870 results about "Lidar point cloud" patented technology

The invention discloses an automatic classification method of airborne laser radar point cloud data, belonging to the technical field of airborne laser radar. In order to improve the efficiency and the precision of the automatic classification of the point cloud data, the method comprises the steps of: firstly, splitting the point cloud data; secondly, counting the each attribute information of splitting segments; and judging the category of the splitting segments according to the spatial space relationship among the splitting segments and the attribute information of the segments. Compared with the existing automatic classification method based on the points, the classification method based on the segments has higher robustness and precision.

The invention discloses a classification method based on vehicle-mounted LiDAR point cloud data. The method comprises the steps that streetscape vehicle-mounted LiDAR point cloud data are obtained and preprocessed so as to remove redundant data and noise in the point cloud data, and the point cloud normal vector, curvature and density are calculated in the preprocessing process; the preprocessed point cloud data are classified; according to the point cloud normal vector, curvature and density, power lines, building vertical face points and tree points are extracted. According to the classification method, implementation is easy, calculated quantity is small, implementation is fast and effective, the foundation is laid for streetscape facet extraction, and facet extraction is more accurate.

The embodiment of the invention discloses a semantic segmentation method based on multi-scale convolutional neural networks (CNNs). The method includes: acquiring intra-modal features in high-resolution aerial-images and LiDAR (Light Detection And Ranging) point cloud data; carrying out inter-modal feature extraction and classification on the basis of the multi-scale convolutional neural networks;and using a multi-scale segmentation method to extract ground object boundaries, eliminate salt and pepper effects, and optimize classification results. By implementing the example of the invention,a method of combining the multi-scale CNNs and multi-scale segmentation (MRS) post-processing is used for semantic segmentation of the high-resolution aerial-images and the LiDAR point cloud data.

The invention discloses an automatic segmentation method for a point cloud of a facade of a large scene city building. The method comprises steps of (1) performing fusion registration on airborne LiDAR point cloud data and on-vehicle LiDAR point cloud data, (2) extracting airborn LiDAR building point cloud data from the airborne LiDAR point cloud data which goes through registration in the step (1), (3) performing segmentation on the point cloud data of a single building based on the airborne LiDAR point cloud data extracted from the step (2), (4) performing contour tracking on the single building segmented by the step (3), (5) performing simplification and normalization processing on a contour line extracted from the step (4), (6) performing rough segmentation on the point cloud of the facade of the building based on the contour line which goes through the simplification and normalization processing in the step (5), and (7) performing fine segmentation on the building facade point cloud which is roughly segmented in the step (6). The automatic segmentation method of the invention can fast and accurately separate the building facade point cloud from the on-vehicle LiDAR point cloud.

The invention belongs to the field of intelligent unmanned intelligent ships, and relates to an unmanned ship water surface target detection, identification and positioning method based on monocular camera and lidar information fusion. The detection, identification and positioning of a water surface target by an unmanned ship are influenced by distances and the fluctuation of the target, so that alidar and a monocular camera are integrated to accurately detect, identify and position the target within a sensing range. According to the method, acquired water surface target images are adopted totrain a neural network-based target detection and recognition model; the lidar employs a conditional removal filter and Euclidean clustering to obtain the position of the water surface target in a world coordinate system; and finally, a camera image information and lidar information fusion method is designed, and therefore, the method is highly robust to uncertain factors. With the method adopted, the unmanned ship is capable of accurately detecting, identifying and positioning the water surface target; and good environment perception can be realized for the target tracking, path planning andautonomous navigation of the unmanned ship. The method has a broad application prospect.

The invention discloses a building roof reconstruction method based on airborne LiDAR data. The method comprises the steps of 1, obtaining LiDAR point cloud data, and 2, processing the LiDAR point cloud data; the step 2 comprises the processes of 201, dividing building roof patches to obtain a plurality of roof patches from the building roof, 202, obtaining building roof point cloud data, 203, extracting outer contour points of the building roof and the roof patches, and meanwhile, extracting the outer boundary line of the building roof, 204, obtaining the vector boundaries of the roof patches, including establishing a search information set, carrying out neighborhood search, determining the types of the outer contour points and determining the vector boundaries, and 205, obtaining a vector model of the building roof, namely obtaining the vector model of the building roof after determining the vector boundaries of the plurality of roof patches. The method is simple in steps, reasonable in design and convenient to realize, good in using effect, and capable of completing the reconstruction processes of the building roof simply, conveniently and quickly with high quality.

