91 results about "Minimum bounding box" patented technology

The invention relates to a range image-based 3D spatial data processing method and a device. In the method, ground laser radar is adopted to scan an object to obtain 3D point cloud data on the surface of the object; a database object table is established by a PL/SQL programming approach; a reference datum is fit according to point cloud which is segmented in advance to generate a minimum bounding box (MBB) of the point cloud and subsequently generate the range image according to the reference datum obtained by fitting. Later, data of the MBB and the range image are written into the database object table and the relation between the MBB and a range image object is established, subsequently, R tree 3D spatial database index is established for the database objects which enter the database, finally, visualization is realized based on Oracle Objects for OLE (OO4O) and OpenGL.

A method for detecting moving objects includes: (a) capturing and establishing a background image; (b) capturing at least one current image; (c) transforming the background image and the current image from an RGB color format into an HSI color format; (d) subtracting the background image from the current image according to a background subtraction rule for generating at least one moving object; (e) performing a vertical scanning and a horizontal scanning on the moving object for generating a minimum bounding box of the moving object; (f) calculating a characteristic datum of the moving object according to the minimum bounding box; (g) tracking the moving object according to the characteristic datum with a Euclidean distance rule; (h) classifying the moving object according to the characteristic datum, the tracking result generated by step (g) and a minimum distance classifier.

A mobile communication terminal that comprises a body having a minimum bounding box with a wide dimension that includes an image sensor and an imaging unit. The body comprises an image-capture aperture substantially perpendicular to the wide dimension. The imaging unit is configured for imaging an image captured via said image-capture aperture on said image sensor.

The invention discloses a robot autonomous classification grabbing method based on YOLOv3, and the method is characterized in that the method comprises: collecting and constructing a target object sample data set; training the YOLOv3 target detection network to obtain a target object recognition model; acquiring a color image and a depth image of the target object; processing the color image by using the trained YOLOv3 target detection network to obtain the category information and the position information of the target object to be grabbed, and further processing by combining the depth imageto obtain the point cloud information of the target object; and performing minimum bounding box solving on the point cloud information, calculating the main direction of the point cloud by combining aPCA algorithm, calibrating the coordinate data of the X, Y and Z axes of the target object, and calculating the six-degree-of-freedom pose of the target object relative to the robot coordinate system. According to the method, the YOLOv3 algorithm is adopted, the object grabbing pose is estimated through point cloud preprocessing, PCA and other methods, and then the robot grabs the target objectsin a classified mode.

The invention discloses a method of extracting a three-dimensional axis of a tunnel based on a minimum bounding box algorithm. The method comprises the following steps of: extracting point cloud slices of a lining ring by virtue of the minimum bounding box algorithm of the tunnel with variable-angle step length; then, fitting the extracted point cloud slices according to an oval or circular model to solve coordinates of a center, and performing smoothing processing on the tunnel axis by adopting a horizontal-curve second-order continuous and a longitudinal-curve second-order continuous method; and finally, obtaining high-precision three-dimensional central axis of the tunnel. The method disclosed by the invention has the advantages that the three-dimensional axis of a single targe structure is automatically extracted by utilizing three-dimensional laser scanning original-point cloud data based on the minimum bounding box algorithm; the method is high in precision, high in speed, and capable of being used in tunnel axis detection, diameter convergence extraction, production of orthoimages, and the like. Besides, the defect that the tunnel design axis or actually measured axis is needed to be pre-input in the conventional processing method can be avoided, so that the automation degree of the data processing is greatly improved.

The present invention relates to a method for generating a mesh model representing a 3D surface from unorganized 3D points extracted from a 3D scanner by using a shrink-wrapping scheme of boundary cells. A method for generating 3-dimensional mesh according to the present invention comprises the steps of: (a) receiving unorganized 3D point coordinates extracted by a 3D scanner or a digitizer; (b) extracting a minimum bounding box including all the point coordinates and uniformly dividing the extracted bounding box into cells of a predetermined size; (c) extracting a boundary cell including at least one point from the cells, extracting a boundary surface from all the boundary cells, and generating an initial mesh by summing extracted boundary surfaces; (d) calculating distances between each vertex constituting the mesh and the several points, finding a nearest point, and moving the vertex to the nearest point; and (e) averaging location of each shrink-wrapped vertex and location of the neighboring vertexes, and moving the shrink-wrapped vertex to center of neighboring vertexes.

