sparseness. To this end, many existing studies have improved processing efficiency by projecting 3D point clouds onto 2D images, such as spherical views (range images) or bird's-eye views, or by fusing different views. In addition, there are also studies that directly process point cloud data, aiming to build models that process irregular data more efficiently. These methods can all be used as backbone networks in active learning frameworks.

[0026] Active learning can significantly reduce labeling costs while ensuring model performance close to fully supervised learning by selectively labeling the most informative samples. In particular, in some application scenarios with moderate accuracy requirements and limited labeling budgets, active learning strategies can prioritize samples with higher uncertainty for labeling, thereby achieving an effective balance between cost and accuracy. Therefore, active learning has shown great potential and practicality in the task of semantic segmentation of point clouds in 3D scene understanding, especially in resource-constrained scenarios.

[0027] To cope with the difficulty of labeling large-scale point cloud data, active learning methods gradually select and request the network to label some samples, such as scans, regions, points, or boxes, during the training process. Most selection strategies are based on uncertainty or diversity principles, for example, uncertainty sampling, which selects the most challenging samples by measuring the uncertainty of the model's prediction score (such as softmax output) for each point, or the gap between the two highest scores.

[0028] Existing active learning methods usually select samples based on uncertainty and diversity, but these methods are difficult to fully capture the geometric and topological features when directly processing 3D point cloud data. In particular, when processing large-scale outdoor LiDAR point clouds, there are still problems of low efficiency or complex preprocessing, which cannot meet the actual needs of some scenes. In addition, the strong correlation between point cloud data samples makes it difficult for existing active learning strategies to achieve ideal results, resulting in poor accuracy of the final segmentation results.

[0029] In view of the problems existing in the above-mentioned prior art, the present invention proposes a point cloud semantic segmentation method based on voxel cohesion-guided active learning. When training the point cloud semantic segmentation model, the method first combines the local and global geometric features of the point cloud data to voxelize the point cloud, and measures its information complexity by calculating the entropy value of each voxel block, and records it as the voxel cohesion to screen out the voxel set with the highest uncertainty. Secondly, by measuring the similarity relationship between data points in each voxel grid in the set, accurate filtering of redundant information points in the voxel is achieved. Finally, an online category coordination module is designed to optimize the threshold setting according to the data density and labeling requirements of the category, and dynamically adjust the number of candidate points of each category during the labeling process, thereby reducing the negative impact of category imbalance on the model training effect and improving labeling efficiency and data utilization.

[0030] Figure 1 A flow chart of the point cloud semantic segmentation method based on voxel cohesion guided active learning provided by the present invention is shown in FIG. Figure 1 As shown, the present invention provides a point cloud semantic segmentation method based on voxel cohesion guided active learning, comprising: Step 101: Obtain target three-dimensional point cloud data corresponding to the scene to be detected.

[0031] In the present invention, laser radar scanning is used to emit a laser beam and measure the time it takes for it to reflect back to calculate the distance, thereby obtaining the three-dimensional coordinates of objects in the scene to obtain three-dimensional point cloud data in the scene. Three-dimensional point cloud data is a collection of a series of points in space, each point usually contains three-dimensional coordinates (X, Y, Z), and these points together describe the shape, position and surface characteristics of the object.

[0032] Step 102: input the target three-dimensional point cloud data into a point cloud semantic segmentation model to obtain a semantic segmentation category result corresponding to the target three-dimensional point cloud data output by the point cloud semantic segmentation model, wherein the point cloud semantic segmentation model is trained by sample three-dimensional point cloud data marked with semantic segmentation category labels, and the sample three-dimensional point cloud data is point cloud data in a sample voxel grid determined based on voxel cohesion, and the voxel cohesion is determined according to the number of data points corresponding to point cloud data points with different semantic segmentation category labels in the sample voxel grid.

[0033] Point cloud semantic segmentation refers to classifying each point in the point cloud data or the object to which each point belongs into a specific semantic category, such as cars, trees, buildings, and pedestrians. In the present invention, the input of the model is the target three-dimensional point cloud data, that is, the point cloud data set that needs to be semantically segmented, and the output is the semantic segmentation category results corresponding to these point cloud data, that is, the semantic label assigned to each point or point set.

[0034] In the present invention, the point cloud semantic segmentation model is trained by learning sample three-dimensional point cloud data marked with semantic segmentation category labels, which sample data contains rich scene and object information, and each point or point set is correctly labeled with a semantic category.

[0035] During the training process, the model learns how to extract features from the input point cloud data through an iterative optimization algorithm (such as gradient descent), and classifies the points into the correct semantic categories based on these features. Specifically, in order to process large-scale point cloud data, the point cloud data is organized into a voxel grid, which is the smallest unit in three-dimensional space, similar to a pixel in a two-dimensional image. The three-dimensional space is divided into a series of small cubes through the voxel grid, and each cube can contain a certain number of points.

