424 results about "Object tracking algorithm" patented technology

The invention discloses a tracking-learning-detection (TLD)-based video object tracking method, which comprises the following steps of: (1) when a start frame of an object is tracked, generating an image sub-window according to given position and size information, and training a detector; (2) estimating the area of the object in a current frame according to the position and size information of the object in a previous frame by using a tracker; (3) detecting the current frame by using the detector, and finding all possible objects in the current frame; and (4) performing fusion processing on results of the tracker and the detector, judging whether the object exists in the current frame or not, returning to the step (3) to process the next frame if the object does not exist in the current frame, judging whether a tracking trajectory of the current frame is effective or not if the object exists in the current frame, giving the position of the object, returning to the step (2) to process the next frame if the tracking trajectory of the current frame is ineffective, finishing the online updating of the detector if the tracking trajectory of the current frame is effective, and returning to the step (2) to process the next frame. According to the method, the conventional TLD algorithm is improved to obtain a video object tracking algorithm which is more ideal than the TLD algorithm.

The invention discloses an interactive remote expert cooperation maintenance system and a method based on an augmented reality technology. An on-site client side carries out video acquisition or image shooting on equipment to be detected and maintained, and transmits to a remote server side; the remote server side receives a video or an image sent by the on-site client side, receives drawing processing on the video or the image carried out by an expert, and sends the drawn target area image to the on-site client side; after receiving the drawn target area image, the on-site client side completes recognition, location and tracking of a target area in the target area image and an on-site video image through an object tracking algorithm; a detector finds out the target area image; a tracker tracks a target; the detection accuracy of the detector is improved continually through P-N Learning; a marked pattern or a text comment, and current equipment state information are overlapped at the position of the target area in the on-site video image, and an augmented reality image video is generated; the augmented reality technology is utilized to synthesize the augmented reality video, and the fault diagnosis and the maintenance of on-site equipment are completed with the cooperation of the expert.

The invention discloses a heavy-duty lorry driving barrier detection and tracking method based on a binocular fisheye camera, and belongs to the technical field of traffic vehicle active safety. The method mainly comprises the steps: starting two calibrated and corrected infrared fisheye cameras, disposed at two ends of the tail of a vehicle, to synchronously collect back-up environment image information when the back-up of a heavy-duty lorry is detected, outputting a back-up information after jointing, and displaying the back-up image on a display screen of a cab. A binocular fisheye visual system is used for obtaining the initial position and distance of a to-be-tracked barrier object, and then a video target tracking algorithm combined with the low-rank matrix theory based on particle filtering is used for achieving the tracking of a barrier, achieves the real-time updating of an image template in a tracking process, and detects whether there is a new barrier or not. When a vehicle is detected to turn around, a supersonic radar is started to detect that there is a barrier in a one-meter range beside the lorry, the position of the barrier is displayed and the early warning is sent. The method is mainly used for a large-size lorry.

The invention relates to an on-duty person violation behavior detection method and system, belongs to the technical field of intelligent video analysis, and solves the problems of low efficiency, highcost and low recognition accuracy of an existing detection method. The method comprises the following steps: constructing a target detection network model, and training by using a data set; acquiringmultiple paths of videos at different angles in the same scene in real time, carrying out multi-target detection and tracking by utilizing a trained target detection network model and a target tracking algorithm, acquiring personnel information in each path of videos, carrying out integration processing, and judging whether behaviors of personnel on duty are abnormal or not. According to the invention, an environment camera video is used as an input video source for intelligent video analysis; the system supports multi-channel video source input and fusion analysis, greatly improves the recognition accuracy of illegal behaviors through deep learning and data modeling means, and achieves the real-time and accurate monitoring of the duty behaviors of personnel on duty in scenes such as a monitoring center, a duty room, a command center and the like.

