108 results about "Point cloud segmentation" patented technology

This application discloses a method, apparatus, storage medium, electronic device, and computer program product for point cloud segmentation of suspended objects based on multi-frame point clouds, belonging to the field of construction technology. The method includes: during the lifting of suspended objects by engineering machinery with lifting function, collecting a point cloud sequence of the area where the suspended object is located, the point cloud sequence including at least two frames of point clouds; determining multiple changing point cloud clusters in each frame of point cloud based on adjacent frames of point clouds in the point cloud sequence; determining multiple candidate point cloud clusters of suspended objects in the area from below the hook to the ground from each frame of point cloud; and segmenting the point cloud of the suspended object in each frame of point cloud based on the changing point cloud clusters and the candidate point cloud clusters of suspended objects. Therefore, it is not only applicable to point cloud segmentation of various types of suspended objects, but also has a wide range of applications, high accuracy in suspended object recognition, and good recognition effect. Moreover, it does not require the use of deep learning models or manual point cloud calibration.

This invention discloses an intelligent measurement method and system for structural cracks based on laser measurement, belonging to the field of structural inspection and measurement technology. The method includes acquiring a high-density three-dimensional point cloud of the structural surface through laser scanning, segmenting the point cloud of candidate crack regions and the point cloud of the complete structural surface. Geometric morphological refinement and skeleton extraction are performed on the candidate point cloud to generate crack skeleton lines and construct its topological connection map, thereby accurately calculating the crack length, width, and orientation angle. By fusing the crack geometric parameters with the structural surface reference information, the spatial position parameters of the crack on the structure are calculated. This method ensures the accuracy of calculating the geometric parameters of complex cracks through skeleton extraction and topological analysis, and provides a precise spatial orientation description of the crack by fusing structural reference information, ultimately generating a comprehensive measurement report including geometric morphology and spatial position.

This invention belongs to the field of medical image processing technology and proposes a method for 3D tooth point cloud segmentation and tooth type recognition based on deep learning. The main scheme is as follows: acquire 3D tooth point cloud data, construct a 3D tooth point cloud dataset and perform data preprocessing; construct a 3D tooth point cloud semantic segmentation network and determine the loss function, and train the 3D tooth point cloud semantic segmentation network based on the loss function; use the trained 3D tooth point cloud semantic segmentation network to perform point-by-point semantic prediction on the point cloud of the tooth row to be tested, and obtain the point-level category probability distribution; perform multi-level probability smoothing and interpolation recovery on the point-level category probability distribution to improve the spatial continuity of the prediction results and restore the original point cloud resolution; perform density-adaptive connectivity analysis and denoising processing on the restored point cloud segmentation results to extract single tooth instances; and use instance-level category statistics for each single tooth instance to determine the final semantic category, thereby realizing tooth type recognition.

This invention discloses a fast point cloud segmentation method based on image mapping for local workpiece grinding and polishing, belonging to the fields of robot vision perception, image processing, and 3D point cloud segmentation technology. The method first sets a red closed boundary around the area to be ground and polished. Then, a robotic arm equipped with a depth camera simultaneously acquires RGB color images and depth data, constructing an ordered point cloud with a one-to-one correspondence between image pixels. Subsequently, the red boundary is detected by fusing features from multiple color spaces (HSV, RGB, and Lab). A mask for the area to be ground and polished is obtained through morphological closing operations, maximum contour filtering, and region filling. Finally, the point cloud of the area to be ground and polished is extracted based on the correspondence between the target pixel positions in the mask and the midpoint positions in the ordered point cloud. This invention reduces the computational complexity of direct 3D point cloud segmentation and improves the accuracy, stability, and processing efficiency of point cloud extraction for local areas to be ground and polished.

