24 results about "Distance sampling" patented technology

The invention discloses a multi-source information collaborative power equipment three-dimensional temperature field construction method. Firstly, an infrared camera, an IMU and a laser radar are utilized to obtain an accurate external parameter relation through joint calibration, point cloud distortion is eliminated, angular points and plane points are extracted, a re-projection residual error, a distance sampling residual error and an IMU pre-integration residual error are constructed, an error state iteration Kalman filter is adopted to optimize a global pose, and positioning is achieved. Providing a self-supervised depth completion network, combining an infrared temperature image and a sparse depth map generated by a laser radar as input, adopting a depth completion strategy guided by an infrared image, estimating relative motion of adjacent frames by using pose information, introducing a feature alignment module to reduce alignment errors, and combining the depth map, the infrared image and IMU data to obtain a self-supervised depth completion algorithm; and efficient construction of the three-dimensional temperature field of the power equipment is realized. According to the invention, the three-dimensional temperature field of the power equipment is constructed more accurately, and the capability of the substation inspection robot for state monitoring and fault diagnosis of the power equipment is improved.

The present invention proposes a lightweight industrial environment defect detection method based on a feature memory bank. First, the top view / left view or right view of qualified products on the production line is photographed, preprocessed after being sorted into a data set. Secondly, the data set is input into a pre-trained standard teacher model and an initialized lightweight student model to perform feature knowledge distillation training on the student model. After that, the defect-free product data set is input into the lightweight student model trained by feature knowledge distillation. After extracting the corresponding feature set of the product image and performing multi-scale feature set fusion, maximum core set distance sampling is carried out to establish a feature memory bank of qualified products on the current production line. Finally, the production line images to be detected are input into the lightweight student model and multi-scale feature fusion is performed, and then compared with the feature memory bank by the nearest neighbor algorithm to detect whether the current product contains defects. The present invention can improve the efficiency of anomaly detection without sacrificing accuracy as much as possible.

The invention relates to a point cloud downsampling method with any sampling size, which belongs to the field of point cloud data, and comprises the following steps: S1, inputting original point cloud data, and determining a target sampling size range; s2, in the training process, each Epoch dynamically selects the sampling size; s3, generating offset through a point cloud feature extraction module, and generating a final sampling point cloud in combination with an original sampling point obtained by farthest distance sampling FPS; s4, calculating a point-to-point distance value between the sampling point cloud and the original point cloud, and taking the point-to-point distance value as similarity loss; s5, inputting the sampling point cloud into a downstream task network, and calculating task related loss; and S6, repeating the steps S1 to S5 until the model converges. Through one-time training, point clouds of any size can be sampled, limitation of fixed sampling size is avoided, and multi-scale feature learning of the point clouds is realized through different-dimension learning of point cloud features.

The embodiment of the invention provides an outer wall hollowing and falling risk prediction method and system based on infrared detection, and the method comprises the steps: obtaining the wall materials of a to-be-detected building outer wall at different time points based on a preset time interval; acquiring a plurality of sampling distances corresponding to the wall material from a preset and determined mapping relation library according to the wall material of each time point in different time points; based on the plurality of sampling distances, performing physical sampling on the to-be-detected building outer wall to generate a multi-distance sampling image; based on the multi-distance sampling images, image fusion is carried out, and wall body temperature field images of the to-be-detected building outer wall at different time points are generated; determining a temperature field image sequence based on the wall temperature field images at different time points; based on the temperature field image sequence, outer wall risk detection and risk grade division are carried out, and the current risk grade of the to-be-detected building outer wall is determined. According to the scheme, the accuracy of outer wall hollowing and falling risk prediction can be improved.

The invention discloses a modular XL-MIMO near-field beam training method, which belongs to the technical field of XL-MIMO beam training, and comprises the following steps: activating antennas of a head module and a tail module in an array to carry out far-field beam scanning, determining a code word index with the maximum beam gain of the head module and the tail module, estimating the code word index with the maximum beam gain of the whole array, and taking the code word index with the maximum beam gain as a center, the method comprises the following steps: selecting a plurality of candidate angles, carrying out non-uniform distance sampling by adopting a polar domain codebook, generating candidate sampling points, distributing each candidate sampling point to a collaborative module for training, determining the energy of each candidate sampling point, further determining an optimal polar domain near-field codeword index, and completing modular XL-MIMO near-field beam training. The method not only can effectively reduce the training overhead, but also can maximize the accuracy of beam gain and angle estimation under the condition of limited resources.