The invention discloses a vehicle-mounted video target detection method based on deep learning, and the method comprises the steps: employing an improved Faster R-CNN algorithm to realize target detection in a complex traffic environment, to provide a driving safety auxiliary function. An existing target tracking algorithm has a serious problem of small target missed detection. The depth information channel is added, the depth information channel is connected with an original color image channel in parallel, fusion is carried out in the channel dimension, candidate frame extraction and targetdetection are carried out on the fused feature image, the detection rate of a small target is improved, in addition, training of a difficult sample is added in training, and the overall target recognition rate of the algorithm is improved. According to the method, a problem of small target missed detection by using the Faster R-CNN algorithm is fully considered. The accuracy of vehicle recognitionin a complex traffic scene is improved through depth image feature fusion and a difficult sample mining method.

The invention discloses a method for generating video abstract on the basis of a deep learning technology. The method includes modeling backgrounds of video stream frame by frame and acquiring moving foregrounds to be used as candidate moving objects; tracking the candidate moving objects of each frame by the aid of a multi-object tracking algorithm and updating candidate objects which form movement tracks; training object classifiers by the aid of convolutional neural networks, confirming the candidate objects and determining categories of the objects by the aid of the classifiers after real moving objects are confirmed; fitting all the real moving objects and relevant information on a small quantity of images, forming video snapshots and displaying the video snapshots to users. The method has the advantages that the real objects and noise can be accurately differentiated from one another by the aid of the deep learning technology; the objects do not need to be confirmed frame by frame owing to an accurate multi-object tracking technology, accordingly, the computational complexity can be greatly reduced, an omission factor of the objects and a false alarm rate of the noise can be effectively reduced, the video processing speeds can be increased, and the method can be applied to various complicated scenes.

A method and system is configured to characterize regions of an environment by the likelihoods of transition of a target from each region to another. The likelihoods of transition between regions is preferably used in combination with conventional object-tracking algorithms to determine the likelihood that a newly-appearing object in a scene corresponds to a recently-disappeared target. The likelihoods of transition may be predefined based on the particular environment, or may be determined based on prior appearances and disappearances in the environment, or a combination of both. The likelihoods of transition may also vary as a function of the time of day, day of the week, and other factors that may affect the likelihoods of transitions between regions in the particular surveillance environment.

A method for automatically optimizing a parameter set for a tracking algorithm comprising receiving a series of image frames and processing the image frames using a tracking algorithm with an initialized parameter set. An updated parameter set is then created according to the processed image frames utilizing estimated tracking analytics. The parameters are validated using a performance metric that may be manually or automatically preformed using a GUI. The image frames are collected from a video camera with a fixed set-up at a fixed location. The image frames may include a training traffic video or a training video for tracking humans.

The invention provides a target tracking algorithm based on the self-adaptive particle filter and sparse representation. According to the target tracking algorithm based on the self-adaptive particle filter and the sparse representation, the improved self-adaptive particle filter technique is adopted to serve as a tracking algorithm framework, a block sparse representation model is used for establishing an observation similarity model of a target, partitioning of the target is achieved by means of the self-adaptive partitioning technique, a structural sparse column diagram of a current target state is constructed to calculate the observation similarity of the current target state, blocking is detected by means of a blocking detection mechanism, a target / background dictionary template and a target template column diagram are updated to capture the change of the appearance of the target and the change of the environment during tracking, L1 optimization in the sparse representation is achieved by means of the variable-direction multiplicator method, and then the execution speed of the target tracking algorithm is increased. The target tracking algorithm based on the self-adaptive particle filter and the sparse representation has the advantage that the robustness to the conditions of the posture change of the tracking target, the change of the environment and lighting and blocking is strong.

The present invention belongs to the visual tracking field, and provides a target tracking algorithm based on the scale adaptive correlation filtering and feature point matching which solves a long-time target tracking problem. The target tracking algorithm comprises establishing a scale adaptive correlation filtering tracking module CFF to process each frame of image; establishing a tracking module MTF based on the feature pint matching and an optical flow; and establishing a cooperative processing determination module of the CFF and the MTF. According to the present invention, a tracking problem is decomposed into the CFF and the MTF which can assist mutually, whether the algorithm is updated is decided by determining the shielded degree of a target or determining whether the target has disappeared in the view field, thereby preventing a model from being polluted by the background information to generate a drift phenomenon. When appearing in the view field again, the target can be detected again, and the corresponding modules are updated to track continuously and stably for a long time. Moreover, the processing speed of the target tracking algorithm satisfies the real-time processing requirement completely, and the target tracking algorithm has a very good effect aiming at an actual complicated scene.