[0036] Furthermore, in the present invention, the voxel cohesion is determined based on the number of data points corresponding to the point cloud data points of different semantic segmentation category labels in the sample voxel grid, that is, the number (or density) of data points in the voxel grid and the distribution of the semantic categories to which these data points belong, which can reflect the "cohesion" of the voxel. Voxels with high cohesion may contain more or more concentrated points of a certain type. Therefore, based on the voxel cohesion, more representative point cloud data in the sample voxel grid can be selected as training samples, which helps the model learn more accurate features and classification rules.

[0037] After the model is trained, new, unlabeled target 3D point cloud data can be input into the model. The model will process the point cloud data and output the semantic segmentation category results corresponding to each point or point set.

[0038] The point cloud semantic segmentation method based on voxel cohesion guided active learning provided by the present invention determines the point cloud data in the sample voxel grid through voxel cohesion, and then performs point cloud semantic segmentation processing based on the point cloud semantic segmentation model trained with these point cloud data, thereby improving the accuracy and generalization ability of the point cloud semantic segmentation results.

[0039] Based on the above embodiment, the point cloud semantic segmentation model is trained by the following steps: Determine, according to the feature map corresponding to the sample laser radar point cloud data, the semantic category label of each point cloud data point in the sample laser radar point cloud data; voxelize the sample laser radar point cloud data to obtain a plurality of pending voxel grids, wherein the data points in the pending voxel grids are determined based on the coordinate information of the point cloud data points in the sample laser radar point data and a preset voxel size, and the voxel point category labels of the data points in the pending voxel grids correspond to the semantic category labels of the point cloud data points in the sample laser radar point data; Calculating the voxel cohesion of the undetermined voxel grid according to the uncertainty score in the undetermined voxel grid, wherein the uncertainty score is calculated based on the number of data points of each different voxel point category label in the voxel grid and the number of all data points in the voxel grid; Based on the voxel cohesion of each of the to-be-determined voxel grids, determining the to-be-determined voxel grid with the lowest voxel cohesion as the sample voxel grid; Marking the point cloud data points corresponding to each data point in the sample voxel grid in the sample lidar point data to obtain the sample three-dimensional point cloud data marked with the semantic segmentation category label; Based on the sample three-dimensional point cloud data, a preset three-dimensional neural network model is trained to obtain the point cloud semantic segmentation model.

[0040] Figure 2 The schematic diagram of the point cloud semantic segmentation framework based on active learning provided by the present invention can be referred to Figure 2As shown, in the present invention, a sparse point voxel convolutional network (SPVCNN) is used as a three-dimensional backbone network in a preset three-dimensional neural network for training. First, the present invention randomly selects a portion of points in the sample lidar point cloud as initial marker points, and these marker points are used to guide subsequent model training. Then, the point cloud data is input into the three-dimensional backbone network SPVCNN for training, and the feature map of each point is extracted. After the feature extraction is completed, the input point cloud is voxelized. After voxelization, the present invention selects voxel grids based on the voxel cohesion metric. Preferably, the present invention provides a voxel cohesion and redundancy filtering module (Voxel Cohesion Measurement and Redundancy Filtering Module, referred to as VCMRF), which evaluates its uncertainty by measuring the cohesion within each voxel grid, and accurately filters out redundant information points based on similarity calculation, retaining representative key points. Optionally, the filtered point cloud data will be processed by the Online Adaptive Category Coordination Module (ACCM), which dynamically adjusts the number of points of each category in the candidate pool to balance the representativeness and quantity distribution of each category in the dataset. Finally, the points in the processed candidate pool will be marked online, and the marked data points will be updated to the training set for the next iteration to further optimize the performance and accuracy of the model.

[0041] In the present invention, sample LiDAR point cloud data are obtained, which represent sampling points of the surface of an object or environment in three-dimensional space. Then, a semantic category label, such as "vehicle", "pedestrian", "tree", etc., is assigned to each point cloud data point through manual labeling or prediction based on other trained models. In the present invention, feature maps of the sample LiDAR point cloud data can be extracted through a three-dimensional backbone network to obtain semantic category labels for each point cloud data point in the sample LiDAR point cloud data.

[0042] Furthermore, the entire point cloud data space is divided into small three-dimensional cubes, called voxel grids. The size of each voxel grid is preset, and the data points in the voxel grid are assigned to corresponding voxels according to the coordinate information of the point cloud data points in the point cloud data space. In the present invention, the data points in each voxel grid inherit the semantic category label of any point cloud data point it contains, so the voxel grid may contain multiple points of different categories, or all of them may belong to the same category.

[0043] In the present invention, for each voxel grid, an uncertainty score is calculated, which reflects the degree of inconsistency of the category labels of the point cloud data points within the voxel grid. The uncertainty score can be calculated based on the ratio of the number of data points with different category labels within the voxel grid to the total number of data points in the voxel grid.