The invention discloses a sight tracking method and system based on cascade deep learning and distance correction, and a storage medium. The method comprises the following steps: processing an eye image through a cascade deep learning model, and obtaining pupil center coordinates and cornea reflection light spot coordinates of two eyes; calculating a binocular pixel distance as a distance correction factor based on the pupil center coordinates, and calculating a relative position difference vector based on the pupil and light spot coordinates; establishing a three-dimensional nonlinear polynomial regression model taking the relative position difference vector and the distance correction factor as input; obtaining calibration data through a multi-distance sampling strategy and solving a model coefficient; and in the real-time tracking stage, calculating the coordinates of the fixation point of the screen by using the solved model. According to the method, the distance correction factor and the three-dimensional mapping model are introduced, so that geometric errors caused by forward and backward movement of the head of the user are effectively eliminated, meanwhile, the feature point positioning precision is ensured through cascaded deep learning detection, and the precision and robustness of a sight line tracking system are remarkably improved.

The invention discloses a mixed field channel representation and acquisition method based on a main lobe energy coverage criterion. The method comprises the following steps: S1, determining a Rayleigh distance and a minimum sampling distance based on communication system parameters; s2, determining an angle domain sampling interval, and generating angle sampling points according to 3dB power overlapping; s3, deriving and applying a 3dB distance recursion analytic expression, determining a self-adaptive distance sampling interval and iteratively generating distance sampling points at each angle according to a nonlinear attenuation rule of beam focusing, and ensuring that main lobe energy coverage overlapping of adjacent dictionary code words in a distance domain meets a preset threshold value; and S4, gathering the beam atoms corresponding to all angle and distance sampling points to obtain a final dictionary as mixed field channel representation. According to the invention, the design of the code words in the adjacent angle and distance domains meets the condition that the power overlap is not less than 3dB, so that the channel representation blind area is greatly reduced in the whole angle and distance domain range.

The invention relates to the technical field of rock sampling equipment, in particular to geological field grooving sampling equipment, which comprises a supporting guide rail, a sampling device, a sampling device, a sampling device, a sampling device, a sampling device, a control device and a control device, and is characterized in that the supporting guide rail is laid on a sampling surface in a mine and is used for limiting a cutting stroke; the movable cutting assembly is arranged on the supporting guide rail in a guiding mode and used for moving in a limited cutting stroke along the supporting guide rail, and the movable cutting assembly comprises cutting pieces of at least two sizes so as to meet different cutting requirements; the discharging assembly is arranged on the movable cutting assembly and used for sweeping sundries located on the supporting guide rail along with movement of the movable cutting assembly. The supporting guide rail is further provided with a discharging channel, and the discharging channel is used for discharging sundries cleaned by the discharging assembly. Through integration of the supporting guide rail, the movable cutting assembly of the multi-size cutting piece, the follow-up discharging assembly and the discharging channel, the problems that in the prior art, long-distance sampling is tedious, secondary cutting operation is complex, sundries are prone to clamping stagnation, and installation and movement are inconvenient are solved.

The present invention provides a method and apparatus for processing point cloud spatial data. The method for processing point cloud spatial data includes: receiving point cloud spatial data; constructing an octree model according to the total amount of the point cloud spatial data, a preset node resolution, and a preset node segmentation threshold; and performing distance sampling on the point cloud spatial data according to the octree model. The method and apparatus for processing point cloud spatial data provided by the present invention are based on the octree structure as the spatial data organizational structure. During the construction of the octree, a sampling process with the node resolution as the spacing is added. Finally, the constructed spatial data organizational structure retains a low-resolution point cloud data subset for low-level nodes and a high-resolution point cloud subset for high-level nodes.