This invention discloses a method and system for three-dimensional phenotypic analysis of potatoes based on computer vision. The method includes: first, constructing an acquisition system with multiple high-resolution RGB cameras in a triangular layout and a uniform light source, including a motion turntable; after distortion correction via checkerboard calibration, simultaneously acquiring multi-view images of potatoes; next, generating a dense depth map through sparse point cloud reconstruction combined with an improved multi-view stereo vision reconstruction mechanism including dynamic adaptive aggregation range and curvature-weighted matching cost calculation; registering and fusing to obtain three-dimensional point cloud data containing color and texture; then extracting basic parameters such as length, volume, and RGB features; and using an improved point cloud segmentation model including curvature-sensitive sampling, geometric attention modules, and a hybrid loss function to segment buds and extract parameters such as number and depth. This provides technical support for potato breeding and intelligent sorting.

The invention discloses a photovoltaic component segmentation positioning method and a related device, and belongs to the field of photovoltaic constructions.The method comprises the following steps that pre-processed point cloud data are preliminarily segmented until all the pre-processed point cloud data have affiliation sets, and then a point cloud set is obtained; calculating a domain point corresponding to each point in the point cloud set, constructing a domain point matrix according to the domain points, processing the domain point matrix by adopting a vector cross multiplication method, and normalizing to obtain a unit normal vector; and after the point cloud set is segmented again, the point cloud data segmented again is screened according to the unit normal vector, an expected contour point set of the BIM model is extracted, the screened point cloud data and the expected contour point set are registered, and a point cloud segmentation positioning result of the photovoltaic component is obtained. According to the method, the problems that images are lost and photovoltaic component positioning analysis cannot be accurately carried out due to the fact that a photogrammetry method is adopted by a photovoltaic power station in a complex environment can be solved.

This application relates to the field of point cloud segmentation technology, disclosing a method, apparatus, device, and medium for point cloud segmentation in non-flat road environments. The method includes: acquiring point cloud data and obtaining a voxel cloud based on voxelization of the point cloud data; when a preset update condition is met, translating the voxel cloud so that the voxel cloud is located in the central region of the target device, and updating the translated voxel cloud; adding the height value of the target point to the array corresponding to each target voxel of the target point; generating a terrain point cloud based on each array and the updated voxel cloud; generating a traversable area contour based on the traversable area, and identifying step-type obstacles based on the traversable area contour. This invention can segment terrain point clouds of traversable areas and obstacle areas using point cloud data, and identify step-type obstacles, thereby significantly improving the point cloud segmentation accuracy in complex, non-flat road environments.

The application provides a three-dimensional point cloud segmentation method, system, device and medium, the method comprising: constructing an initial three-dimensional Gaussian point cloud according to multi-view image data of a target scene; obtaining first base elements located in a boundary blur region in all three-dimensional Gaussian base elements of the initial three-dimensional Gaussian point cloud according to initial semantic masks of each frame of image in the multi-view image data, and performing an adaptive splitting operation on each first base element to obtain a plurality of second base elements; performing multi-scale semantic feature training on a target three-dimensional Gaussian point cloud with the initial semantic mask as a supervision signal to obtain a semantic feature vector of each base element in the target three-dimensional Gaussian point cloud; and assigning a corresponding semantic label to each base element in the target three-dimensional Gaussian point cloud according to the semantic feature vector to obtain a three-dimensional Gaussian point cloud after semantic segmentation. The application realizes optimization of a boundary splitting process of Gaussian splashing and a semantic feature matching process, and improves the accuracy, robustness and flexibility of three-dimensional Gaussian scene segmentation.

This invention discloses a method and system for measuring vehicle dimensions based on a mobile acquisition device. The method includes: time synchronization and joint calibration of a lidar and a camera; obtaining the target vehicle position as the target detection result; stitching together each frame of point cloud data to complete the three-dimensional reconstruction of the current space and reproduce the vehicle outline point cloud information; segmenting to obtain ground point cloud and non-ground point cloud; combining the target detection result to obtain the target point cloud; extracting vehicle outline points from the target point cloud; fitting the length L and width W of the target vehicle's outline dimensions based on the yaw angle of the vehicle outline points; obtaining the height H of the target vehicle's outline dimensions. The system includes: a mobile acquisition device, a data preprocessing module, a target detection module, a point cloud stitching module, a ground point segmentation module, a target point cloud segmentation module, a vehicle outline point segmentation module, a first dimension calculation module, and a second dimension calculation module.

