120 results about "Visual target tracking" patented technology

A visual target tracking method of a full convolution class and regression twin network structure comprises the following steps: (1) according to the position of a target in an image, cutting a targettemplate image and a search area image in an original training set, and forming a training data set by cut image pairs; (2) establishing a full convolution twin network to extract image features; (3)establishing a classification regression network; (4) in response to the fact that each pixel point on the image has a corresponding foreground score and a predicted bounding box, calculating the total score of each pixel point by combining the information of the foreground score and the information of the bounding box, wherein the pixel point with the highest total score is the center of the tracking target; and (5) training the full convolution twin network and the classification regression network by using the training data set to obtain the trained full convolution twin network and the classification regression network, calculating a score graph of a target in the to-be-tested image sequence by using the trained networks, and performing target positioning based on the score graph. According to the invention, the tracking precision and speed are improved.

The invention relates to a combined judging strategy-based visual target tracking method. The method comprises the following steps of: (1) aiming at a target size change problem, establishing a self-adaptive scale so as to automatically adjust the size of a tracking frame; (2) aiming at a problem that a position of a target cannot be continuously determined after the target is sheltered, predicting a current observed quantity by adoption of Kalman filtering, and after the target reappears, continuously tracking the target by using a KCF algorithm; and (3) aiming at a problem of tracking failure caused by rapid movement of the target. According to the method, a movement speed of the target is calculated through detecting a position movement distance of the target in adjacent frames, and a detection area size extension coefficient is self-adaptively adjusted, so that relatively high precision can be obtained under different environments, the robustness of the whole tracking system is effectively improved, a relatively high calculation speed is kept, and a high engineering practical value is provided.

The invention discloses a quad-rotor unmanned aerial vehicle visual target tracking method based on a binocular camera. The method comprises the following steps of detecting a tracking target througha target recognition algorithm, finishing the position tracking and scale tracking of the target in pixel significance by using a visual tracking algorithm based on correlation filtering, and judgingwhether a repositioning program needs to be started or a long-term tracker needs to be updated according to a tracking effect; secondly, according to an image area selected by the tracking frame, calculating the relative distance between the quad-rotor unmanned aerial vehicle and the tracking target by using an LK optical flow method, and after coordinate conversion, realizing the global state estimation of the tracking target by using a Kalman filter; and finally, according to the estimated global position and speed state of the target, designing the state quantity of an outer loop control system of the unmanned aerial vehicle to realize the delay-free stable tracking of the tracked target by the unmanned aerial vehicle.

The invention relates to a smart visual holder system and a realization method thereof. The system comprises an image collection module which is used for collecting high-speed high-definition sequence images and sending the collected images to a holder master control module; a tri-axial stabilizing holder which is used for controlling a posture of the image collection module; the holder master control module which interacts with the image collection module, the tri-axial stabilizing holder and an image processing module and is used for carrying out video coding and storage on the images collected by the image collection module; and the image processing module which is used for carrying out format conversion on the images received by the holder master control module from the image collection module and processing the images, thereby finishing visual target tracking under moving target indication and a variable field of view, and feeding back a processing result to the holder master control module. The technical problem that an existing system is low in intelligentization degree and single in function is solved. Two intelligent visual functions of moving target indication and variable field of view target continuous tracking are integrated, thereby facilitating use.

The invention discloses an optimization method of a visual target tracking method based on particle filtering and optical flow vector, belonging to the technical field of image/video post-processing. The method comprises the following steps: performing the parallel computing of each huge computation algorithms involved in the visual target tracking method based on particle filtering and optical flow vector in multiple CPUs, running a thread on each CPU to process a part of line data, evenly distributing all the line data to each CPU; and sending an event report to a thread synchronization manager after completing the task of each thread, and starting all the threads by the thread synchronization manager, when all the threads complete the current task, to start the subsequent task event. By using the method of the invention, the efficiency of the visual target tracking method based on particle filtering and optical flow vector can be increased, and the requirement on the real-time of the tracking method can be satisfied.

A method for controlling a movable object to track a target object may be provided. The method may comprise: determining a change in one or more features between a first image frame and a second imageframe, wherein the one or more features are associated with the target object, and wherein the first image frame and the second image frame are captured at different points in time; and adjusting a movement of the movable object based on the change in the one or more features between the first image frame and the second image frame.