The invention discloses a surveying and mapping device and a surveying and mapping method, and the surveying and mapping device comprises a roller distance measuring device which comprises a roller and a handle and is used for obtaining the data of the distance that the roller travels; the RTK device comprises a receiving module and a base station module, the receiving module is fixedly installed on the handle, and the RTK device is used for obtaining three-dimensional coordinates of sampling position points on the roller advancing route at a first sampling frequency; and the data processing module is connected with the roller distance measuring device and the receiving module, and is used for sampling the distance data at a second sampling frequency so as to obtain distance sampling data between two adjacent sampling position points, and drawing the motion trail of the roller based on the three-dimensional coordinates of the sampling position points and the distance sampling data. According to the invention, the roller distance measuring device and the RTK device are combined to obtain the motion trail, and the measurement precision of the motion trail is improved to a millimeter level.

The present invention discloses a point cloud semantic segmentation method and system based on adaptive feature sampling, including: performing color channel conversion on point cloud data; performing farthest point sampling on the converted point cloud data and conducting neighborhood search; performing adaptive feature transformation on each neighborhood; performing non-local feature transformation on the features of all points; fusing the features after the above-mentioned downsampling, adaptive features and non-local features; repeating downsampling three times; performing three-layer upsampling based on neighboring point interpolation; using the softmax function to predict the result and conducting model training based on the cross-entropy loss function. Using the trained model to predict the point cloud data to be segmented. The method of the present invention re-weights the neighborhood range of the downsampled points, adaptively adjusts the sampled points, and at the same time constructs the correlation between long-distance sampled points through the fusion of local and non-local features, effectively reducing the influence of noise points on the segmentation result and improving the segmentation effect of data-imbalanced objects.

The present invention discloses a navigation radar multiple reflection false target identification technology based on arithmetic sequence extraction. The technology utilizes the rule that false echoes formed by multiple reflections appear at equal intervals along the same azimuth and their energy decays successively. First, a group of equally spaced extreme points is extracted for each pulse echo. Extreme points distributed according to the arithmetic sequence rule are extracted using a distance sampling rate δr as a stepping interval. Then, "distance-azimuth-equally spaced value" clustering and "group number-class number" clustering are performed in sequence. Finally, multiple reflection false echoes are identified and marked, thereby eliminating or reducing their impact on the navigation radar's detection of real surface targets. The present invention can effectively identify and mark false targets formed by multiple reflection echoes in navigation radar echoes, avoiding false alarms and missed detections caused by false echoes on surface target observations. Furthermore, ship scale information is extracted from the echoes, facilitating subsequent surface target identification and tracking.

The invention discloses a marine radar image simulation method under the condition of ship shielding, and the method comprises the steps: modeling a to-be-detected ship target into a plurality of scatterers, determining the effective scatterers, and obtaining the reflection power of the effective scatterers received by a radar; the voltage signal of the distance sampling unit on the sight line of the main shaft of the radar antenna is obtained, and the voltage signal is displayed through a radar echo image after being subjected to coordinate conversion. The method is a radar signal simulation method under the condition of considering physical shielding, shielding object characteristics of a navigation water area and target ship characteristics are fully combined, a ship target is modeled into a plurality of independent scatterers, and the shielding conditions of the scatterers are analyzed one by one, so that the simulation accuracy of the radar signal is improved. And the radar echo image of the target is synthesized by using the effective reflectors which are not shielded, so that the stability of target tracking is high, the probability of tracking failure is greatly reduced, and the ship driving safety is improved.

The application discloses a kind of multi-source information coordination's electric power equipment three-dimensional temperature field construction method.Firstly, precise external parameter relationship is obtained by joint calibration using infrared camera, IMU and laser radar, eliminate point cloud distortion and extract corner point and plane point, construct re-projection residual, distance sampling residual and IMU pre-integral residual, adopt error state iterative Kalman filter to optimize global pose, obtain positioning.Self-supervised depth completion network is proposed, combine infrared temperature image and sparse depth map generated by laser radar as input, adopt infrared image guided depth completion strategy, and use pose information to estimate relative motion of adjacent frame, introduce feature alignment module to reduce alignment error, combine depth map, infrared image and IMU data, realize the efficient construction of electric power equipment three-dimensional temperature field.The application more accurately constructs electric power equipment three-dimensional temperature field, improves the ability of substation inspection robot to electric power equipment state monitoring and fault diagnosis.