[0044] Next, this uncertainty score is used to calculate the cohesion of the voxel grid. A lower cohesion means that the category labels of the point cloud data points in the voxel grid are more diverse, and may contain object parts or edges of multiple different categories. Furthermore, the voxel grid with the lowest voxel cohesion is selected from all the pending voxel grids as the sample voxel grid. In the present invention, all point cloud data points belonging to the sample voxel grid are labeled to obtain a sample three-dimensional point cloud data set with accurate semantic segmentation labels. Then, the preset three-dimensional neural network model is trained using this labeled sample three-dimensional point cloud data. During the training process, the model learns how to predict the semantic category of each point based on the spatial distribution and characteristics of the point cloud data. After sufficient iterations and optimization, a model that can accurately perform point cloud semantic segmentation, namely a point cloud semantic segmentation model, is finally obtained. The model can understand and interpret complex three-dimensional environments, thereby supporting more advanced functions such as path planning and obstacle avoidance for self-driving cars.

[0045] On the basis of the above embodiment, the voxel cohesion degree of the to-be-determined voxel grid is calculated according to the uncertainty score of the to-be-determined voxel grid, including: Based on the voxel cohesion calculation formula, the voxel cohesion of the to-be-determined voxel grid is calculated according to the uncertainty score of the to-be-determined voxel grid. The voxel cohesion calculation formula is:

[0046] in, Indicates The first The number of data points with voxel class labels, Indicates A grid of undetermined voxels, represents a collection of voxel grids, Indicates The first voxel in the undetermined voxel grid The predicted category corresponding to the data point, Represents the total number of voxel point category labels, Indicates The voxel cohesion corresponding to the undetermined voxel grid is Indicates The number of all data points in the undetermined voxel grid, Indicates The uncertainty score corresponding to the undetermined voxel grid, Represents the total number of data point categories within the pending voxel grid.

[0047] In the present invention, it is necessary to select and label the point cloud data points with the highest value from the sample lidar point cloud data to maximize the model performance, that is, identifying and using the labeled data with the most information for training is the key to obtaining a high-quality model, thereby achieving accurate segmentation prediction.

[0048] Specifically, the present invention first voxelizes the sample lidar point cloud data and converts it into a three-dimensional voxel grid set. ,The process involves sampling the continuous 3D spatial input into a discrete voxel grid, and points falling within the same grid are merged. Each voxel grid is regarded as an independent selection unit.

[0049] For sample lidar point cloud data The Point cloud data points , accordingly, the point cloud data points in the voxel grid coordinates are: ; in, Represents a predefined voxel grid size. Point cloud data points As a point in three-dimensional space, its coordinates are expressed as . is a point cloud data point Coordinates in the voxel grid. In the present invention, the point cloud data points Each coordinate component of , , Divide by the preset voxel size Δ respectively, and then round down (that is, take the largest integer not greater than this number), so as to divide the continuous three-dimensional space into discrete voxel grids.

[0050] Furthermore, for each voxel grid , evaluate its semantic cohesion, i.e., voxel cohesion. Since point cloud data usually exhibits local semantic similarity, it is assumed that points in the same voxel grid have similar semantics. Therefore, the present invention evaluates the importance of each voxel grid by voxel cohesion. The number of data points corresponding to the category label of each voxel point in each voxel grid The calculation process is as follows:

[0051] The uncertainty score of the undetermined voxel grid is calculated as follows:

[0052] in, Indicates Uncertainty score for the voxel grid.

[0053] The voxel cohesion calculation formula is:

[0054] In the present invention, the voxel cohesion of the voxel grid is calculated by the uncertainty score of the voxel grid. In each 3D point cloud scan, the voxel grid with the lowest voxel cohesion is selected. The lower cohesion indicates that the category diversity within the voxel is greater, and the labeling of these voxels contributes more to the model training, so that the model is trained or the subsequent model is updated according to these selected voxel grids.

[0055] On the basis of the above embodiment, the step of marking the point cloud data points corresponding to each data point in the sample voxel grid in the sample lidar point data to obtain the sample three-dimensional point cloud data marked with the semantic segmentation category label includes: According to the preset perception distance, each data point in the sample voxel grid is judged. If there is a key data point within the preset perception distance with any data point to be marked as the center, a feature difference value is calculated according to the feature vectors corresponding to the data point to be marked and the key data point, wherein the key data point is a data point in the sample voxel grid that has been determined to be a point cloud data point to be marked; When it is determined that the feature difference value is less than or equal to a preset feature difference threshold, the data points to be marked are removed to obtain a sample voxel grid after redundant data points are removed; Marking the point cloud data points corresponding to each data point of the sample voxel grid after the redundant data points are removed in the sample lidar point data to obtain the sample three-dimensional point cloud data marked with the semantic segmentation category label; When it is determined that the feature difference value is greater than a preset feature difference threshold, or when the key data point does not exist within the preset perception distance centered on the data point to be marked, the data point to be marked is retained as the key data point.

[0056] In the present invention, redundancy filtering can also be performed on voxels, thereby further streamlining the data points within the voxels and retaining only key data points that can represent the entire voxel information, thereby avoiding redundant labeling and increasing labeling costs.

[0057] Specifically, the present invention predefines a perception distance and feature difference threshold , for any data point in the voxel grid , that is, The first data points, check at data points When the center is the preset sensing distance, Are there other data points in the range that have been selected as key data points to establish an evaluation set .