The invention discloses a method for reconstructing a scattered point cloud with boundary constraint. The method comprises the steps of: firstly mapping a three-dimensional scattered point cloud to a two-dimensional plane, obtaining a minimum bounding box for the obtained two-dimensional point collection and then dividing the bounding box into multiple rectangular grids, judging whether the grid is a boundary grid based on existence of a two-dimensional point in an adjacent grid of each grid, extracting a two-dimensional boundary point from each boundary grid and then obtaining boundary points of the three-dimensional scattered point cloud through the mapping relationship; removing the boundary points in the three-dimensional scattered point cloud and then simplifying rest point cloud by utilizing a fuzzy entropy iteration based point cloud simplified algorithm, and combining the simplified rest point cloud with the boundary points of the point cloud and then composing a simple point cloud; performing triangulation on the two-dimensional point collection corresponding to the simple point cloud by utilizing a Delaunay criterion based optimized algorithm and then mapping the triangulation result to a three-dimensional space to finally realize curved surface reconstruction of the three-dimensional scattered point cloud.

The invention discloses a convolutional neural network RGB-D significance detection method based on multilayer fusion. The method comprises the following steps: converting full connection layer modules FC6 and FC7 in a VGG16 network into full convolution layers, and combining the full convolution layers CONV1-CONV5 to form a new convolutional neural network; Sequentially carrying out dimension reduction and fusion operation on the new convolutional neural network to obtain an initial iteration significance detection result; Refining the saliency detection result of the initial iteration by adopting iterative optimization; Adopting different training data to sequentially carry out initialization training and first and second fine tuning training on the new convolutional neural network; Andfor the result after training, using the minimum bounding box of the saliency object to perform cutting and mirror overturning on the input image to obtain a saliency detection result. According to the method, an effective CNN model is designed, RGB and depth information is fused, multi-scale features of a significant object are captured, and convolution features from a higher layer to a lower layer are fused, so that the scale problem of the significant object is solved.

A cutting apparatus is disclosed in which a cutting blade and an object to be cut are moved relative to each other so that a desired pattern is cutout of the object. The cutting apparatus includes an arranging unit arranging the pattern in a cut-allowable region of the object, a frame setting unit setting a minimum boundary frame which is polygonal or curved in shape and includes a contour of the pattern arranged by the arranging unit, and a cutting data generating unit generating outer line cutting data for cutting an outer line dividing a first region near the pattern within the cut-allowable region and a second region other than the first region, outside the boundary frame, based on the boundary frame. The pattern and the outer line are cut out of the object based on pattern cutting data for cutting the pattern and the outer line cutting data.

The invention discloses a rapid acquisition and calibration method of an image data set in deep learning. After electronic image stabilization processing is carried out on a collected video, denoisingand smoothening are carried out by median filtering, and foreground extraction is performed by using a Gaussian mixture model; according to the position of an outer profile of the extracted foreground, corresponding position information in an original video frame is located to obtain a minimum bounding box of a specific target automatically; foreground target classification of the video frame iscarried out by using a hog-feature-based local template matching algorithm; with an inter-frame matching algorithm, same-class marking is carried out on a corresponding connection region between adjacent frames; and the position information and type information of the foreground are obtained to realize rapid calibration of an image data set. Therefore, problems that only a few of data carry tags in deep learning and new data with tags are obtained difficulty with high cost are solved; and the great convenience is provided for training and identification of the specific type of image data.

The invention discloses a texture dynamic organization method for a fine three-dimensional model. The method comprises the following steps of: opening a three-dimensional scene model, and solving the size of a minimum bounding box of a three-dimensional model; establishing an index tree based on an octree and specifying a maximum recursion depth; recording space coordinates x, y and z of leaf nodes occupied by each part in a building so as to generate a Morton retrieval code of a texture of each part; extracting root nodes of the index tree established by the three-dimensional model; carrying out texture file association on the index tree by use of a dynamic association method; judging whether texture files which have the same texture name and the same resolution exist in brother nodes with the same father node; and repeatedly and circularly traversing and judging until the establishment of the index tree and dynamic association and aggregation of all nodes are completed, and finally outputting a texture index tree with clear hierarchy and good organization and an index file of the texture index tree.

The invention provides a virtual measurement principle and method based on augmented reality. The virtual measurement principle of minimum bounding boxes based on augmented reality comprises a reality measurement environment model, an augmented reality measurement environment model and a human-computer interaction model. The virtual measurement realization method of minimum bounding boxes based on augmented reality comprises visual scanning tracking technology, virtual ruler measurement technology and human-computer interaction interface technology. The invention can be used for contactless measurement and calibration of the minimum bounding boxes of spatial three-dimensional objects, has the advantages of simple concept, flexible structure, convenient application and the like and is augmentation and expansion of the existing manual or instrument measurement modes.