The application provides a perception method based on a light-weight geometric information decoupling graph convolution point cloud perception model, and belongs to the technical fields of point cloud perception and deep learning. The method comprises the following steps: obtaining candidate region retrieval by discretizing the original point cloud data space coordinates through 2D voxel down-sampling based on the light-weight geometric information decoupling graph convolution point cloud perception model; obtaining adjacent edge index by using a scale factor-based expansion KNN algorithm based on the candidate region. The encoder performs multiple graph convolution and down-sampling processes based on the adjacent edge index and the point cloud subset, and the decoder performs multiple up-sampling and graph convolution processes based on the point cloud subset and the adjacent edge index. The application can complete point cloud classification by extracting multi-layer graph convolution features through the encoder. For point cloud segmentation and target identification tasks, the decoder gradually reconstructs spatial details in the up-sampling stage, effectively compensating for the loss of geometric information caused by down-sampling in the encoding process.

This invention relates to the field of pose estimation technology, specifically to a multi-view, multi-person 3D pose estimation method, apparatus, and device. The method includes: acquiring image information and depth information of the target using multi-view acquisition technology; segmenting the person in a single view based on the image information and extracting 2D semantic features; reconstructing the point cloud based on the depth information to obtain a 3D point cloud in a preset world coordinate system; segmenting the point cloud for different human figures based on the 3D point cloud and calculating the point cloud boundaries as human body boundaries; mapping the 2D semantic features to the 3D point cloud to obtain 3D features; and classifying different human figures and constructing 3D poses based on the 3D features and human body boundaries. This technical solution can segment the point cloud for different human figures based on the 3D point cloud, accurately define the joint positions of different figures, and achieve differentiation and matching of different figures. Furthermore, based on discrete point cloud data, it has better flexibility and fault tolerance.

The application discloses a kind of unmanned vehicle fast point cloud segmentation algorithm under complex environment, by original laser radar point cloud is encoded as initial equilateral triangle grid chart structure, introduce regularized space representation, to enhance the expression ability of point cloud local geometry;For each point cloud in equilateral triangle grid, the local geometric feature is extracted and the grid flatness is calculated using principal component analysis method, so as to realize the effective distinction of ground area and non-ground area;For the grid area with higher flatness, directly determine the internal point cloud as ground point cloud;For the grid area with lower flatness and complex terrain, then improve the local modeling accuracy by adaptive subdivision operation;In the non-ground point cloud segmentation stage, an adaptive euclidean clustering algorithm based on laser radar angle resolution is used to accelerate the segmentation, and the clustering scale is set adaptively for point clouds in different distance ranges, so as to significantly reduce the computational complexity while ensuring the accuracy of segmentation.

This invention discloses an adaptive point cloud completion method based on rail geometric features, solving the problem that sparse rail point clouds during high-speed train operation lead to decreased rail point cloud segmentation accuracy, resulting in the inability to effectively identify potential foreign object intrusion risks. The method includes: constructing a rail geometric feature library; acquiring sparse point cloud data of the rail ahead of the train and performing coordinate system transformation; extracting rail geometric features and calculating feature vectors; matching the extracted geometric features with the rail geometric feature library to obtain the optimal matching rail model; generating geometric constraints; correcting curvature features and track superelevation features based on the optimal matching rail model; dynamically adjusting point cloud completion parameters according to train operating status parameters; and generating a completed point cloud based on the corrected point cloud in S3. The completed point cloud accurately determines the rail position to achieve boundary determination, thus providing a key geometric basis for foreign object intrusion identification and ensuring train operation safety.

