34 results about "Trajectory segmentation" patented technology

The invention discloses a preprocessing method of a mobile phone signaling trajectory based on clustering outlier analysis, which reasonably and effectively cleans and denoises base station location data to remove abnormal data, so as to acquire relatively accurate trajectory data on the premise of ensuring original motion characteristics of the data. The method comprises the following steps: on the basis of removing a time repeated sampling value and an invalid sampling value of the trajectory data, performing sub-trajectory segmentation by recognizing the abnormal sampling interval; using an outlier detection algorithm to detect outliers, and dealing with larger system noises; and using Kalman filtering to deal with inherent noises, so that the trajectory data which can perform subsequent mining processing can be obtained. Compared with the prior art, the preprocessing method of the mobile phone signaling trajectory provided by the invention has the following beneficial effects: reasonable and effective multi-level pretreatment, cleaning, denoising and anomalies removal are performed on the base station location data, so that relatively accurate trajectory data can be obtained on the premise of ensuring the original motion characteristics of the data.

The invention discloses a road network hot spot region mining method, belongs to the technical field of data mining and solves the problem when trajectory spatial-temporal similarity measurement and clustering calculation is adopted for trajectory clustering in the prior art. The method includes: step 1, performing trajectory segmentation on all trajectory sections, and calculating spatial-temporal similarity and spatial-temporal distance between two sub trajectory sections acquired after segmentation; step 2, performing clustering calculation on all trajectory section data in a grid space according to the spatial-temporal similarity and the spatial-temporal distance of the sub trajectories and a DBSCAN algorithm based on dynamic neighbor; step 3, selecting a remarkable class cluster set from class clusters of clustering calculation, and extracting stay spots from the remarkable class cluster set; step 4, according to the number of trajectory sections carried by the stay spots, acquiring hot regions of the stay spots, and acquiring hot spot regions in a road network from the regions where the stay spots are positioned. The method is used for spatial position locating.

The invention discloses an automatic parallel parking path planning method based on two sections of second-order Bezier curves. The method comprises the steps of obtaining target parking space information to determine a parking target position and establishing a global coordinate system by taking the parking target position as an original point; constructing a parking trajectory segmentation judgment model to determine the number of planned trajectory segments, planning the end position range of the first section of trajectory, determining the start point, the end point and the control point of each section of planned trajectory, performing path planning by using a second-order Bezier curve, performing monitoring at set intervals, and adjusting the planned trajectory by using a dynamic adjustment planning trajectory method. According to the method, the requirement for the initial pose of the to-be-parked vehicle is lowered, and the environmental adaptability of the path planning methodis improved; the vehicle can be driven into the planned area manually through simple adjustment, so that the cost is reduced, and the method can be applied to medium and low-grade vehicles; control points of parking trajectory planning are easy to determine, a trajectory generation algorithm is simple, and the path curvature is continuous. The correction difficulty of the track error is reduced,and the obstacle avoidance performance and parking efficiency of the vehicle are improved.

To provide a moving object detection apparatus which accurately performs region extraction, regardless of the pose or size of a moving object. The moving object detection apparatus includes: an image receiving unit receiving the video sequence; a motion analysis unit calculating movement trajectories based on motions of the image; a segmentation unit performing segmentation so as to divide the movement trajectories into subsets, and setting a part of the movement trajectories as common points shared by the subsets; a distance calculation unit calculating a distance representing a similarity between a pair of movement trajectories, for each of the subsets; a geodesic distance calculation unit transforming the calculated distance into a geodesic distance; an approximate geodesic distance calculation unit calculating an approximate geodesic distance bridging over the subsets, by integrating geodesic distances including the common points; and a region extraction unit performing clustering on the calculated approximate geodesic distance.

The embodiment of the invention discloses a target tracking method and device, electronic equipment and a computer storage medium. The method includes: establishing target object trajectories of eachtarget object in a video according to first detection boxes in a first image frame of the video and second detection boxes in a second image frame, wherein the second image frame is a frame of image before the first image frame in the video; respectively extracting features of a first detection box corresponding to each target object; carrying out trajectory segmentation on each first detection box according to features of the first detection frame corresponding to the each target object and the respective target object trajectories of each target object in the first detection boxes and the second detection boxes to obtain first detection box information after segmentation; and carrying out target tracking according to the first detection frame information after segmentation. The above-mentioned embodiment of the invention realizes possibility of eliminating detection tracking errors while noises are filtered out.