The present invention discloses an intelligent monitoring method for the real-time trace of a moving target region, comprising a video input device, a visual analysis system and a network input-output device. In order to track the moving target in high real time and at high precision, the system adopts a target track algorithm which is based on the model dynamic switching, adopts a simple quick model, which is based on the region trace, according to the effective determination of a shielding state, for a single moving target that is not shielded, and adopts a narrow baseline image matching model, which is based on the SIFT feature, for compound motion targets which are mutually shielded. The system has the advantages of simple structure, high real time performance, high precision, and strong expandability, and the advantage that the system has cable Ethernet and wireless GPRS multiple network access facility and effectively realizes the real time trace function for the moving target region.

The invention belongs to the technical field of intelligent automobiles, specifically relates to a complex environment radar multi-target tracking and road driving environment prediction method, whichparticularly aims to solve the problems that the position relationship identification of a lane where a target vehicle locates is inaccurate and the robustness and precision of a target tracking algorithm are not high through determination based on a radar original target measurement value in the process that a vehicle having a self-adaptive control function drives in a curve or an intelligent vehicle having an autonomous valet parking function enters and exits from a curved ramp of an underground parking lot. The complex environment radar multi-target tracking and road driving environment prediction method is mainly implemented by present vehicle motion state estimation, millimeter wave radar signal conversion, time synchronization, target motion compensation, data rationality judgment,target measurement value noise reduction, road curvature estimation, target aggregation, target motion attribute and motion state identification, improved adaptive extended Kalman filtering algorithmtracking and data association, road driving environment prediction and key target generation.

A multitarget tracking antispoofing receiver utilizes multitarget tracking algorithms and multiple correlators for tracking signals of interest in a field of view about a nominal trajectory of a desired true signal for tracking targets within the code phase and carrier frequency signal space so as to predict when true and spoof signals will cross paths in the signal space without spoofing with loss of tracking of the desired signal so that true tracking of a desired target is maintained in the presence of a spoofing signal moving along a spoofing signal track and crossing a true path of the true signal.

The invention discloses a gray level target tracking algorithm based on marginal information and mean shift, which comprises the following steps: (1) preprocessing a gray level target image, including image de-noising, contrast enhancement and Sobel filtration; (2) extracting a feature template of the target in a preprocessed initial frame image; (3) forecasting an initial position of the target in the current frame by using a Kalman filter; (4) searching near the initial position forecasted by the Kalman filter in a mean shift tracking algorithm and obtaining an optimal position of the target in the current frame; and (5) updating the target template in combination with Canny filtration at regular intervals. Based on the marginal information and the mean shift algorithm, effective tracking of the gray level target is realized, and the tracking can be steadily finished in real time when the size, shape, gray level distribution and background of the target change.

The invention belongs to the technical field of video target tracking, and provides an online video target tracking method based on depth cross similarity matching. The method comprises the followingsteps of designing a depth feature cross similarity module, capturing all local similarity information of a template and a sample, wherein obtained similar features are no longer sensitive to displacement and deformation; designing a similarity attention layer, distributing weight coefficients for cross similarity results of different spatial positions, and enabling a tracking algorithm not to respond to edge background interference; and designing a loss function containing a parameter regularization item, and rapidly optimizing the parameter to an optimal value. Based on the above three-pointbasic scheme, the deep learning twin network based on matching is used as a basic framework, and any target in the video is tracked in a precise and robust manner. From the tracking effect, the method provided by the invention has the capabilities of distinguishing similar objects, re-identifying a reappearing target after shielding and coping with rotation and deformation, and can be applied tovideo applications such as automatic driving of a front scene, autonomous flight of an unmanned aerial vehicle, traffic or a safety monitoring video and the like.