The invention discloses a 3D target motion analysis method based on visual and radar information fusion. The method comprises: constructing and training an initial target detection model to obtain a target detection model; obtaining a camera image in real time; detecting the camera image to obtain a 2D frame and a target mask of the target; tracking the target by using a multi-target tracking algorithm to obtain a target id; acquiring laser radar point cloud data in real time; performing joint calibration on the camera image and the laser radar point cloud data to obtain a coordinate conversion relation; projecting the laser radar point cloud data to an image and obtaining point cloud data; filtering the point cloud data to obtain point cloud data only belonging to the target; performing 3D rectangular frame fitting to obtain 3D coordinates of the target; and calculating the speed and the speed direction of the target, and completing the motion analysis of the 3D target. According to the method, the 3D target can be rapidly, accurately and scientifically operated, analyzed and predicted, and the method is high in reliability, good in accuracy and excellent in performance.

The invention discloses a negative obstacle detection method based on the local structure feature of a laser radar point cloud. In collected laser radar point cloud data, three structure features including local point cloud distance hopping, local point cloud dense distribution and local point cloud height decreasing are detected line by line; on the basis of the three structure features, candidate point pairs which may belong to a negative obstacle are obtained through extraction and screening in single line laser point cloud; all candidate point pairs obtained by the laser point cloud of each line laser are clustered on the basis of point pair length consistency and spatial position consistency to obtain a negative obstacle candidate area; and through area filtering and point pair amount filtering, obtaining a negative obstacle area. By use of the method, the negative obstacle in environment can be effectively detected, and the method has the advantages of good detection success rate, small calculation cost and high instantaneity.

The invention relates to a target detection method based on laser radar and image pre-fusion. The laser radar and the camera are combined for calibration; laser radar point cloud data of the visual angle of the camera is obtained; then, before the data is input into the neural network; firstly, spherical projection is performed on a laser radar data set, dense and two-dimensional data is obtained,then an image obtained by laser radar projection and a camera image are taken as two inputs of a network to respectively learn color and three-dimensional features, and finally the features of the two modes are fused through a 1 * 1 convolution block to realize target detection based on radar and pre-vision fusion. The method is different from traditional vision and radar respective detection, finally adopts a weighted post-fusion mode, adopts two corresponding inputs of a laser radar and an image, respectively learns features, and finally performs fusion, so that the accuracy of target recognition can be improved, and the category and three-dimensional information of a target can be obtained at the same time.

The invention provides a three-dimensional bridge reconstruction method based on vehicle-mounted LiDAR point cloud data. The three-dimensional bridge reconstruction method based on the vehicle-mounted LiDAR point cloud data can be used for visualizing an acquired three-dimensional point cloud data implementing three-dimensional model on the bottom of a bridge and comprises the following steps of (1) acquiring the vehicle-mounted LiDAR point cloud data; (2) uniformly diluting the point cloud data so as to reduce data volume; (3) calculating a normal vector, curvature and density of the point cloud data and filtering out noises; (4) registering the point cloud data and diluting the point cloud data; (5) extracting surface plates of the bridge, restraining the surface plates according to priori knowledge and establishing a TIN model; (6) performing TIN model and texture image mapping; and (7) visualizing the three-dimensional model. By the three-dimensional bridge reconstruction method based on the vehicle-mounted LiDAR point cloud data, the surface plates of the bridge can be matched, incomplete data are supplemented, thick scanning data are effectively combined to thin scanning data, and the three-dimensional model of the bridge is established quickly and precisely in real time.

Disclosed is a power line extracting and fitting method based on in-vehicle LiDAR data. The method includes the following steps: first, dividing in-vehicle LiDAR point cloud by using volume elements and determining points in all the volume elements, second, removing the volume elements without the power line point cloud according to distribution characteristics of true power lines, third, dividing the power line point cloud acquired by filtration according to power line corridors and initially clustering the power line point cloud with the AutoClust algorithm, fourth, clustering and merging the initial point cloud belonging to the same power line with a merging cluster method based on end portion fitting line segments, and fifth, recovering fractured power lines according to characteristics of the power lines to obtain point cloud capable of being used for representing a single power line finally and carrying out three-dimensional fitting based on the point cloud. The method can achieve automatic and quick extraction of a large amount of power line point cloud in the in-vehicle LiDAR data, and also can achieve accurate recognition on the single power line and accurate fitting of three-dimensional models of the power lines.