The invention relates to a CNN-based automatic identification method of lepidopteron types. During preprocessing, background removing is carried out on a collected insect specimen image; a minimum bounding box of a foreground image is calculated; and an effective foreground area is obtained by cutting out. Feature extraction is carried out by using an Imagenet pre-training depth learning neural network model. And classification and identification are carried out based on two kinds of conditions; when samples are sufficient, parameters of a DCNN classification layer are trained based on a fine tuning network structure, thereby realizing end-to-end classification and identification; and when a sample data set is small and insufficient samples are provided for training DCNN parameters, an X<2> kernel SVM classifier for a small sample set is used for realizing classification and identification. The automatic identification method has advantages of simple operation, high identification precision, high fault tolerance, good time performance and the capability of improving lepidopteron type identification obviously.

The invention relates to a rapid sectioning method for a high-level radioactive waste repository 3D mesh model, belonging to the technical field of 3D modeling. The 3D rapid sectioning of any cross section of the high-level radioactive waste repository 3D mesh model based on tetrahedron construction can be supported. According to the method, through establishing the minimum bounding box of the high-level radioactive waste repository 3D mesh model, the mesh model is subjected to adaptive regional division, the rapid search of the divided areas of the repository model is carried out according to section information, the divided areas are determined, in mesh sectioning, the space information and topology information of the sectioned mesh are subjected to synchronous processing, and thus the geometric integrity of the model is maintained. According to the method, any times of sectioning of any cross section of the high-level radioactive waste repository 3D mesh model can be supported, the cross section area generation speed is rapid, when the tetrahedron quantity in the high-level radioactive waste repository 3D mesh model is larger than 1 million and is smaller than 5 million, the cross section generation time is smaller than three seconds, and the repository model can be dynamically refreshed in real time.

The invention discloses a graph pretreatment method for removing negative OPC (Optical Proximity Correction). The method comprises the following steps of setting a short hypotenuse size selection standard; screening graphs with short hypotenuses in a graph segment, selecting a target graph meeting the short hypotenuse size selection standard; expanding the short hypotenuse in the target graph towards the inside of the target graph to form a rectangle; expanding outside the rectangle to form a minimum bounding box region, wherein the minimum bounding box region can cover the short hypotenuse in the target graph; combining the target graph with the minimum bounding box region to remove the short hypotenuse from the target graph. In the method, by expanding the short hypotenuse and finally combining the minimum bounding box region with the target graph, the short hypotenuse graph negative for the OPC is removed, after the short hypotenuses are removed, and the target graph is changed into an orthogonal right-angle side graph, the result of the OPC can be consistent with the expectation effect well.

The invention provides a point cloud repair method and system. The method comprises introducing point cloud data of a product, receiving point spacing and a curvature filtering parameter which are input by a user, determining a to-be-repaired area in the point cloud data and the minimum bounding box of the to-be-repaired area, selecting a preset number of points in the minimum bounding box and outside the to-be-repaired area, fitting the selected points into a B-spline surface, performing interpolations on the B-spline surface according to the point spacing to obtain a plurality of discrete points, filling the to-be-repaired area with the discrete points, determining and deleting unorganized points of the to-be-repaired area according to the point spacing and the curvature filtering parameter, and performing smoothing processing on the point cloud data after filling when other to-be-repaired areas do not exist in the point cloud data. High-precision repairing of a missing point cloud can be achieved.

The invention provides an automatic labeling method and device for a target object. The automatic labeling method comprises the steps of determining a first labeling box according to a first projection of a target object on a first projection plane in first point cloud data, obtaining second point cloud data in the first labeling box, determining a second labeling box and a third labeling box according to the second projection of the target object on the second projection plane and the third projection of the target object on the third projection plane in the second point cloud data, determining a minimum bounding box coordinate of the target object according to the first labeling box, the second labeling box and the third labeling box, and displaying the minimum bounding box of the targetobject in the first point cloud data according to the minimum bounding box coordinates. According to the embodiment of the invention, the target object can be automatically labeled in the laser pointcloud without manual labeling.