The invention discloses a vehicle trajectory data processing method, device and equipment and a storage medium. The method comprises the steps: sorting the obtained trajectory points of each vehicle according to the sequence of timestamps from small to large, and obtaining a trajectory point sequence of each vehicle, wherein the track points comprise position information, timestamps and license plate information; aiming at the track point sequence of each vehicle; performing trajectory segmentation on the trajectory point sequence according to the target area to obtain a plurality of trajectory segments, and applying the logically segmented trajectory segments as source data to a preset algorithm, such as an artificial intelligence algorithm or a business statistics algorithm, wherein thedata volume needing to be processed by the algorithm can be reduced, and the operation efficiency of the algorithm can be improved. The scheme can be applied to the traffic field of smart cities, thereby promoting the construction of the smart cities.

The invention discloses a multimodal surgical trajectory fast segmentation method based on unsupervised deep learning, belonging to the field of robot assisted minimally invasive surgery. Firstly, therobot system collects kinematics data and video data for a robot-assisted minimally invasive surgery process. Then wavelet transform is used to smooth and filter the short trajectory noise and data jitter in kinematics data. At the same time, a DCED-Net network structure is used to extract features from video data. The kinematics data after smoothing filtering and the video data image after feature extraction are sent to the improved TSC model for clustering, and the pre-segmentation results of n surgical demonstration trajectories are obtained. Finally, the PMDD merging algorithm is used tosimilarly merge the pre-segmentation results of each trajectory, and the merging result is the final trajectory segmentation result. The invention provides an optimization scheme and unsupervised deeplearning for over-segmentation and other problems, accelerates the extraction speed of video features, improves the quality of features and makes the clustering result more accurate.

The invention relates to a ship trajectory prediction method based on a one-dimensional convolutional neural network and LSTM. The method comprises the following steps: a data preprocessing step: preprocessing trajectory data including ship position, navigational speed and course information collected by a ship AIS to obtain trajectory segmentation data; a feature extraction step: adopting a one-dimensional convolutional neural network to perform feature extraction optimization on the trajectory segmentation data, and combining the extracted advanced features with the trajectory segmentation data to construct input data of trajectory prediction training; a trajectory prediction model training step: importing the input data into an LSTM neural network model to learn a ship motion law implied in trajectory data; and a trajectory prediction step: predicting the position of the ship at the next moment by using the ship motion law. Compared with other existing prediction methods, the methodhas the advantages of better prediction precision, lower mean square error and quicker prediction.

The invention discloses a privacy protection method for track data release based on track segmentation. The privacy protection method comprises the following steps: applying an equivalence class division algorithm based on track segmentation filling to an original track data set accumulated by a location-based service application provider; applying a clustering group construction algorithm based on trajectory segmentation clustering to each equivalence class; determining the starting time of a clustering group and performing dividing to obtain candidate clustering groups; traversing each candidate clustering group, determining a trajectory set, and constructing trajectory segments outside a clustering group time interval; inserting the tracks which are not added into the clustering group into the clustering group with the same time interval and the nearest spatial position; and performing spatial disturbance on each position point on each track in each clustering group, and changing each clustering group into an anonymous track set as a track data set which can be directly published. According to the method, an equivalence class division algorithm is adopted, the number of deletedspace-time points is reduced, it is guaranteed that the equivalence class contains enough tracks, and the availability of the to-be-published data is improved.

The invention discloses a trajectory segmentation method and a device. The trajectory segmentation method comprises an acquisition step used for acquiring multiple trajectory sampling points, a segmentation step used for segmenting the trajectory to obtain at least two candidate segmentation points, a feature extraction step used for extracting a set of features describing the relative relationship between the at least two candidate segmentation points, and a classification step used for classifying the at least two candidate segmentation points as true or false segmentation points using a classifier based on the extracted features.

The invention relates to a fast automatic density clustering-based detection method for a variable-scale infrared small target and belongs to the image analysis and image understanding field. According to the method, an SURF (speeded-up robust feature) operator is utilized to realize feature extraction of the variable-scale infrared small target; the problem that features are sensitive to the environment is solved; according to a problem domain, a fast automatic density clustering algorithm directly completes trajectory segmentation and extraction from a space domain, and therefore, the problem of exponential complexity caused by data fusion in original sequence detection can be solved; the problems of over-segmentation and under-segmentation in a clustering process can be solved, and the integrity and independence of trajectory extraction and the automatic selection of a clustering center are ensured; In later-stage trajectory extraction, a backtracking algorithm is used to find an optimal solution, the smooth invariant constraint of trajectories is fused into the design of a pruning function, so that unrelated clutter branches can be cut off quickly, and therefore, the speed of solution search can be increased; and the robust feature detection operator and the backtracking strategy are used in combination, and therefore, the detection problem of the variable-scale infrared small target can be solved, and the real-time performance and robustness of the algorithm can be improved.