The invention discloses a dynamic obstacle tracking method and system based on sparse laser radar data, and belongs to the technical field of dynamic obstacle tracking, and the method comprises the following steps: S1, eliminating a static background; S2, performing dynamic point cloud clustering; S3, performing convex hull extraction; S4, performing feature extraction; S5, performing data association; S6: predicting a trajectory. According to the method, static obstacle laser radar point clouds are filtered by using a grid map, and extremely sparse point clouds are filtered by a small amountof residual static point clouds through a dbscan algorithm, so the final filtering effect is enhanced; point clouds are classified into ellipses, rectangles and straight lines in a fuzzy mode by extracting angle information of point cloud convex hulls, then correct point cloud position points are obtained by assisting the point clouds in fitting the sizes of graphs, and the accuracy of data association is guaranteed. In addition, the characteristics of the nearest neighborhood and the multi-target hypothesis algorithm are integrated, the multi-target association algorithm is improved, and dataassociation work can be efficiently completed on the premise of ensuring accuracy.

The invention discloses a LiDAR point cloud filter method based on gradient and area growth. The method mainly includes the steps of firstly, acquiring the original LiDAR point cloud data; secondly, rejecting gross errors; thirdly, gridding rules; fourthly, selecting ground seed points; fifthly, performing area growth; sixthly, using the grids where the new ground points satisfying the conditions as centers and using the point, with the lowest elevation, in the grids as the new seed points for growing according to the fourth step until no points satisfy the conditions, and stopping growing; seventhly, sequentially performing area growth on seed points and centralized seed points according to the fourth step and the fifth step until the whole process is finished. By the method, the elevation threshold size is limited effectively, the calculation quantity of the whole method is lowered effectively, and the ground points and non-ground points in LiDAR point cloud data of complex cities can be separated fast. In addition, the method is simple, and the result obtained by the method is relatively small.

The invention relates to an external parameter calibration method, device and apparatus for a laser radar and a binocular camera. The method comprises the following steps of acquiring point cloud dataof the laser radar and image data of the binocular camera; obtaining a corresponding point cloud picture according to the image data; obtaining initial external parameters of the laser radar and thebinocular camera, performing coordinate conversion on the point cloud data according to the initial external parameters to obtain point cloud data under a camera coordinate system, and obtaining the initial external parameters according to the relative attitude of the laser radar and the binocular camera; and performing registration processing according to the point cloud picture corresponding tothe image data and the point cloud data under the camera coordinate system, and obtaining a conversion matrix between the laser radar and the binocular camera according to a registration processing result. Coordinate conversion is carried out on laser radar point cloud data, registration processing is carried out on the converted point cloud data and a point cloud picture corresponding to a binocular camera, a corresponding conversion matrix is obtained, and therefore external parameter calibration of the laser radar and the binocular camera can be carried out more conveniently and accurately.

The invention discloses a multi-line laser radar and multi-path camera mixed calibration method. The method comprises the following steps of S1, collecting the original image data of a multi-path camera, the point cloud data of a multi-line laser radar and the point cloud data of a static laser radar; S2, solving an internal reference model of each camera; S3, subjecting images acquired by each camera to de-distortion treatment and obtaining corrected images; S4, registering the point cloud data of the static laser radar into the point cloud coordinate system of the multi-line laser radar; S5, acquiring the position (Xs,Ys,Zs) of each camera in the point cloud coordinate system of the multi-line laser radar in the point cloud data which are well registered in the step S4; S6, selecting the pixel coordinates (u,v) of at least four target objects in the corrected images of each camera and the corresponding three-dimensional coordinates (Xp,Yp,Zp) of the target objects in the point cloud with the multi-line laser radar as an coordinate origin; S7, according to the internal reference model of each camera, the position (Xs,Ys,Zs) of each camera, the pixel coordinates (u,v) of target objects corresponding to the camera, and the corresponding three-dimensional coordinates (Xp,Yp,Zp), establishing a collinear equation. In this way, the attitude angle elements of each camera and the cosines of the camera in nine directions can be figured out. Therefore, the calibration is completed.

