38 results about "Representative point" patented technology

The invention discloses a point cloud semantic segmentation method based on two-dimensional and three-dimensional convolution. Aiming at the problem that the dependency relationship between pixels or points is explored and the relationship between objects is neglected in the existing method, and considering that the variance of ALS point cloud coordinates on an XY plane is greater than the variance along a Z-axis direction, firstly, the representativeness of features obtained by 3D convolution is enhanced by adding 2D convolution; in this way, the distribution of points on the XY plane can be more concerned. In the feature extraction module, the original single 3D KPConv operator in the KPConv is replaced by 2D KPConv and 3D KPConv, and features of more representative points are extracted by two-dimensional and three-dimensional convolution information; at the object level, a spatial channel attention mechanism is applied to account for global information. According to the method, information among all point pairs is obtained by applying a space attention matrix, and the output of two-dimensional and three-dimensional convolution is embedded among the point pairs, so that information gain is carried out, the relation between points and objects is improved, and semantic consistency is enhanced.

The method for predicting a future position of an object comprises capturing a video feed including a two-dimensional (2D) image data; detecting an object in the two-dimensional (2D) image data; obtaining a pose of the object which comprises a plurality of key points; generating a history of poses of the object; determining movement characteristics of two or more key points of the object; obtaining a predicted position data of each of the two or more key points by at least estimating an estimated state of each of the two or more key points based on the movement characteristics and the history of poses; and predicting a future position of at least a predefined representative point of the object from the predicted position data, wherein said future position is predicted on a two-dimensional (2D) image data of the video feed.

To provide a calibration method or the like that can improve the accuracy of identifying a relation among multiple state images based on the multiple state images acquired by a stereo camera at each of multiple different positional attitudes.SOLUTION: A calibration sheet 20 (object) having, on a flat designated surface, designated areas 21, 22, and 23 in which local areas 210, 220, and 230 having three different attributes (RGB values) are arranged so as to be distributed. Three-dimensional positions of respective representative points (centers of gravity) of the designated areas 21, 22, and 23 reflected in each of multiple state images acquired by stereo cameras at different positional attitudes are identified. An affine transformation matrix is obtained so that the three-dimensional position of each representative point in a first state image coordinate system coincides with the three-dimensional position of each representative point in a second state image coordinate system.SELECTED DRAWING: Figure 2

The present invention proposes an optical flow semantic SLAM method for any dynamic object, including: obtaining continuous RGB images, and every eight frames of images, using a deep learning-based optical flow estimation module to estimate the optical flow of the current and the next frame of images to generate an optical flow map; using a saliency region detection module to extract the dynamic region in the optical flow map as a mask, and finding two representative points in the dynamic mask to prompt the semantic segmentation module to perform segmentation, so as to eliminate any dynamic object and remove the interference of the dynamic object to the system; using a lightweight CNN architecture to extract and match feature points from the image with the dynamic region removed, and sending the RGB image and the original depth map after feature matching points and removing the dynamic region into the ORB-SLAM2 system, and improving the accuracy of positioning and mapping by virtue of high-quality feature points. The present invention proposes a more efficient SLAM method, which can have higher positioning and mapping accuracy, and at the same time can have higher robustness in practical engineering.

The present invention provides a three-dimensional target detection method, system and storage medium based on an auxiliary task learning network, including: step one, set abstraction: design a set abstraction layer, the set abstraction layer is composed of a sampling layer, a grouping layer and a point network layer; step two: input the output results of D-FPS and FS in parallel into the voxel feature encoding and self-attention mechanism layer to form an auxiliary task learning network; step three: in the candidate generation layer, use the representative points in F-FPS as the initial center points, the initial center points are transferred to their corresponding instances under the supervision of their relative positions and the correction of the auxiliary network center point estimation, so as to generate new candidate points; step four, data enhancement. The beneficial effects of the present invention are: the target detection of the present invention has high accuracy, good real-time performance, strong generalization ability, and can meet the requirements of online processing and high accuracy of target detection in the application scenario of autonomous driving.