This invention proposes a PointNeXt point cloud segmentation method and system that integrates a linear attention mechanism, relating to the field of semantic segmentation technology. The method includes: acquiring target 3D point cloud data and inputting it into an improved PointNeXt network; using a progressive linear attention module to extract local features and model the global context of the target 3D point cloud data to obtain corresponding local and global point cloud features; performing channel-dimensional concatenation and nonlinear fusion to obtain a first intermediate feature; downsampling the first intermediate feature to obtain a downsampled feature; enhancing and dynamically weighting the downsampled feature using the channel attention and spatial attention mechanisms of an enhanced dual-attention multilayer perception module to obtain a target fused feature; and using the upsampled target fused feature to perform semantic segmentation of the target 3D point cloud data to determine the semantic label corresponding to each point in the target 3D point cloud data.

This invention relates to the field of infrastructure engineering technology, specifically to a machine vision-based method and system for locating spraying areas. The method includes: using a depth camera to capture the original point cloud of the beam surface, segmenting and selecting target planes on the beam surface using the RANSAC algorithm; converting the image to two dimensions, extracting line features using Hough transform and edge detection, inversely mapping back to three dimensions, and selecting the uppermost point of the line; moving the camera upwards to capture the point cloud of the upper part of the beam, removing interference to extract two target planes; calculating the normal vectors of the two planes, comparing the point cloud to obtain the coordinates of the calibration points, constructing a new three-dimensional coordinate system and generating a rotation matrix; and planning the robot spraying path based on the beam surface design parameters. The system includes: a depth camera, point cloud segmentation, two-dimensional mapping and feature extraction, three-dimensional coordinate recovery and key point selection, spatial calibration, and path planning modules, which work together to implement the above method, eliminate point cloud blind spots, improve positioning accuracy, and meet the requirements for precise spraying of beam surfaces.

The application discloses a point cloud data wood and leaf separation method, acquires tree three-dimensional point clouds, and pre-processes to obtain normalized point clouds; a point feature vector containing coordinates, curvatures and linearity is constructed, a point cloud Transformer segmentation network integrating a scale self-adaptive deformable attention is input, and each point trunk, branchlet or leaf label is output, so that wood point cloud and leaf point cloud separation results are obtained; in the training stage, a branchlet F1 guide loss is introduced and combined with a reweighted or focal loss; after reasoning, a skeleton is extracted based on voxelization refinement, connectivity constraints are applied to correct misclassification and remove isolated points, and a structure-continuous wood point cloud is output.

This invention discloses a method and apparatus for segmenting transmission line point clouds based on geometric enhancement and obfuscation constraints, belonging to the field of image data processing technology. The method includes: preprocessing the original 3D point cloud of the transmission line to extract multi-scale geometric features and highly normalized features, generating a multi-dimensional enhanced feature vector; inputting the vector into a Point Transformer V3 backbone network for sequential attention encoding-decoding, outputting a geometric structure feature map of each point in the transmission line scene; jointly training the network parameters using combined loss and obfuscation constraint loss; and performing structured post-processing on the prediction results to obtain the final geometric structure segmentation result at the original point level. This invention solves problems such as terrain undulation interference, cable obfuscation, weak small-sample recognition, and sampling recovery misalignment in large-scene transmission line point clouds, improving segmentation accuracy and engineering stability, and can directly output point-by-point classification results conforming to the LAS standard.

This invention discloses a semantically driven adaptive graph boundary point cloud segmentation and decoding method, comprising the following steps: S1, acquiring 3D point cloud data, image data, human behavior text data, and object name text data from the application scenario, and encoding them to obtain point cloud feature sets, image features F1, human behavior text features, and object name text features, respectively; S2, cross-modal conditional feature preprocessing, fusing image features F1 with human behavior text features to obtain image conditional features Ih; S3, semantically driven point cloud feature enhancement and multi-scale graph structure modeling; S4, cross-modal guided fine-grained point cloud segmentation and boundary uncertainty refinement decoding. This invention enables the model to more sensitively perceive boundary transition regions, thereby effectively alleviating the boundary ambiguity problem in traditional methods and improving the boundary clarity of the segmentation results.