The invention discloses a high-performance point cloud distance sampling simulation method, and belongs to the technical field of point cloud simulation. A linear depth test method is introduced, a graphic processing flow of a game engine is modified, high-performance detection of the distance of an object in a visual range of a camera is achieved, and the distance of the object in the visual range of the camera is detected through the linear depth test. According to the method, the depth testing process of the GPU during rendering is fully utilized, the requirement for detailed collision detection on each ray is reduced, the depth map is sampled in the fragment coloring function, efficient calculation of the surface depth of the object is achieved, and compared with a traditional method, the method has the advantages that the calculation overhead is remarkably reduced, and the calculation efficiency is improved. Compared with the prior art, the point cloud distance sampling method has the advantages that the point cloud distance sampling method is simple, the performance is higher in a large-scale scene, and the simulation efficiency is greatly improved, so that high-performance and real-time point cloud distance sampling can be realized, and more reliable and efficient data support is provided for simulation of an automatic driving system.

This invention provides a fully automated pretreatment method, system, and storage medium for soil samples. The method involves acquiring a three-dimensional environmental model of the soil sample pretreatment process, determining a task sequence consisting of a starting point, task points, and target points based on the process flow, and planning a segmented, collision-free path using adjacent point pairs as start and end points. A Gaussian mixture probability model is constructed based on the endpoints, generating sampling points in the probability space to guide random tree growth. The node closest to the sampling point is selected as an expansion node, and multiple alternative directions are generated around the expansion direction. The direction with the lowest energy consumption is selected using a joint energy consumption model. New nodes are generated with a preset step size, and these nodes are added only when the weighted sum of the angular acceleration and angular jerk of the path segment is below a disturbance threshold. When a new node enters the endpoint's attraction domain, an initial path is generated and locally optimized to obtain a segmented path. All segmented paths are then concatenated into a global path, and a robotic arm is controlled to perform the soil pretreatment operation.

The invention discloses a privacy protection system for protecting user local dx-private, the system comprises a bounded disturbance generation mechanism of a user side and a TK-means clustering method of a server side, belongs to the field of data mining, and relates to a method for generating local synthetic data and a clustering method, and the system comprises the following steps: carrying out noise distance sampling through a bounded exponential function; then randomly generating a batch of data and generating synthetic data through a gradient descent method; and then a traditional K-Means algorithm is improved by using an EM algorithm based on the T hybrid model. According to the method provided by the invention, the data utility and the accuracy of the clustering result are improved while the dx-private privacy protection is provided.

The invention discloses a ScanSAR full-scene rapid splicing imaging method. The method comprises the following steps: acquiring BURST data; based on the BURST data, obtaining the normalized sampling rate of the azimuth direction and the range direction of the whole scene, obtaining the zero padding number of the range direction and the zero padding number of the azimuth direction, and updating the scene center slant distance of the BURST data according to the zero padding number of the azimuth direction; based on the azimuth starting time and the distance starting time, the azimuth starting time and the distance starting time of the spliced image are determined, and the azimuth time offset and the distance time offset are obtained; according to the zero padding number and the time offset corresponding to the distance direction and the azimuth direction, performing imaging processing on the BURST data by adopting an SEPCAN algorithm to obtain focusing imaging data; and carrying out full-scene splicing according to a time sequence on the basis of the focused imaging data and the azimuth sampling interval and the distance sampling interval of the BURST data to obtain a spliced image. The method can lay a foundation for application such as observation SAR image target detection and interpretation.

The utility model discloses a long-distance safety analysis sampling device, relates to a sampling device, and in particular relates to a long-distance safety analysis sampling device which adopts a telescopic structure and can realize long-distance sampling and improve the safety of sampling personnel. The long-distance safety analysis sampling device is characterized in that the sampling device comprises a telescopic assembly, a sampling probe, a stepless motor and a take-up assembly, the telescopic assembly and the take-up assembly are both connected with a driving gear on a rotating shaft of the stepless motor, the sampling probe is installed at the end of the telescopic assembly, and the sampling probe is installed at the end of the take-up assembly. The sampling probe is connected with an analysis instrument through a data line, and the data line is wound on the take-up assembly. The long-distance safety analysis sampling device is simple in structure, scientific in design and convenient to use, the motor is adopted to drive the telescopic assembly and the take-up assembly to work synchronously, the threaded sleeve stretches out and draws back through meshed teeth between the telescopic assemblies so as to achieve the purpose of automatic stretching and retracting, sampling work can be conveniently carried out in a limited space, and the sampling efficiency is improved. And the sampling safety is improved.