[0058] If the evaluation set is empty, then the data point Points that are considered representative can be used as key data points for subsequent labeling. If not empty, the data point is calculated With the evaluation set Each data point in The characteristic difference value of the key data point , feature difference value The calculation method is as follows:

[0059] in, and Represents data points With data points The feature vector of can be obtained through the trained feature extraction model.

[0060] In the present invention, if the evaluation set There is at least one data point in ,satisfy , that is, the feature difference between the two data points is small, indicating that the data points On the features and data points Too close, so the data points The redundant data points are defined as redundant and removed from the sample voxel grid. The redundant data point removal strategy provided by the present invention ensures the independence of the selected data points in space and features, and effectively improves the expression and utilization efficiency of point cloud data.

[0061] Based on the above embodiment, the method further includes: Sort the uncertainty scores in descending order, and sort the top scores in the sorting results. The undetermined voxel grids corresponding to the uncertain scores are used as the category coordination voxel grids; Acquire pseudo label data of each data point in the category coordination voxel grid and a preset voxel point category weight corresponding to the pseudo label data; Obtaining a data point uncertainty score corresponding to each data point in the category-coordinated voxel grid according to a product result between the uncertainty score of the category-coordinated voxel grid and the preset voxel point category weight; Sort the uncertainty scores of the data points in descending order, and put the sorting results first The data points corresponding to the uncertain scores of the data points are taken as the data points to be marked; Mark the point cloud data points corresponding to the data points to be marked in the sample laser radar point data.

[0062] Active learning strategies can effectively select samples with rich information, but these samples may not be enough to maintain the balance between categories, especially when the categories are extremely unbalanced. Without class balancing, the model may perform well on the majority class and poorly on the minority class. This deviation may reduce the effectiveness of the point cloud semantic segmentation model in practical applications, especially in scenarios where the minority class is crucial. To this end, the present invention introduces an online class balancing module to ensure that all categories are adequately covered by samples and provide a more balanced class representation, thereby improving the overall performance of the model.

[0063] Specifically, according to the order of uncertainty scores from large to small, select the first Undetermined voxel grids are used as category coordination voxel grids. Assume that these voxel grids contain Points. The uncertainty score for a voxel grid is ,in, For voxels For each data point in , its uncertainty score is also recorded as .

[0064] set up is the preset voxel point category weight vector, where is the number of voxel point categories, The weight of the corresponding voxel point category c. The pseudo label of each data point is obtained by predicting the relevant model. , i represents the i-th data point, , whose uncertainty score is , the data point i in the category coordination voxel grid is updated by the following formula to obtain the data point uncertainty score corresponding to each data point in the category coordination voxel grid :

[0065] in, Represents pseudo labels The preset voxel point category weight vector corresponding to the data point.

[0066] Finally, the categories are coordinated in the voxel grid in descending order of uncertainty scores. The point cloud data points corresponding to the data points are marked.

[0067] On the basis of the above embodiment, the preset three-dimensional neural network model is trained based on the sample three-dimensional point cloud data to obtain the point cloud semantic segmentation model, including: Based on the sample three-dimensional point cloud data and the cross entropy loss function, the preset three-dimensional neural network model is trained to obtain the point cloud semantic segmentation model.

[0068] In this invention, in order to accurately train the point cloud semantic segmentation model and optimize its parameters, the standard cross entropy loss function is used as the overall loss function. The cross entropy loss function is a standard loss function for classification tasks, which calculates the difference between the predicted probability distribution output by the model and the target probability distribution of the true label. The formula of this loss function can be expressed as:

[0069] in, represents the number of samples, represents the number of categories, Representation sample For category The true label (usually 0 or 1), Represents the model prediction sample Belongs to category probability.

[0070] In the present invention, the cross entropy loss function is used to optimize the model by minimizing the distance between the predicted probability and the target probability, which can directly provide feedback for the prediction of each category, which not only improves the model's prediction probability for the correct category, but also suppresses the prediction probability for the wrong category, thereby effectively improving the accuracy of classification.

[0071] In one embodiment, the performance of the point cloud semantic segmentation method based on voxel cohesion guided active learning provided by the present invention is experimentally verified, and the selected benchmark data sets include nuScenes and SemanticKITTI.

[0072] The nuScenes dataset includes 1,000 driving scenes from different cities, covering a variety of weather and lighting conditions, reflecting the complexity of real-world driving. Each scene is 20 seconds long, with a total of 28,130 training frames and 6,019 validation frames. Each frame contains a point cloud generated by 32 laser radars, with point-by-point annotations for each point, and RGB images captured by six cameras from different perspectives, segmented using sixteen categories.

[0073] The SemanticKITTI dataset is a popular segmentation dataset focused on LiDAR point clouds, providing a total of 43,000 scans with point-by-point semantic annotations, of which 23,201 scans from sequence 00 to 10 are used for training and validation, including 19,130 ​​training scans and 4,071 validation scans, supporting deep segmentation studies of 19 categories. These features make SemanticKITTI an ideal choice for understanding and evaluating point cloud data processing techniques.