The application relates to the technical field of silicon crystal bar measurement and classification, and discloses a silicon crystal bar diameter measurement and intelligent classification method based on point cloud segmentation fitting, which comprises the following steps: obtaining the original three-dimensional point cloud of a silicon crystal bar to be measured, carrying out pretreatment, extracting a main shaft and normalizing coordinates to obtain aligned three-dimensional point cloud; segmenting and cutting the aligned three-dimensional point cloud along the main shaft direction, screening a plurality of effective subsegments, and carrying out local cylindrical fitting on the effective subsegments to obtain local geometric parameters; carrying out feature statistical analysis according to the local geometric parameters and carrying out deviation compensation; matching the compensated geometric feature parameters with preconfigured product model templates, identifying and determining the model to which the silicon crystal bar to be measured belongs, and carrying out quality judgment and outputting results according to the quality standard of the model.

This disclosure provides a method, apparatus, device, and storage medium for point cloud segmentation of a target object. It aims to improve the accuracy of point cloud segmentation. The method includes: acquiring point cloud data and an RGB image containing the target object during crane lifting; extracting contours from the RGB image to obtain various contours; for any contour, obtaining a current feature vector of the contour based on the position of each pixel and the RGB value of each pixel; determining the target contour of the target object using the current feature vectors of each contour and the standard contour feature vector of the target object; determining the three-dimensional position coordinates of the target object's center point in a radar coordinate system based on the target contour; determining the circumscribed sphere radius of the target object based on the extreme values ​​of the horizontal and vertical position coordinates of the target contour; and segmenting the point cloud data containing the target object using the three-dimensional position coordinates of the target object's center point in the radar coordinate system and the circumscribed sphere radius to obtain the point cloud data of the target object.

The present application relates to the technical field of point cloud processing, and provides a track structure point cloud weak supervision segmentation method and system based on SAM and PointNet++. The method comprises the following steps: collecting track structure point cloud data to obtain an unlabeled point cloud data set with RGB color; constructing a point cloud processing module based on SAM and a weak supervision point cloud segmentation module based on PointNet++; preprocessing the unlabeled point cloud data set to obtain point cloud training data with pseudo-labels and confidence weights; inputting the point cloud training data into the weak supervision point cloud segmentation module based on PointNet++ for training to obtain a trained point cloud semantic segmentation model, and performing semantic segmentation on the unlabeled point cloud by using the trained point cloud semantic segmentation model. The method realizes a training effect close to that of full supervision point cloud data under weak supervision, reduces the dependence of track structure point cloud segmentation on high-precision labeled point cloud data, and improves the segmentation efficiency and accuracy.

The present application relates to point cloud semantic segmentation technology, especially to a confidence-guided multi-expert fusion open vocabulary three-dimensional point cloud segmentation method, comprising obtaining three-dimensional geometric features of to-be-segmented point cloud data from the to-be-segmented point cloud data by using a pre-trained student encoder, obtaining an adaptive text feature dictionary by using a CLIP text encoder, multiplying the obtained three-dimensional geometric features and the adaptive text feature dictionary, and obtaining a final segmentation result; the present application combines multi-model fusion, confidence-guided distillation and learnable prompt optimization, solves the problem of insufficient semantic supervision robustness and limited alignment accuracy of fixed text features and three-dimensional features caused by the dependence of the existing open vocabulary three-dimensional point cloud segmentation method on a single two-dimensional model, proposes a confidence-guided multi-model feature fusion strategy and a scene-adaptive text prompt optimization scheme, and solves the problems of multi-source model feature heterogeneity and poor scene adaptability of text features.

The present application relates to the technical field of three-dimensional point cloud, and discloses a sparse point cloud semantic segmentation method based on mask modulation noise regression.The method is as follows: training a point cloud semantic segmentation model based on mask modulation noise regression; in the training stage, a double-path joint optimization strategy is adopted to jointly optimize the semantic segmentation loss of the semantic prior segmentation network and the mask modulation noise regression loss of the mask guided diffusion branch; in the inference stage: collecting the original point cloud of the target region, preprocessing and segmenting the original point cloud to obtain a plurality of point cloud subblocks; extracting features of the point cloud subblocks to obtain point cloud subblock features, down-sampling the point cloud subblock features to obtain bottleneck features, up-sampling the bottleneck features to obtain semantic probability labels, and forming a point-level semantic segmentation probability graph according to the semantic probability labels.The present application solves the technical problem of insufficient robustness of the existing point cloud segmentation method to small targets and thin boundaries in real sparse and noisy point cloud scenes.