The invention discloses a mechanical arm task planning system based on a behavior tree and an application method. The mechanical arm task planning system comprises a track segmentation module, a target pose detection module, a behavior tree design module, a dynamic motion primitive generalization module, an upper computer system and a Kinova Jaco mechanical arm. On the one hand, a Kinova Jaco mechanical arm is controlled by teleoperation to perform multiple grabbing task demonstration, an upper computer system obtains action elements from collected demonstration data through a track segmentation module, and an action element library is constructed; on the other hand, according to the grabbing task execution logic, a grabbing task behavior tree is created through a behavior tree design module; the Kinect sensor captures an object depth image, after the object position and posture are estimated through the target posture detection module, generalization is carried out through the dynamic motion primitive generalization module in combination with corresponding motion primitives in the motion primitive library, the motion primitives obtained through generalization serve as action nodes of a behavior tree, the behavior tree transmits an execution instruction to the upper computer system, and the action primitives serve as action nodes of the dynamic motion primitive generalization module. And the Kinova Jaco mechanical arm is controlled to execute the grabbing task.

The invention provides a trajectory semantic segmentation method based on dynamic pedestrian agent hybrid model. The main content comprises: a dynamic pedestrian agent hybrid model (MDA), a hidden Markov model (HMM) and trajectory semantic segmentation. The process is as follows: the MDA is a layered Bayesian model, and the model predicts the dynamics and reliability parameter of each agent through a modified Kalman filter (used for processing an observed value of trajectory missing trace) and an iterative expectation maximum algorithm (EM algorithm); then the trajectories are gathered according to the estimated agent; then, the trajectory is transmitted to the HMM, and the parameter phi is estimated; and finally, semantic segmentation is performed on the trajectory in the HMM by using aviterbi algorithm and the learned HMM parameter phi. According to the trajectory semantic segmentation method provided by the invention, the behavior of a pedestrian can be analyzed during the trajectory segmentation of the pedestrian, and corresponding estimation is performed, so that the accuracy and the efficiency are higher.

The invention discloses a track compression method. Aiming at the defects in the prior art, the solution thought of the method is to globally optimize a feature trajectory sub-chain obtained by simplifying and compressing a line segment based on a global optimization thought, and comprehensively consider various influence factors such as a deviation angle, a vertical Euclidean distance, a trajectory segmentation number and the like; evaluation indexes capable of comprehensively reflecting space and time characteristics of a compression result are established, so that the contradiction between the compression efficiency and effective information retention and the contradiction between the compression efficiency and the original track as similar as possible are well balanced. The invention further discloses a track compression device. Compared with the prior art, the method and the device have the advantages that off-line compression that all tracks contain a large amount of effective information and are high in accuracy can be achieved, and the accuracy and efficiency of overall operation of track compression are improved.

The invention discloses a vibration suppression method of a flexible mechanical arm based on trajectory segmentation optimization. The approximate delay amount of the conventional vibration suppression method is derived based on the characteristics of the conventional vibration suppression method; the approximate delay compensation amount of the original track is deduced, the original track and the non-delay track are optimized in a segmented mode and then combined through cubic curve transition optimization, and finally the optimized track which is good in vibration suppression effect, free of delay and good in track middle section precision is achieved. According to the method, the vibration suppression function of the flexible mechanical arm is achieved, the characteristics of zero delay and good track middle section precision keeping are achieved, and parameters in the method can be modified and adjusted according to actual working conditions and requirements so as to meet requirements.

The invention provides a target motion track segmentation compression method based on improved particle swarm optimization, and the method comprises the steps: constructing a particle swarm track segmentation approximation scheme optimization basic frame based on contour preservation, building a mapping relation between track data and particle states, defining a track contour fitting degree optimization target function, and carrying out the optimization of the target motion track. Individual trajectory segmentation scheme search based on the particle swarm optimization is realized; a particle updating strategy based on neighborhood adjustment and random jump is designed, and the track segmentation scheme searching efficiency of the particle swarm algorithm is improved; and for each trajectory in the data set, particle swarm trajectory segmentation scheme optimization with the segmentation number increasing from small to large is carried out, a trajectory segmentation compression effect evaluation index is established, and a trajectory data overall segmentation compression scheme which comprehensively considers the compression ratio and the segmentation precision is obtained. According to the method, global-oriented group intelligent track segmentation compression is realized, and compared with an existing method, the track data compression effect and segmentation approximation precision are both improved.