A region within a field of view of a video stream associated with an Internet Protocol (IP) camera can be identified. The region can be associated with one or more logical boundaries which can correspond to one or more physical boundaries within a customer service area of a place of business. The customer service area can be a domain in which a customer interacts with a business product and / or a business service. A customer crossing a logical boundary of the customer service area can be detected in real-time. The detecting can be performed by a directional tripwire analytics which can comprise of an object tracking algorithm, a face detection functionality, and / or a shape detection procedure. The traffic flow associated with the region can be programmatically determined which can include flow density, flow rate, and flow speed.

The invention discloses an RA-Singer-EKF (Random Access-Singer-Extended Kalman Filter) maneuvering target tracking algorithm based on radial acceleration, which belongs to the field of radar maneuvering target tracking. According to the method, the radial acceleration and radial speed information of a maneuvering target can be rapidly and accurately provided, and the tracking performance of a radar on the maneuvering target is improved effectively. The method comprises the following steps of: (I) sampling a radar receiving signal, and obtaining a target radial acceleration and a radial speed by using a matching pursuit (OMP (Operation Management Platform)) method; (II) performing coordinate conversion on the radial acceleration and the radial speed at a data processing stage, and introducing into a measuring equation and a state equation; and (III) realizing maneuvering target tracking by adopting a Singer model and an EKF algorithm. Due to the adoption of the RA-Singer-EKF maneuvering target tracking algorithm, the maneuvering situation of the target can be reflected accurately in real time, the target tracking accuracy is increased, the speed and acceleration estimation accuracies are improved, engineering implementation is easy, and a high engineering application value and a good popularization prospect are achieved.

The invention is suitable for the field of safety monitoring, and provides a method for controlling the automatic zoom of a PTZ (pan / tilt / zoom) camera. The method comprises the following steps: establishing archival information comprising the hoisting height of a PTZ and the information of the PTZ camera; receiving an instruction, then opening the automatic zoom function of the PTZ camera, and acquiring the vertical rotation angle of the PTZ in real time; according to the current vertical rotation angle and archival information of the PTZ, calculating the focal distance of the PTZ camera; and according to the calculated focal distance of the PTZ camera, zooming the PTZ camera to the focal distance, or zooming the PTZ camera to an integer multiple of the focal distance closest to the calculated focal distance. The automatic zoom method provided by the invention has the advantages that because the zoom parameter of the PTZ camera can automatically be set according to the actual needs of a monitoring scene, the method can be applied more flexibly without carrying out zoom control by detecting the actual size of a target, thereby simplifying the complexity of a target tracking algorithm, facilitating implementing the target tracking more easily, and improving the working efficiency of the monitoring personnel.

The invention discloses a robot dynamic capture method and system. The method comprises the steps that S1, a collected image transmitted from an industrial camera is received in real time, and the obtained image is preprocessed; S2, a template tracking algorithm based on multiple clues is utilized to determine a foreground target region; S3, a template matching algorithm based on an edge gradientis adopted to precisely recognize a foreground target; S4, a random sampling consensus algorithm is adopted to perform modeling on motion information of an object in a view field; and S5, the information is fed back to a robot in real time for dynamic capture in combination with camera calibration and hand-eye calibration parameters. According to the robot dynamic capture method and system, by theadoption of a target tracking algorithm based on multiple clues, the motion target foreground region can be extracted efficiently and accurately; and under the application scene of robot sorting, therunning speed of a visual motion target tracking algorithm can be effectively increased, high-frame-rate real-time recognition is realized, and therefore the precision of visual servo is improved.

The invention discloses an overlapped domain dual-camera target tracking system and method. According to the system, an HSV-based space background weighted Meanshift algorithm is connected with an improved visual field boundary target. According to the target tracking algorithm, a relatively-independent tracking task is executed on a video sequence collected by each camera within the visual field range of the camera through a front-end single-camera tracking subsystem, relevant information of the tacked target is obtained, data communication between the cameras is achieved by utilizing a target connection algorithm, and then high-class video processing processes like behavior judgment are carried out.