The present disclosure relates to an apparatus and method for object detection. The object detection device comprises a sensor (e.g., LIDAR) and a processor.The processor may extract a plurality of representative points, determine a first reference contour point, a first representative point, a second representative point, and a third representative point from the plurality of representative points, determine a first distance between the first reference contour point and the second representative point; determine a first angle between a first ray extending from the first reference contour point to the first representative point and a second ray extending from the second representative point to the third representative point, determine a fourth representative point, and determine as a contour point the first representative point, the second representative point, or the fourth representative point.

The invention relates to a project supervision quality verification method based on BIM and on-site monitoring data fusion. The method comprises the following steps: constructing a BIM model of a target project; carrying out representative point random sampling on the BIM model, and calculating a quality verification index of a representative point; constructing a representative point set based on all the representative points, wherein the representative point set comprises the representative points and related data thereof; obtaining a target quality verification point outside the representative point set, and obtaining a second three-dimensional coordinate and a quality verification index of the target quality verification point based on the BIM model; and engineering supervision quality verification is carried out on the target quality verification point based on the representative point set and the second three-dimensional coordinates through a preset quality verification model, whether a quality abnormity risk exists is judged, and a result is output. According to the method, the abnormal threshold value can be dynamically adjusted through the spatial relevance between the target quality verification point and the sampling point of the target project, the target quality verification point with the quality abnormal risk can be sensitively captured, and the quality verification result is more accurate.

Aligning a source and target model includes calculating a shape descriptor for a plurality of edge points, grouping the edge points by shape descriptor and selecting representative points for each group. Target and source point pair features (PPFs) are calculated between pairs of representative points on the target and source models, where PPFs defines the relative position and orientation of point pairs. Target PPFs are matched with each source PPF and the point pairs associated with the matched PPFs are transformed to align the location and edge direction of a target PPF point with the location and edge direction of a source PPF point. An angle is determined to align the second target PPF point with the second source PPF point, and a modal angle is found among the determined rotation angles. An output transformation is calculated using the transformations associated with the modal angle.

This position estimation device is configured to: acquire an existence probability, at each of a plurality of representative points, that a positioning target exists; acquire a transition probability indicating a probability that the positioning target moves to each of the plurality of representative points; acquire, on the basis of the existence probability and the transition probability and as a pre-positioning existence probability, an existence probability, at each of the plurality of representative points, that the positioning target exists at the time of positioning for next positioning data of the positioning target; acquire the next positioning data; and acquire a post-positioning existence probability by updating the existence probability, at each of the plurality of representative points, that the positioning target exists at the time of positioning for the next positioning data, on the basis of the next positioning data.

An alignment system (1) for aligning a workpiece (10) comprises: an image processing unit (400) for determining the center of vibration of a representative point of the workpiece (10) on the basis of a captured image of the workpiece (10); and an operation control unit (200) for controlling actuators (111, 112, 113) for moving the workpiece (10). The operation control unit (200) controls the actuators (111, 112, 113) on the basis of the center determined by the image processing unit (400).

A point cloud with attributes is obtained. Its geometry is encoded and decoded. Holes are detected in the decoded geometry and metadata representative of coordinates of the holes are generated. In an embodiment, points are added at these locations to obtain a reconstructed geometry. The attributes of the point cloud are transferred on the decoded or on the reconstructed geometry and, the decoded geometry, the transferred attributes, and the metadata are encoded in a data stream. At the decoding side, a data stream representative of a point cloud is obtained. The geometry is decoded. In an embodiment, metadata are used to add point to the decoded geometry. The attributes of the point cloud are decoded the decoded or reconstructed geometry. In the first case, metadata are used to add points to the point cloud and their attributes are estimated according to the decoded attributes.