[0074] The present invention uses SPVCNN as the backbone, the hidden dimension of SemanticKITTI is 64, and the hidden dimension of NuScenes is 128 to speed up the network. Random perturbations and random flipping are used to enhance the input point cloud data, and the stochastic gradient descent (SGD) optimizer is used. The initial learning rate lr is 0.24, the momentum is 0.9, and the cosine annealing strategy is used to decay the learning rate, and the weight decay coefficient is 1e-4. The total training rounds are set to 65 and the batch size is set to 8. All codes are implemented based on the PyTorch framework and experiments are carried out on a server equipped with a GeForce RTX 4090, the device video memory is 24GB, the operating system version is Ubuntu22.04, and the cuda version is 12.1.2.

[0075] In this paper, the mean intersection over union (mIoU) is mainly used to evaluate the performance of the method. mIoU is a commonly used indicator for evaluating classification accuracy, especially for semantic segmentation tasks, because it comprehensively considers the prediction accuracy and recall rate of each category. The calculation method of mIoU is as follows: First, the intersection over union (IoU) is calculated for each category separately, and then the IoU of all categories is averaged. The IoU of a category is calculated by dividing the number of points that the model correctly classifies for that category by the sum of the total number of points in the predicted set and the true set. The specific calculation formula is:

[0076] in, , and Represent the number of true positives, false positives, true negatives and false negatives, respectively. is the total number of categories.

[0077] In order to verify the effectiveness of VCMRF and OCC, the present invention conducted a series of ablation experiments on the SemantiKitti dataset. Table 1 shows the performance comparison results of the method provided by the present invention and other common strategies under different benchmark active learning strategies: Table 1 Comparison results of different benchmark active learning strategies

[0078] Furthermore, in order to evaluate the efficiency of the proposed method, several benchmark active learning strategies were replaced on a unified network architecture (SPVCNN), and the labeling budget was strictly controlled (i.e., only 0.1% of the points were labeled). The experimental results show that compared with the random point selection method, the method of the present invention significantly improved by 11.5%, reaching 61.7% mIoU, while other methods, such as the entropy-based strategy and the edge strategy, achieved 56.8% and 52.1% mIoU respectively. In contrast, the method of the present invention can more efficiently identify and label information-rich points, while effectively addressing the problem of class imbalance, thereby significantly improving the model performance.

[0079] Furthermore, in order to verify the efficiency of each module in the method of the present invention, several modules provided by the present invention (i.e., backbone network, voxel cohesion and redundant filtering module, and online adaptive category coordination module, etc.) are used in different combination experiments to verify their effectiveness. In the case of a total of identically labeled points (0.1%), the backbone network of the present invention is used for direct training, with an initial random one thousandth of the point labeling, and the mIoU is only 54.3%. After using the voxel cohesion and redundant filtering module, the segmentation accuracy is improved to 60.2%. If the voxel cohesion and redundant filtering module is combined with the online adaptive category coordination module, the performance can be as high as 61.7%.

[0080] Furthermore, the present invention randomly selects five typical scenes, and the real value, full supervision results, the method of the present invention, and the visualization comparison results of the random active learning strategy. From the comparison, it can be intuitively seen that the method of the present invention is very close to the results of full supervision learning in terms of overall effect, and even in some scenes, the method of the present invention shows better performance for the recognition of rare categories. In contrast, the results of the random active learning strategy are not as good as the learning method of the present invention.

[0081] In some scenarios, the semantic segmentation accuracy of the random strategy is low, especially for some rare categories (such as traffic signals, tree trunks and other motor vehicles in the scene) and scene junctions (such as the sidewalk junction in the scene). The randomly selected samples failed to fully cover the important information in the data, resulting in a high segmentation error rate in these scenarios. This shows that the random active learning strategy is difficult to effectively capture the complex patterns and detailed information in the data, thereby affecting the overall performance of the segmentation model. In contrast, the method proposed in the present invention intelligently selects samples with high information content for labeling, so that the model can better learn key features, thereby reducing the labeling cost and still achieving a segmentation effect that is basically equivalent to the full supervision method.

[0082] Table 2 shows the comparison results between the proposed method and the existing methods on the SemantiKitti dataset, covering 19 semantic segmentation categories: Table 2 Comparison of SemantiKitti datasets

[0083] SPVCNN is a fully supervised method and the three-dimensional network skeleton used in the present invention. Although it achieved the highest mIoU (64.5%), the labeling cost is extremely high. ReDAL achieved an mIoU (59.8%) close to that of full supervision at a labeling rate of 5%, demonstrating the high efficiency of active learning. SLiDR and LESS also achieved good results using extremely low labeled data, especially in the truck and road categories, but had lower accuracy in rare categories such as cyclists. In contrast, the method of the present invention achieved an mIoU of 61.7% using only 0.1% of the labeled points, and some categories even surpassed the fully supervised results. This shows that the method of the present invention can maintain excellent segmentation performance while reducing the labeling cost.