This invention discloses a modular XL-MIMO near-field beam training method, belonging to the field of XL-MIMO beam training technology. The method comprises: activating the antennas of the first and last modules in the array to perform far-field beam scanning, determining the codeword index with the maximum beam gain of the first and last modules, estimating the codeword index with the maximum beam gain of the entire array, and using this as the center, selecting several candidate angles, using a polar codebook for non-uniform distance sampling to generate candidate sampling points, assigning each candidate sampling point to a collaborative module for training, determining the energy of each candidate sampling point, and then determining the optimal polar near-field codeword index to complete the modular XL-MIMO near-field beam training. This method not only effectively reduces training overhead, but also maximizes the accuracy of beam gain and angle estimation under resource-limited conditions.

The application relates to a deviation summation two-parameter fast estimation constant false alarm rate detection method, which realizes the following steps: 1) setting parameters such as a protection window, a reference window and a reference sub-window; 2) receiving logradar pulse echo data; 3) adopting a sliding window mode to sequentially count the background amplitude mean value of each distance sampling unit along the distance direction; 4) calculating the deviation square value of each echo amplitude and the background amplitude mean value; 5) adopting the sliding window mode to sequentially count the background amplitude standard deviation of each distance sampling unit along the distance direction; and 6) calculating a detection threshold according to the background amplitude mean value and the standard deviation, and sequentially performing threshold detection on each distance sampling unit. The application estimates two parameters of the background amplitude mean value and the standard deviation, so that the detector can adaptively cope with a non-uniform clutter environment, a fast parameter estimation means of approximate sorting is adopted, the multiple target environment can be effectively coped with, and the processing speed can be greatly improved.

The embodiment of the application provides a kind of based on infrared detection's outer wall hollowing out and falling risk prediction method and system, the method comprises: based on preset time interval, the wall material of the building outer wall to be detected at different time points is acquired;And for the wall material of each time point in different time points, from the mapping relationship library determined in advance, obtain the multiple sampling distances corresponding to wall material;Based on multiple sampling distances, the building outer wall to be detected is physically sampled, and multiple distance sampling images are generated;Based on multiple distance sampling images, image fusion is carried out, and the wall temperature field image of the building outer wall to be detected at different time points is generated;And based on the wall temperature field image of different time points, determine temperature field image sequence;Based on temperature field image sequence, outer wall risk detection and risk level division are carried out, and the current risk level of the building outer wall to be detected is determined.The above scheme can improve the accuracy of outer wall hollowing out and falling risk prediction.

The application discloses a path trajectory correction method, device, equipment and medium. In the method, the coordinate set of the initial path point of the to-be-corrected path is acquired, B-spline fitting is performed to obtain the position of each control point of the to-be-corrected path, for each control point, the angle difference between the line direction of the control point and the last control point and the line direction of the last control point and the control point before the last control point is determined according to the position of the control point and the positions of the adjacent two control points before the control point, and / or the distance of the nearest obstacle corresponding to the control point is determined according to the pre-stored ESDF map, the updated position of the control point is determined through angle optimization and / or obstacle distance optimization, and the position of each path point after trajectory modification is obtained through B-spline distance sampling, so that the problems of path close to obstacles, danger and / or large angle deflection leading to position deviation are solved, and the operation efficiency of the forklift AGV is improved.

The invention discloses an automatic adjusting and arranging device for a dense archive bookshelf. The automatic adjusting and arranging device comprises a two-phase four-wire stepping motor, a relay, a laser distance measuring sensor, a power supply and a linear module guide rail, the laser distance measuring sensor is used for collecting distance sampling data of files, opposite to the laser distance measuring sensor, on the bookshelf; the relay is used for controlling the two-phase four-wire stepping motor to move on the linear module guide rail according to the distance sampling data; and the two-phase four-wire stepping motor is used for pushing the files on the bookshelf to fix the files under the control of the relay, and arranging the books to be fallen down into the original shape, so that the problem that the files fall down due to taking or are required to be put into the bookshelf without space is solved.