The invention discloses a target tracking method which comprises the steps of: obtaining the initial position of a target and confirming the initial area at which the target locates according to the initial position; calculating a color histogram and an illumination robustness feature histogram of the initial area at which the target locates; utilizing the color histogram and the illumination robustness feature histogram for expressing a target model of the target; and searching the optimal position of the target from the current images according to the target model. In addition, the invention also discloses a target tracking device. The technical proposal disclosed by the invention can improve the stability of the target tracking algorithm to the illumination change.

The invention discloses a method and device of video image zooming processing. The method of the video image zooming processing includes the steps: adopting a video target tracing algorithm to confirm an optimal cuboid from video image cuboids composed of current image frames and a certain number of image frames in front or at the back of the current image frames based on important index values of the current image frames of videos to be processed and a certain number of the image frames in front or at the back of the current image frames, confirming an image area included in the optimal cuboid in the current image frames to be as an interest area of the current image frames, extracting an important target area of the interest area from the current image frames and acquiring a background image of the current image frames after the important target area is extracted, and acquiring current image frames through zooming processing by carrying out the zooming processing on the important target area and the background image respectively. According to the scheme, the video image zooming processing method and device is capable of improving a display effect of key content in the videos through the zooming processing.

The invention discloses a method and system for monitoring traffic accidents. The method for monitoring the traffic accidents comprises the steps that frame drawing processes are conducted on a traffic monitoring video obtained in real time to obtain a plurality of real-time frame image data; target inspection is conducted on the real-time frame image data through a deep learning framework to ensure a target area; target tracking is conducted on targets in the target area through a target tracking algorithm; and targeted traffic accidents are monitored according to the traffic accidents corresponding to different targets, and the states of the target traffic accidents are renewed according to monitoring results. According to the method and system for monitoring traffic accidents, based onthe deep study framework and the target tracking technology, automobiles and pedestrians can be effectively recognized and traced, the method and system for monitoring the traffic accidents can directly connected with a highway management system, highway video data can be automatically analyzed, and traffic accident information is output; a road camera can be effectively used, and the covering rate of the traffic accident collecting on the highway is increased; and special video vehicle inspection equipment does not need to be purchased, and the cost is saved.

The invention discloses a visual significance-based interference sensing and tracking algorithm. The visual significance-based interference sensing and tracking algorithm comprises the steps of S1, inputting a video image; S2, representing the appearance model of a to-be-tracked target in the image by utilizing the characteristics of a gradient orientation histogram, and calculating to obtain a histogram disturbance model in the gradient direction; S3, calculating the output response and the context awareness correlation tracking response; S4, carrying out weighted fusion to obtain a target weighted response, adopting a position where the maximum response is located as the position of the to-be-tracked target, and estimating the target scale and the position change; S5, when the to-be-tracked target is shielded, calculating to obtain a visual saliency map, and estimating the position of the to-be-tracked target according to the position of a candidate target; S6, according to the condition of the to-be-tracked target, updating the appearance model and the disturbance model; S7, inputting the next frame of the image, and returning to the step S1. According to the algorithm, the problem in the prior art that the existing target tracking method is easily influenced by factors such as illumination change, low resolution, scale change, shielding, similar targets, noisy backgrounds and the like so as to cause the poor tracking effect can be solved.

The invention provides a target tracing method which is based on a particle filter and active contour model, and relates to a target tracing method for a gradient vector flow-parameter active contour model. The method firstly adopts a background difference method to obtain target initial profile, and the improved gradient vector flow-parameter active contour model is utilized, thus leading the parameter active contour model to be converged to true profile of a moving target; control points are increased or reduced according to the distance of the control points, so as to achieve the purpose of tracing moving and morph targets in a self-adapting way; then energetic particle filter target tracing algorithm is used for tracing the targets by combining the particle filter and the improved gradient vector flow-parameter active contour model; and tracing strategy when the target is sheltered is used, thus overcoming the effect of sheltering in the tracing process. In a whole tracing process, profile point information is not sheltered mostly by fully utilizing GVF-Snake model target profile, so as to effectively overcoming the effect such as complex environment and the like in the tracing process.