[0084] In view of the problems existing in existing 3D point cloud semantic segmentation methods, such as long data set annotation time, high labor cost, noise sensitivity and insufficient precision, the present invention proposes a point cloud semantic segmentation method based on voxel cohesion guided active learning.

[0085] Specifically, the present invention proposes a voxel cohesion measurement and redundancy filtering module, which evaluates the uncertainty of a single voxel through the voxel cohesion within each voxel to obtain the similarity relationship between the points within the voxel and achieve accurate filtering of redundant information points within the voxel. This process not only filters out redundant data points, but also retains key information points that can represent all points within the voxel. In addition, the module avoids the practice of directly labeling the entire voxel in the existing method, reduces the labeling budget wasted due to redundant labeling, and effectively solves the problem of redundant labeling in each voxel.

[0086] On this basis, we use an online adaptive category coordination module to dynamically adjust the number of candidate points of each category during the annotation process by setting specific thresholds for different categories, and adopt an adaptive threshold adjustment mechanism to automatically optimize the threshold setting according to the data density and annotation requirements of the category, thereby ensuring that the representativeness and balance of each category in the data set are effectively improved, reducing the negative impact of category imbalance on model training results. At the same time, by optimizing the allocation of annotation resources, we improve the annotation efficiency and data utilization rate, solve the problem of category imbalance in the point selection process, and avoid the network from showing overconfidence in the dominant category.

[0087] The present invention not only effectively improves the labeling efficiency, but also ensures the category balance of the data set through the innovative combination of the voxel cohesion and redundant filtering module and the online adaptive category coordination module, thereby significantly improving the accuracy and generalization ability of the model. It has important practical application value for improving the efficiency of point cloud labeling and the accuracy of data processing, and is particularly suitable for autonomous driving, robot navigation, building information modeling and other fields.

[0088] The point cloud semantic segmentation system based on voxel cohesion guided active learning provided by the present invention is described below. The point cloud semantic segmentation system based on voxel cohesion guided active learning described below and the point cloud semantic segmentation method based on voxel cohesion guided active learning described above can be referenced to each other.

[0089] Figure 3 A schematic diagram of the structure of the point cloud semantic segmentation system based on voxel cohesion guided active learning provided by the present invention is shown in FIG. Figure 3 As shown, the present invention provides a point cloud semantic segmentation system based on voxel cohesion guided active learning, including a point cloud data acquisition module 301 and a point cloud semantic segmentation module 302, wherein the point cloud data acquisition module 301 is used to obtain target three-dimensional point cloud data corresponding to the scene to be detected; the point cloud semantic segmentation module is used to input the target three-dimensional point cloud data into a point cloud semantic segmentation model to obtain a semantic segmentation category result corresponding to the target three-dimensional point cloud data output by the point cloud semantic segmentation model, wherein the point cloud semantic segmentation model is trained by sample three-dimensional point cloud data marked with semantic segmentation category labels, and the sample three-dimensional point cloud data is point cloud data in a sample voxel grid determined based on voxel cohesion, and the voxel cohesion is determined according to the number of data points corresponding to point cloud data points with different semantic segmentation category labels in the sample voxel grid.

[0090] The point cloud semantic segmentation system based on voxel cohesion guided active learning provided by the present invention determines the point cloud data in the sample voxel grid through voxel cohesion, and then performs point cloud semantic segmentation processing based on the point cloud semantic segmentation model trained with these point cloud data, thereby improving the accuracy and generalization ability of the point cloud semantic segmentation results.

[0091] The system provided by the present invention is used to execute the above-mentioned method embodiments. Please refer to the above-mentioned embodiments for the specific processes and detailed contents, which will not be repeated here.

[0092] Figure 4 A schematic diagram of the structure of an electronic device provided by the present invention, such as Figure 4 As shown, the electronic device may include: a processor (Processor) 401, a communication interface (Communications Interface) 402, a memory (Memory) 403 and a communication bus 404, wherein the processor 401, the communication interface 402, and the memory 403 complete mutual communication through the communication bus 404. The processor 401 may call the logic instructions in the memory 403 to execute the point cloud semantic segmentation method based on voxel cohesion guided active learning, the method comprising: obtaining target three-dimensional point cloud data corresponding to the scene to be detected; inputting the target three-dimensional point cloud data into the point cloud semantic segmentation model, and obtaining the semantic segmentation category result corresponding to the target three-dimensional point cloud data output by the point cloud semantic segmentation model, wherein the point cloud semantic segmentation model is obtained by training sample three-dimensional point cloud data marked with semantic segmentation category labels, the sample three-dimensional point cloud data is point cloud data in a sample voxel grid determined based on voxel cohesion, and the voxel cohesion is determined according to the number of data points corresponding to point cloud data points of different semantic segmentation category labels in the sample voxel grid.