The invention discloses a multiple-divided-conductor automatic extraction and fine modeling method based on LiDAR point clouds. The method comprises the steps that 1, power line and power tower point clouds are extracted from the LiDAR point clouds; 2, according to the characteristics that elevations of power line points are basically the same in a local region and elevations of power tower points vary greatly in the local region, power tower point clouds are further extracted from the power line and power tower point clouds, and the power tower point clouds are removed to obtain power line point clouds; 3, the power line point clouds are subjected to space division to obtain all-phase power line point clouds, a random consistency detection method is adopted to detect noise points in the all-phase power line point clouds, and the noise points are removed; and 4, divided sub-conductor point clouds are extracted from single-phase conductor point clouds based on dichotomy, and the divided sub-conductor point clouds are subjected to catenary fitting. Through the method, the efficiency of a three-dimensional line patrol system can be improved, more precise three-dimensional coordinates can be obtained, and corridor line patrol cost of power lines can be lowered.

The invention discloses a three-dimensional space index method aiming at massive laser radar point cloud models. Aiming at characteristics of massive data and high spatial resolution of point cloud models, the method indexes data to a single three-dimensional space point by adopting a multi-level hybrid index strategy; aiming at an integrity characteristic of point cloud models, the method continuously subdivides massive cloud points from a minimum bounding rectangle (MBB) of an overall point cloud model by adopting a cotree index, and three-dimensional points in cotree index nodes are distributed uniformly in space; and aiming at a scattered characteristic and the data post-processing requirement of large-scale point cloud models, the method indexes a single three-dimensional space point by adopting a three-dimensional K-D tree, and the quick inquiry and processing of a single point coordinate and the attribute data of the single point coordinate can be achieved. On the basis of above integrated space index construction, the method adopts a three-dimensional R tree to manage the MBBs of a plurality of point cloud models in three-dimensional scenes, and a multi-level hybrid space index mode is formed finally.

The invention discloses a 3D vehicle detection method based on multi-sensor fusion, and the method comprises the steps: Step 1, obtaining the semantic information (i.e., an RGB image) of a vehicle through a camera installed on the vehicle, carrying out the scanning of the surrounding environment of the vehicle through a laser radar located at the top of the vehicle, and obtaining the precise depthinformation (i.e., laser radar point cloud) of the environment; Step 2, preprocessing the laser radar point cloud, taking Z-axis [0, 2.5] m according to the height of the automobile, and equally dividing the point cloud into five height slices along the Z-axis direction; Step 3, generating a 3D vehicle region of interest on the laser radar point cloud; Step 4, performing feature extraction on theprocessed radar point cloud and RGB image and generating corresponding feature maps; Step 5, respectively mapping the 3D vehicle region of interest to feature maps of the point cloud and the RGB image; and Step 6, fusing the feature maps of the mapping part in the step 5, and finally realizing 3D positioning and detection of the vehicle target.

Described is a system and method for detecting elevated structures, such as bridges and overpasses, in point cloud data. A set of data from a three-dimensional point cloud of a landscape is received by the system. The set of data points comprises inlier data points and outlier data points. The inlier data points in the three-dimensional point cloud data are identified and combined into at least one segment. The segment is converted into an image comprising at least one image level. Each image level is processed with an edge detection algorithm to detect elevated edges. The elevated edges are vectorized to identify an elevated structure of interest in the landscape. The present invention is useful in applications that require three-dimensional sensing systems, such as autonomous navigation and surveillance applications.

The invention provides a ground object classification method based on electric power corridor airborne LiDAR point cloud dada, and relates to the airborne LiDAR point cloud data processing field. In order to solve the problem of insufficient utilization of information included in the electric power corridor airborne LiDAR point cloud dada in the prior art, the invention provides a ground object classification method based on electric power corridor airborne LiDAR point cloud data features. The method comprises the following steps: obtaining airborne LiDAR point cloud dada; removing gross error points of the point cloud dada; carrying out point cloud dada feature extraction and processing, wherein the point cloud dada features comprise single point feature and neighbourhood feature, and carrying out normalization processing on the extracted features; and selecting samples having labels from a LiDAR point cloud dada set, separating target information of power lines, vegetation, buildings and earth surface and the like in the LiDAR point cloud dada, and classifying the data set by utilizing the samples having labels to obtain class information of power lines, vegetation, buildings and earth surface and the like.