The invention discloses a slicing method and system, computer equipment, a computer readable storage medium and a computer program product, and the slicing method comprises the following steps: creating a plurality of continuous layers capable of completely wrapping an implicit expression model; and selecting any number of continuous layers from the plurality of layers to form a subspace so as to carry out step-by-step gridding on the subspace, which comprises the following steps of: recursively executing to respectively determine a value interval of implicit expression of each subgrid of the current level so as to carry out next-level gridding on the subgrids of which the value interval comprises a threshold interval, each sub-grid of the current level is a single-layer grid meeting the corresponding layer resolution; and selecting a sub-grid of which the value interval of the implicit expression comprises the threshold interval from the single-layer grid meeting the corresponding layer resolution as a key grid, and filling an output value of the implicit expression of at least one representative point in the key grid, so as to determine the slice contour of each selected layer.

This application provides an assembly method and apparatus based on contact surfaces. When assembling precision-machined parts, the method places a movable surface in a non-contact state at the initial assembly position of a fixed surface, determines a representative point of a convex peak region and a fixed representative point in the fixed surface; obtains any one of the convex peak regions to determine a first contact point; determines a second contact point based on the convex peak representative point and the first contact point; determines a third contact point by rotating the line connecting the target intersection point and the second contact point; checks whether the triangular region formed by the first, second, and third contact points meets the assembly conditions. If it does, a pose transformation matrix is ​​output to assemble the movable and fixed surfaces according to the pose transformation matrix; otherwise, the third contact point is eliminated according to a set elimination rule, and the third contact point is re-determined until the assembly conditions are met. Therefore, this application can meet the assembly requirements of precision-machined parts.

The invention provides an automatic endoscope following method and device, electronic equipment and a storage medium, and relates to the technical field of automatic control. The method comprises the following steps: identifying a plurality of instruments in an endoscope image; based on the endoscope image, obtaining position information of the instrument in the endoscope image; based on the position information of the multiple instruments, comprehensive representative points of the multiple instruments are obtained; and if the comprehensive representative point deviates from the center of the visual field of the endoscope, performing visual angle following adjustment on the endoscope. The method is used for dynamically balancing the influence of multiple instruments on the view, ensuring that key instruments are always in the center of the view, and solving the view management problem in a multi-instrument scene in the prior art.

In some implementations, a device may receive an image that depicts an environment associated with a vehicle. The device may partition the image into a plurality of subsections. The device may analyze the plurality of subsections to determine respective subsection information, wherein subsection information, for an individual subsection, indicates: a probability score that the subsection includes a line segment associated with an object class, a position of a representative point of the line segment, and a direction of the line segment. The device may identify, based on the respective subsection information of the plurality of subsections, a line associated with the object class that is associated with a set of subsections of the plurality of subsections. The device may perform one or more actions based on identifying the line associated with the object class.

In various examples, an environment surrounding an ego-object is visualized using an adaptive 3D bowl that models the environment with a shape that changes based on distance (and direction) to one or more representative point(s) on detected objects. Distance (and direction) to detected objects may be determined using 3D object detection or a top-down 2D or 3D occupancy grid, and used to adapt the shape of the adaptive 3D bowl in various ways (e.g., by sizing its ground plane to fit within the distance to the closest detected object, fitting a shape using an optimization algorithm). The adaptive 3D bowl may be enabled or disabled during each time slice (e.g., based on ego-speed), and the 3D bowl for each time slice may be used to render a visualization of the environment (e.g., a top-down projection image, a textured 3D bowl, and / or a rendered view thereof).

A method for calibration of a microscope-based-system includes generating a path, defined by movement controls of start points, for controlling an illuminating position with scanning mirrors directing an illuminating light; opening a camera shutter; illuminating with the illuminating light after a first waiting time for each movement control; obtaining an image; repeating "opening the shutter" to "obtaining image" twice, with a third waiting time longer than a second which is longer than the first, obtaining a second and third image; identifying representative points of the first to third image; calculating offsets of each movement control by comparing the representative points of the first and second image with the third; determining stable times of each start point, that are the shortest waiting times for the corresponding image offset to meet a predetermined criterion; fitting a curve by a mathematical model; and setting the model in the system.