[0093] In addition, the logic instructions in the above-mentioned memory 403 can be implemented in the form of a software functional unit and can be stored in a computer-readable storage medium when it is sold or used as an independent product. Based on such an understanding, the technical solution of the present invention, in essence, or the part that contributes to the prior art or the part of the technical solution, can be embodied in the form of a software product, and the computer software product is stored in a storage medium, including a number of instructions for a computer device (which can be a personal computer, a server, or a network device, etc.) to perform all or part of the steps of the method described in each embodiment of the present invention. The aforementioned storage medium includes: U disk, mobile hard disk, read-only memory (ROM, Read-Only Memory), random access memory (RAM, Random Access Memory), disk or optical disk, etc. Various media that can store program codes.

[0094] On the other hand, the present invention also provides a computer program product, which includes a computer program stored on a non-transitory computer-readable storage medium, and the computer program includes program instructions. When the program instructions are executed by a computer, the computer can execute the point cloud semantic segmentation method based on voxel cohesion guided active learning provided by the above methods, and the method includes: obtaining target three-dimensional point cloud data corresponding to the scene to be detected; inputting the target three-dimensional point cloud data into a point cloud semantic segmentation model to obtain a semantic segmentation category result corresponding to the target three-dimensional point cloud data output by the point cloud semantic segmentation model, wherein the point cloud semantic segmentation model is trained by sample three-dimensional point cloud data marked with semantic segmentation category labels, and the sample three-dimensional point cloud data is point cloud data in a sample voxel grid determined based on voxel cohesion, and the voxel cohesion is determined according to the number of data points corresponding to point cloud data points with different semantic segmentation category labels in the sample voxel grid.

[0095] On the other hand, the present invention also provides a non-transitory computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, is implemented to execute the point cloud semantic segmentation method based on voxel cohesion guided active learning provided in the above-mentioned embodiments, the method comprising: obtaining target three-dimensional point cloud data corresponding to a scene to be detected; inputting the target three-dimensional point cloud data into a point cloud semantic segmentation model, and obtaining a semantic segmentation category result corresponding to the target three-dimensional point cloud data output by the point cloud semantic segmentation model, wherein the point cloud semantic segmentation model is trained by sample three-dimensional point cloud data marked with a semantic segmentation category label, and the sample three-dimensional point cloud data is point cloud data in a sample voxel grid determined based on voxel cohesion, and the voxel cohesion is determined according to the number of data points corresponding to point cloud data points with different semantic segmentation category labels in the sample voxel grid.

[0096] The device embodiments described above are merely illustrative, wherein the units described as separate components may or may not be physically separated, and the components displayed as units may or may not be physical units, that is, they may be located in one place, or they may be distributed on multiple network units. Some or all of the modules may be selected according to actual needs to achieve the purpose of the scheme of this embodiment. Ordinary technicians in this field can understand and implement it without paying creative labor.

[0097] Through the description of the above implementation methods, those skilled in the art can clearly understand that each implementation method can be implemented by means of software plus a necessary general hardware platform, and of course, can also be implemented by hardware. Based on this understanding, the above technical solution is essentially or the part that contributes to the prior art can be embodied in the form of a software product, and the computer software product can be stored in a computer-readable storage medium, such as ROM / RAM, a disk, an optical disk, etc., including a number of instructions for a computer device (which can be a personal computer, a server, or a network device, etc.) to execute the methods described in each embodiment or some parts of the embodiments.

[0098] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, rather than to limit it. Although the present invention has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that they can still modify the technical solutions described in the aforementioned embodiments, or make equivalent replacements for some of the technical features therein. However, these modifications or replacements do not deviate the essence of the corresponding technical solutions from the spirit and scope of the technical solutions of the embodiments of the present invention.

Claims

1. A point cloud semantic segmentation method based on voxel cohesion guided active learning, characterized in that: include: Obtain target three-dimensional point cloud data corresponding to the scene to be detected; The target three-dimensional point cloud data is input into a point cloud semantic segmentation model to obtain a semantic segmentation category result corresponding to the target three-dimensional point cloud data output by the point cloud semantic segmentation model, wherein the point cloud semantic segmentation model is trained by sample three-dimensional point cloud data marked with semantic segmentation category labels, and the sample three-dimensional point cloud data is point cloud data in a sample voxel grid determined based on voxel cohesion, and the voxel cohesion is determined according to the number of data points corresponding to point cloud data points with different semantic segmentation category labels in the sample voxel grid.

2. The point cloud semantic segmentation method based on voxel cohesion guided active learning according to claim 1, characterized in that: The point cloud semantic segmentation model is trained by the following steps: Determine, according to the feature map corresponding to the sample laser radar point cloud data, the semantic category label of each point cloud data point in the sample laser radar point cloud data; voxelize the sample laser radar point cloud data to obtain a plurality of pending voxel grids, wherein the data points in the pending voxel grids are determined based on the coordinate information of the point cloud data points in the sample laser radar point data and a preset voxel size, and the voxel point category labels of the data points in the pending voxel grids correspond to the semantic category labels of the point cloud data points in the sample laser radar point data; Calculating the voxel cohesion of the undetermined voxel grid according to the uncertainty score of the undetermined voxel grid, wherein the uncertainty score is calculated based on the number of data points of each different voxel point category label in the voxel grid and the number of all data points in the voxel grid; Based on the voxel cohesion of each of the to-be-determined voxel grids, determining the to-be-determined voxel grid with the lowest voxel cohesion as the sample voxel grid; Marking the point cloud data points corresponding to each data point in the sample voxel grid in the sample lidar point data to obtain the sample three-dimensional point cloud data marked with the semantic segmentation category label; Based on the sample three-dimensional point cloud data, a preset three-dimensional neural network model is trained to obtain the point cloud semantic segmentation model.