The invention provides a target object behavior data acquisition method based on videos, comprising: first, playing a pending video; according to a scheduled target recognition algorithm and a target tracking algorithm, identifying and tracking a target object in the video, and according to the corresponding relation between a video pixel coordinate and a physic coordinate, calculating the physic coordinate of the target object; and finally, calculating the behavior data of the target object according to the physic coordinate, corresponding frames and video frame rates. Compared with the prior art, the method is based on videos, and utilizes a target recognition algorithm, a target tracking algorithm and the corresponding relation between a video pixel coordinate and a physic coordinate, thereby obtaining the physic coordinate of a target object in a video; and then, according to the frame rate and frame number of the video, the method calculates target object behavior data, thereby providing real, effective, objective and vast data support for research of pedestrian microscopic behaviors.

The invention discloses an openCV(open source computer vision library)-based video target tracking algorithm. The openCV-based video target tracking algorithm is characterized in that a selected template is matched with a video frame; a subgraph position which is most similar to a template image in the video frame is found out through calculation of correlative coefficients; updating of the template is determined according to a predicted position of a kalman filter and the correlative coefficient values; and the specific algorithm includes template matching, position prediction and template updating. Compared with the prior art, the openCV-based video target tracking algorithm has the advantages that recognition and tracking of a target are not affected by environment change; the target object is accurately recognized and tracked in time; and the updating of the tracked object is available, so that the template can be dynamically updated during the system tracking, and the tracking is more accurate under the condition of environment and object change. In addition, a plurality of parameters are used as tracking evidences, so that the tracking is more reliable; and under the condition of target object moving, continuous background change and shadow influence, the tracking object cannot be lost..

The invention discloses a Camshift (continuously adaptive mean-shift) robot fish tracking method based on an embedded Kalman filter. The Camshift robot fish tracking method is suitable for complicated underwater environments of robot fishes, the accuracy of the robot fish tracking quick motion is improved, and the real-time performance is good. The Camshift robot fish tracking method comprises the following steps of firstly, utilizing the Kalman filter to predict the possible position of a motion object in the next frame of image; then, using Camshift to search in the relative reduction range, so as to effectively enhance the accuracy of tracking a quick motion object; then, utilizing the observing value obtained by the Camshift to correct the observing value obtained by the Kalman filter, so as to further improve the accuracy of target tracking. Compared with the prior art, the Camshift robot fish tracking method has the advantages that the real-time performance and accuracy of the target tracking algorithm can be comprehensively considered, and the purpose of the robot fish accurately tracking the target in real time is realized.

The invention relates to a binocular-vision-based real-time extraction method and system for three-dimensional hand information. The method comprises: obtaining image information collected by a left camera and a right camera of a binocular camera system in real time; detecting hands in the images collected by the left camera and the right camera in real time; detecting a hand at a first frame by using a histogram of oriented gradient (HOG) and a support vector machine (SVM) linear classifier and detecting hands at follow-up frames by using a target tracking algorithm; extracting information of centers of palms and fingertips of the detected hands in real time and using the information as feature points of the hands; and according to the binocular-vision principle, calculating three-dimension coordinates of the feature points based on the extracted feature points so as to obtain real-time three-dimensional hand information. In addition, the system is composed of a binocular camera system and an information processing device; and the information processing device includes a hand detection module, a feature point extraction module, and a three-dimensional reconstruction module. According to the invention, real-time reliable detection and tracking of hands under a complex background can be realized; and thus reliable reconstruction of the three-dimensional hand information can be realized based on the detection and tracking.

The invention discloses a real-time visual target tracking method based on light streams. The method includes: screening feature points rich in textural features, and tracking the feature points between every two frames of images to obtain feature point matching relations; filtering the feature points with large matching errors through various filter algorithms such as normalized correlation coefficients, forward-reverse tracking errors and random consistency detection to retain most reliable feature points so as to obtain the target motion speed in images; using the target motion speed as observation quantity, and estimating through Kalman filtering to generate the target position. Compared with a tracking algorithm, the method has the advantages that the method is unaffected by light, the target can be stably tracked under the conditions that the target moves, a camera moves, the target is blocked for a short term and the like. The method is high in calculation efficiency, and the tracking algorithm is verified for the first time on an ARM platform.