The invention discloses a small-scale target detection method based on a weak edge, and the method comprises the following steps: 1), establishing a weak edge detection network: enabling a backbone network to be divided into five stages based on VGG16, and obtaining the features of different scales; fusing the results of each stage to obtain a final result of edge detection; 2) constructing loss functions of the trunk network and the bypass branches, and performing weighted summation on the auxiliary loss functions of the plurality of branches and the loss functions on the fusion features; 3)connected region analysis: carrying out binarization processing on the generated edge images; determining a minimum bounding box, and generating a multi-instance suggestion area; and 4) detecting a small-size target by using the RPN, summarizing the detection results of all multi-instance suggested regions, and further adjusting the target detection result by using the Fast R-CNN. According to themethod, a good detection result is obtained when the situation that the target is small in the image and the situation that the targets are mutually adhered and shielded are processed.

The invention relates to a rapid algorithm for a quasi-rectangular object bounding box, and the algorithm comprises the following steps: 1), reading an object, and obtaining the ID of the object; 2) traversing all feature IDs of the object; 3) constructing a Z-axis vector and a Y-axis vector; 4) performing cross multiplication on the Z-axis vector and the Y-axis vector to obtain an X-axis vector;5) constructing a temporary coordinate system: 6) calling an AABB bounding box in the temporary coordinate system to obtain the minimum bounding box of the object, and outputting the sizes of the bounding box in the X, Y and Z directions and the coordinates of the left lower corner and the right upper corner of the box; and 7) setting the temporary coordinate system as WCS, and drawing and displaying the bounding box according to the parameters of the bounding box. According to the method, the defects that an existing AABB algorithm, an OBB algorithm and other algorithms are low in efficiency,complex in processing process and the like are overcome, the calculation time of a computer is shortened, the calculation time of the bounding box is controlled within 0.1 s, and engineering application of the algorithms is facilitated.

The invention is applicable to the technical field of robot spraying, and provides a robot spraying trajectory generating method based on an ROS platform. The method includes the following steps of reading triangular mesh information in an stl model, conducting region segmentation on the stl model on the basis of the continuity among triangular patches, connecting the designated adjacent segmentedregions into a spraying assembly, obtaining the minimum bounding box of the spraying assembly, generating a group of planer clusters parallel to a calibrating face with any section of the minimum bounding box as the calibrating face, sequentially obtaining the intersection points of all planes and triangular mesh of the spraying assembly, sequentially connecting the intersection points to form spraying routes, adding trajectory points to reverse extension cords of the two ends of each spraying route respectively, and connecting the trajectory point of the tail end of the front spraying routewith the trajectory point of the head end of the rear spraying route through a straight line. By means of the spraying trajectory generating method, the robot spraying trajectory can be automaticallygenerated only by manually setting part of parameters, the automation degree is high, and work efficiency is improved.

The invention relates to a super-resolution restoration method based on a monitoring video. The method comprises the following steps: (1) reading and preprocessing a monitoring video: reading the video by applying Matlab software, dividing the video into a form of continuous sequence images, performing Gaussian blur processing on the obtained sequence images, performing down-sampling, and synthesizing the down-sampled sequence images into video frames to simulate and obtain a reduced-quality low-resolution video; and (2) performing moving object detection and segmentation: detecting a moving object by using a three-frame difference method, designing a minimum bounding box, extracting the moving object, and enabling the sizes of output images to be consistent; and (3) only providing high-resolution information for the moving target, namely separating the moving target from the background, adopting a convex set projection method for the moving target, and directly performing bilinear interpolation processing on the background.

Techniques for associating a graphics object with one of a plurality of target regions of a graphical object template are disclosed. The graphics object is associated with one of the target regions based on the distance between the center of mass of the graphics object and the center of mass of the target region and the overlap amount of the graphics object with the target region. A pre-generated graphics object (stamp object) may also be automatically incorporated into one of the target regions based on the alignment of the minimum bounding box of the stamp object and the minimum bounding box of the target region.

The invention relates to a three-dimensional model generation method based on genetic manipulation. The method includes the steps: inputting a three-dimensional model set; selecting optional one three-dimensional model from the three-dimensional model set, and marking the three-dimensional model as a three-dimensional model M1; randomly selecting one part of the three-dimensional model M1, and marking the part as a part P1; based on the part P1, calculating corresponding values of parts of the three-dimensional models, except the three-dimensional model M1, with respect to the part P1; collecting the parts with the corresponding values, with respect to the part P1, larger than 0.9 to form a corresponding part set related to the part P1; calculating minimum bounding box for each part in the corresponding part set to obtain a minimum bounding box set of the parts corresponding to the part P1; calculating the range of allowable deformation of the part P1 according to the minimum bounding box set; based on the range of allowable deformation, deforming the part P1; expanding the deformation to other parts of the three-dimensional model M1 to obtain new three-dimensional models. The method taking advantages of genetic variation is simple, and the three-dimensional model data set is enriched.