3. The point cloud semantic segmentation method based on voxel cohesion guided active learning according to claim 2, characterized in that: The step of calculating the voxel cohesion of the to-be-determined voxel grid according to the uncertainty score of the to-be-determined voxel grid comprises: Based on the voxel cohesion calculation formula, the voxel cohesion of the to-be-determined voxel grid is calculated according to the uncertainty score of the to-be-determined voxel grid. The voxel cohesion calculation formula is: in, Indicates The first The number of data points with voxel class labels, Indicates A grid of undetermined voxels, represents a collection of voxel grids, Indicates The first voxel in the undetermined voxel grid The predicted category corresponding to the data point, Represents the total number of voxel point category labels, Indicates The voxel cohesion corresponding to the undetermined voxel grid is Indicates The number of all data points in the undetermined voxel grid, Indicates The uncertainty score corresponding to the undetermined voxel grid, Represents the total number of data point categories within the pending voxel grid.

4. The point cloud semantic segmentation method based on voxel cohesion guided active learning according to claim 2, characterized in that: The step of marking the point cloud data points corresponding to each data point in the sample voxel grid in the sample laser radar point data to obtain the sample three-dimensional point cloud data marked with the semantic segmentation category label includes: According to the preset perception distance, each data point in the sample voxel grid is judged. If there is a key data point within the preset perception distance with any data point to be marked as the center, a feature difference value is calculated according to the feature vectors corresponding to the data point to be marked and the key data point, wherein the key data point is a data point in the sample voxel grid that has been determined to be a point cloud data point to be marked; When it is determined that the feature difference value is less than or equal to a preset feature difference threshold, the data points to be marked are removed to obtain a sample voxel grid after redundant data points are removed; Marking the point cloud data points corresponding to each data point of the sample voxel grid after the redundant data points are removed in the sample lidar point data to obtain the sample three-dimensional point cloud data marked with the semantic segmentation category label; When it is determined that the feature difference value is greater than a preset feature difference threshold, or when the key data point does not exist within the preset perception distance centered on the data point to be marked, the data point to be marked is retained as the key data point.

5. The point cloud semantic segmentation method based on voxel cohesion guided active learning according to claim 3 or 4, characterized in that: The method further comprises: Sort the uncertainty scores in descending order, and sort the top scores in the sorting results. The undetermined voxel grids corresponding to the uncertain scores are used as the category coordination voxel grids; Acquire pseudo label data of each data point in the category coordination voxel grid and a preset voxel point category weight corresponding to the pseudo label data; Obtaining a data point uncertainty score corresponding to each data point in the category-coordinated voxel grid according to a product result between the uncertainty score of the category-coordinated voxel grid and the preset voxel point category weight; Sort the uncertainty scores of the data points in descending order, and put the sorting results first The data points corresponding to the uncertain scores of the data points are taken as the data points to be marked; Mark the point cloud data points corresponding to the data points to be marked in the sample laser radar point data.

6. The point cloud semantic segmentation method based on voxel cohesion guided active learning according to claim 2, characterized in that: The method of training a preset three-dimensional neural network model based on the sample three-dimensional point cloud data to obtain the point cloud semantic segmentation model includes: Based on the sample three-dimensional point cloud data and the cross entropy loss function, the preset three-dimensional neural network model is trained to obtain the point cloud semantic segmentation model.

7. A point cloud semantic segmentation system based on voxel cohesion guided active learning, characterized in that: include: Point cloud data acquisition module, used to obtain target three-dimensional point cloud data corresponding to the scene to be detected; A point cloud semantic segmentation module is used to input the target three-dimensional point cloud data into a point cloud semantic segmentation model to obtain a semantic segmentation category result corresponding to the target three-dimensional point cloud data output by the point cloud semantic segmentation model, wherein the point cloud semantic segmentation model is trained by sample three-dimensional point cloud data marked with semantic segmentation category labels, and the sample three-dimensional point cloud data is point cloud data in a sample voxel grid determined based on voxel cohesion, and the voxel cohesion is determined according to the number of data points corresponding to point cloud data points with different semantic segmentation category labels in the sample voxel grid.

8. An electronic device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein: When the processor executes the computer program, the point cloud semantic segmentation method based on voxel cohesion guided active learning as described in any one of claims 1 to 6 is implemented.

9. A non-transitory computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the point cloud semantic segmentation method based on voxel cohesion guided active learning as described in any one of claims 1 to 6 is implemented.

10. A computer program product, comprising a computer program, characterized in that When the computer program is executed by a processor, the point cloud semantic segmentation method based on voxel cohesion guided active learning as described in any one of claims 1 to 6 is implemented.

Images