The application discloses a weakly supervised forest point cloud vertical layering method based on disturbance consistency and entropy regularization, and comprises the following steps: random sampling and label division are performed on original point cloud samples and disturbed point cloud samples to obtain double-branch input point clouds; a multi-scale feature extraction network is constructed, the double-branch input point clouds are taken as input, and double-branch point-by-point high-dimensional features are obtained; a classification head is arranged at the end of the multi-scale feature extraction network, the double-branch point-by-point high-dimensional features are taken as input, and double-branch point-by-point prediction probability distributions are obtained through normalization; a total loss function comprising a category perception disturbance consistency loss, an entropy regularization loss and a cross-entropy classification loss is constructed; end-to-end training optimization is performed based on the total loss function, and the to-be-predicted forest point cloud is input to obtain point-by-point vertical layering results of a forest tree layer, a shrub layer and a herb layer. The application adopts a weakly supervised learning framework, significantly reduces the labeling cost, and improves the accuracy of forest point cloud vertical structure layering.

According to at least one embodiment, a method for generating a global 2D map of an indoor environment comprises the steps of: processing a 3D point cloud of the indoor environment; detecting one or more point clusters present in the 3D point cloud; in response to the detecting, removing a cluster from the 3D point cloud to generate a global 3D point cloud on the basis of a determination that an object corresponding to the detected cluster is unnecessary; and dividing the global 3D point cloud into 3D segments. The method further comprises the steps of: for each 3D segment, identifying a local floor as a reference plane of the 3D segment; and collecting 2D slices of the 3D segment at a 2D sensor height of a robot on the basis of the identified local floor. The collected 2D slices of the 3D segments are assembled to form a global 2D map.

The application discloses a mountainous large-span bridge component point cloud segmentation method based on a large model. First, point cloud data of a mountainous large-span bridge in multiple construction stages is collected, the point cloud data in the multiple construction stages is preprocessed, and then a rectangular bounding box is constructed. Taking the number of point clouds in the rectangular bounding box as a judgment criterion, the whole bridge point cloud is screened out. Secondly, based on the obtained whole bridge point cloud and the rectangular bounding box parameters, a mapping relationship between the bridge point cloud and a two-dimensional image is constructed. Then, based on the mapping relationship between the point cloud and the two-dimensional image, target component segmentation is performed on the bridge point cloud in the primary segmentation stage. Finally, taking the bridge point cloud in the primary segmentation stage as a criterion, the bridge point cloud segmentation in the remaining construction stages is completed. The application provides high-precision component point cloud data support for the whole life cycle construction progress monitoring, operation and maintenance of the mountainous large-span bridge, and has important engineering application value.

The present application relates to computer vision and surveying data processing technical field, a kind of based on scene geometric relationship constraint's roof plane accurate extraction method and system, comprising: obtaining roof three-dimensional point cloud, roof three-dimensional point cloud is input to pre-constructed point cloud segmentation model, obtain supervoxel set, confirm out thickness characteristic value set and supervoxel barycenter set based on supervoxel set, confirm out data item set A based on roof first candidate plane primitive set and supervoxel set, confirm out smooth item set A based on roof first candidate plane primitive set, confirm out data item set based on roof candidate plane primitive set and supervoxel barycenter set, confirm out smooth item set based on supervoxel set, confirm out label item set based on roof candidate plane primitive set, perform label merging operation to supervoxel label set, obtain multiple supervoxel sets, summarize roof plane, obtain roof plane set, complete building roof point cloud in plane accurate extraction.The present application can improve the accuracy of roof plane extraction.