The present disclosure relates to an object recognition apparatus and method. The object recognition apparatus includes a sensor (e.g., LIDAR) and a processor. The processor may extract a plurality of representative points, determine a first reference contour point, a first representative point, a second representative point, and a third representative point among the plurality of representative points, determine a first distance between the first reference contour point and the second representative point; determine a first angle between a first ray extending from the first reference contour point to the first representative point and a second ray extending from the second representative point to the third representative point, determine a fourth representative point, and determine, as a contour point, one of the first representative point, the second representative point, or the fourth representative point as the contour point.

Provided is an information processing device which causes an AI model that has learned a first task to continuously learn a second task. The information processing device: classifies, via an unsupervised clustering method, a plurality of pieces of first training data used for the learning of the first task into a prescribed number of clusters; selects, from among the plurality of pieces of first training data, a prescribed number of pieces of first training data that are closest to respective representative points of the prescribed number of clusters; stores, in a buffer, the prescribed number of pieces of first training data that have been selected; and uses the prescribed number of pieces of first training data stored in the buffer and a plurality of pieces of second training data to be used in the learning of the second task, to cause the AI model to continuously learn the second task.

A method for predicting a future position of an object includes capturing a video stream, the video stream including two-dimensional (2D) image data; detecting an object in the two-dimensional (2D) image data; obtaining a posture of the object comprising a plurality of key points; generating a posture history of the object; determining movement characteristics of two or more key points of the object; predicted position data for each of the two or more keypoints is obtained by estimating at least an estimated state of each of the two or more keypoints. Based on the movement characteristics and the attitude history; and predicting a future position of at least a predefined representative point of the object from the predicted position data, where the future position is predicted for two-dimensional (2D) image data of the video stream.

A point cloud with attributes is obtained. Its geometry is encoded and decoded. Holes are detected in the decoded geometry and points are added at these locations to obtain a reconstructed geometry. The attributes of the point cloud are transferred on this reconstructed geometry and, the decoded geometry and the transferred attributes are encoded in a data stream. At the decoding side, a data stream representative of a point cloud is obtained. The geometry is decoded, and holes are detected to obtain a reconstructed geometry. The attributes of the point cloud are transferred on this reconstructed geometry to obtain a point cloud with attributes.

A non-transitory computer-readable storage medium storing an estimation program that causes a computer to execute a process includes specifying representative points of each of training clusters that corresponds to each of labels targeted for estimation; setting boundaries between each of input clusters under a condition that a number of the input clusters and a number of the representative points coincide with each other, the input clusters being generated by clustering in a feature space for input data; acquiring estimation results for the labels with respect to the input data based on a correspondence relationship between the input clusters and the training clusters based on the boundaries; and estimating determination accuracy for the labels by using the machine learning model with respect to the input data based on the estimation results.

An alignment system (1) comprises: an image processing unit (400) for obtaining the center of vibration of a representative point of a workpiece (10) on the basis of a captured image of the workpiece (10), and determining whether the obtained center is within the accuracy reference range of a predetermined representative point; and an operation control unit (200) for controlling drive devices (111, 112, 113) that move the workpiece (10). The operation control unit (200) controls the drive devices (111, 112, 113) on the basis of the result determined by the image processing unit (400).

Provided are an information processing system, an information processing method, and a program that achieve update of a classification algorithm capable of achieving improvement in accuracy. The information processing system comprises a processor and a memory in which an instruction to be executed by the processor is stored. The processor executes the instruction to: select, from among a plurality of feature points respectively representing the feature amounts of a plurality of images, at least one first representative point for each label attached to the images on the basis of the feature amounts by using a classification algorithm; correct the label of the first representative point on the basis of a determination result of whether the label of the first representative point is correct; and update the classification algorithm on the basis of a classification rule corrected according to the correction of the label of the first representative point.

The application belongs to the technical field of robots, and particularly relates to a map exploration method and device, a computer readable storage medium and a robot. The method comprises the following steps: screening a first boundary contour according to a preset first path width; the first boundary contour is a boundary contour of an unknown region and a first passable region, and the first passable region is a known region determined by the first path width; searching for an exploration representative point of the first boundary contour; the exploration representative point of the first boundary contour is located in the first passable region; if the exploration representative point of the first boundary contour is searched, a first target exploration representative point is selected; the first target exploration representative point is an exploration representative point of the first boundary contour with the highest priority; and the robot is controlled to move to the first target exploration representative point to perform map exploration. Through the above method, the exploration representative point with the highest priority can be selected to perform map exploration, so that blind exploration can be reduced, and the efficiency of map exploration can be effectively improved.