The invention relates to the technical field of trajectory analysis, in particular to a marine target trajectory segmentation and description method, electronic equipment and a storage medium. The marine target trajectory segmentation and description method comprises the following steps: dividing a given trajectory into a plurality of sub-trajectories according to a trajectory segmentation rule; determining trajectory types of the sub-trajectories, and combining the sub-trajectories of the same type to optimize the trajectory to be composed of trajectory segments; phagocytosis is carried out on each track segment according to phagocytosis rules, so that the track is optimized to be composed of phagocytosis track segments; each phagocytic track section is infringed according to an infringedrule, so that the track is optimized to be composed of infringed track sections; and evolving each infringement track section according to an evolutionary rule to obtain a track composed of evolutionary track sections. According to the method, the track is segmented fully according to the shape characteristics of the track, so that the influence of environmental factors is eliminated; according to the method, seven behavior modes are adopted to segment and describe the track, and segmentation and description are accurate and reliable; in addition, the method is high in applicability.

The invention discloses a track clustering privacy protection method based on semantics. Vehicle track data mining and privacy protection of interest points in a track are combined. The method comprises the following steps: firstly, carrying out trajectory segmentation by using an improved Clusters function so as to generate a cluster on a single trajectory; then calculating a distance range Eps of trajectory data clustering; determining an Eps neighborhood of the track point according to the Eps, then determining a core point of the cluster, and performing stop area clustering by using the core point and the Eps neighborhood of the core point; adjusting the Eps to enable the Eps to meet a termination condition; performing secondary clustering on the stop area through core attribute selection; before track data is published, carrying out road network segmentation according to a privacy protection level and a Delaunay triangulation method to form a plurality of Voronoi maps (VM); secondly, based on the anonymity degree and the category degree, classifying the interest points in each VM into a plurality of buckets (buckets); and finally, publishing track data containing the pseudo positions of the interest points. In the invention, the pseudo positions of the interest points refer to the interest points closest to the duration time of each interest point in the bucket corresponding to the interest point. According to the method, the privacy information involved in the vehicle track interest point mining process can be effectively protected.

The invention relates to a travel trajectory segmentation method and device, a storage medium and electronic equipment, and the method comprises the steps: determining the number of sectors corresponding to a data segment according to a driving angle corresponding to each data point in the data segment in a travel track of a transport person; according to the time point and the position coordinate corresponding to each data point, determining the static duration and the average driving speed of the transport personnel in the data segment; according to the number of sectors, the static duration and average driving speed, determining whether a staying segment mark is added to each data point or not, so that the data point marking process of the data segment is completed; for each data segment in the travel trajectory, performing a data point marking process repeatedly to determine a stay segment in the travel trajectory. According to the invention, the stay segment can be identified according to the driving data of each data point in the travel trajectory, dependence of the stay segment identification process on manually submitted parameters is reduced, the accuracy of stay segment identification is improved, and the efficiency of travel trajectory segmentation is improved.

The invention discloses a method for rapidly segmenting multimodal surgical trajectories based on unsupervised deep learning, which belongs to the field of robot-assisted minimally invasive surgery. The method first aims at a robot-assisted minimally invasive surgery process. The robot system collects surgical kinematics data and video data; structure for feature extraction on video data. The kinematics data processed by smoothing filter and video data image after feature extraction are sent to the improved TSC model for clustering, and the trajectory pre-segmentation results of n surgical demonstrations are obtained; finally, the PMDD merging algorithm is used to pre-segment each trajectory. The segmentation results are similarly merged, and the merged result is the final trajectory segmentation result. The present invention proposes an optimization scheme and unsupervised deep learning for problems such as over-segmentation, accelerates the extraction speed of video features, improves feature quality, and makes clustering results more accurate.

The invention discloses a segmented clustering-based road network trajectory semantic privacy protection method, and the method comprises the steps: 1) preprocessing an initial trajectory data set, and carrying out the re-sampling operation of a trajectory; 2) based on an MDL (Minimum Description Length) principle, performing segmentation division on the trajectory to obtain a trajectory segmentation data set; 3) clustering the track segments in the track segment data set based on the road network distance to form segment classes; 4) determining the road network field of the segment class, and counting the semantic position distribution of the segment class; and 5) based on the semantic position distribution of the segment class, generating an anonymous segment set in a heuristic manner, and using the anonymous segment set to replace a single track to carry out data release so as to realize road network track semantic privacy protection. An anonymous segment set is used for replacing a single track for data release, the semantic privacy of the track is protected, track segmentation, clustering and heuristic algorithms are combined, the anonymous success rate and the semantic privacy protection effect are improved, and the anonymous data quality is improved.