This invention relates to the field of machine vision technology, specifically to a method and system for positioning the rim of paper cups in a stacked paper cup production line. The invention acquires consecutive frame images of each paper cup; selects representative points of the cup rim contour in each frame; determines the directional deviation angle of each representative point based on its displacement vector in adjacent frames; obtains the sliding influence coefficient of the cup rim position in each frame based on the displacement vector and directional deviation angle; analyzes the fluctuation amplitude of the cup rim position sliding in all frames to determine the degree of sliding influence of each rim; combines this with the hysteresis time to determine the degree of hysteresis deviation of the cup rim position in each frame; and determines the degree of positional mismatch of each rim based on the degree of sliding influence and the degree of hysteresis deviation, thereby determining the correction degree for each rim and correcting its position. This invention can improve the positioning accuracy of paper cup rims and ensure precise alignment in subsequent processes.

The application discloses a kind of map data processing method, device, medium, equipment and cloud platform, belong to map making technical field.The method comprises the following steps: obtaining at least one object surface in map intermediate data and at least one object surface corresponding adjacent surface, at least one object surface and at least one object surface corresponding adjacent surface are simulated combination and obtain simulated combination surface;Simulated combination surface with multiple closed loops is identified as the multi-face with cavity, the closed loop surrounded area inside the multi-face is identified as cavity, and the gap of at least one object surface is determined according to the cavity;The coordinates of one internal point of the gap are extracted as the representative point coordinates of the gap, and the representative point coordinates are de-duplicated to obtain the gap position.The application avoids the problems of large workload, low accuracy, easy redundancy identification and easy omission when identifying the gap in the prior art, improves the efficiency and accuracy of gap identification.

It accurately and reliably recognizes the position and orientation of objects. [Solution] The recognition devices 201 and 202 include: a feature region detection unit 21 that detects at least three feature regions 62 of an object from imaging data 67 obtained by imaging the object 61; a representative point extraction unit 23 that calculates extracted representative points 69 from the imaging data and feature regions and extracts at least three extracted representative points and extracted label information 64 for identifying each extracted representative point; and a position and orientation estimation unit 24 that estimates the position and orientation of an object by aligning the extracted representative points with the reference representative points and the extracted representative points, where the extracted label information corresponding to the extracted representative points and the reference label information 64 for identifying at least three reference representative points having a predefined correct positional relationship coincide.

The present application relates to the technical field of data processing, and particularly relates to a POI recommendation method without leaking position information, comprising: establishing N representative points, disclosing the position coordinates and number of the representative points, and obtaining the representative point number m by a mobile terminal; constructing a storage agent; generating a plurality of pairs of public and private keys; asking the storage agent for an association number s by the mobile terminal, and asking a server for at least two public keys; generating an initial vector by the mobile terminal, and randomly selecting a public key; calculating a feature vector and sending the feature vector to the server; respectively decrypting the feature vector by the server using the private key; searching for a corresponding point of interest by the server, and including the point of interest into a result set; associating the result set with s, and sending the result set to the storage agent; and storing the result set in a storage space corresponding to s by the mobile terminal, and asking the storage agent for the result set in the storage space corresponding to s. The substantial effect of the present application is that the position information of the mobile terminal is protected when the point of interest list is obtained.

The alignment system (1) has: an image processing unit (400) that, based on a captured image of a workpiece (10), calculates a center of vibration of a representative point of the workpiece (10), and determines whether the calculated center is within a range of a predetermined precision reference for the representative point; and a motion control unit (200) that controls a driving device (111, 112, 113) that moves the workpiece (10). The motion control unit (200) controls the driving device (111, 112, 113) based on a result of the determination by the image processing unit (400).