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.

The invention belongs to the visual tracking field and relates to an adaptive sub-block screening-based multi-clue visual tracking method. The method comprises the following steps that: (1) saliency detection is performed on a target region, uniform block division is used in combination, so that candidate sub-blocks can be obtained; (2) multi-scale sampling is performed on the candidate sub-blocks, sub-blocks with large frequency-domain response and the corresponding scales of the sub-blocks are determined, and a candidate sub-block set is updated; (3) motion estimation is performed on the sub-blocks in the candidate sub-block set, and the current location of a tracking target is determined through the multi-clue fusion of the sub-blocks; and (4) a Gaussian kernel corresponding to the location of each sub-block is updated through the current location of the target, and sub-blocks which do not satisfy requirements are re-initialized. According to the adaptive sub-block screening-based multi-clue visual tracking method of the invention adopted, the interference of background can be removed, the visual constraints of middle-level features and the priori constraints of high-level languages are fully utilized, so that the locating of the target is more accurate. The adaptive sub-block screening-based multi-clue visual tracking method has the advantages of simple steps and small computation amount, and is suitable for performing visual target tracking under a blocking condition.

The invention discloses a local distance study and sequencing queue-based visual target tracking method, which comprises: step 1, selecting a target and the adjacent background of the target in a first frame image by using a target frame and a background frame, randomly sampling in each frame to obtain two small image sheet sets representing the target and the local background of the target; studying the local distance metric function of each target small image sheet and establishing a sequencing queue of the function, and calculating the purity of the sequencing queue and establishing a target model; step 2, randomly sampling a next frame image to obtain a new small image sheet set; calculating the distances among each small image sheet in the target model and all new small image sheets, and establishing a sequencing queue; and calculating confidence coefficients of the new small image sheets according to the positions of the new small image sheets in each sequencing queue and establishing a confidence graph; step 3, determining the position of the target in a new frame image by using the confidence graph; step 4, updating the target small image sheet set and a background small image sheet set; and step 5, updating the target model, the local distance metric function and the purity, and returning to the step 2.

The invention relates to a visual target tracking method based on scale adaptation and occlusion detection. The method comprises the following steps of according to a target position and a size determined by a previous frame, cutting an image block in a current frame and extracting the convolution characteristics of different layers as sample characteristic graphs; in each layer of the characteristic graph, using a nuclear correlation filtering method to acquire a response graph, and then linearly superposing the different layers of response graphs to acquire a response general graph, whereina position corresponding to a maximum value is a the target position of a current frame; collecting different sizes of samples at the target position and adjusting to a same size, and acquiring a scale response via a scale filter, wherein a scale corresponding to a maximum value is the optimum scale of the current frame; calculating the peak value side lobe ratio of the response general graph anddetermining whether a target is occluded; when the target is occluded, using a space-time context model to re-determine the target position; and updating the model and preparing for the determinationof the target position and the size of the next frame. In the invention, the accuracy and the robustness of visual target tracking are increased.

The invention relates to a neutrosophic similarity measurement-based scale-adaptive visual target tracking method. The method comprises the following steps of: selecting a to-be-tracked target area inan initial frame and calculating a target feature histogram and an initial background histogram; carrying out truth, falsity and indeterminacy measurement aiming at target feature attributes and background feature similarity attributes; establishing a neutrosophic weight vector; introducing the neutrosophic weight vector into a mean shift strategy to determine a target area of a current frame; calculating corresponding truth, falsity and indeterminacy measuring values aiming at scale reducing and expanding and determining a scale updating strategy according to cosine similarity measurement; and updating a target background feature histogram. The method disclosed by the invention has the beneficial effects that an extremely efficient mean shift algorithm is adopted, the corresponding neutrosophic measurement calculating amount is small, the weight vector and scale estimating is low in complexity and high in efficiency and the requirements of real-time target tracking are met; and by utilizing a neutrosophic set theory, the tracking performance of a tracking algorithm coping with challenges of complex backgrounds and like is effectively improved through taking the change of trackedtarget features and the similarity of target / background features into account.

The invention discloses a weak structure perception visual target tracking method capable of fusing with context detection. During initialization, a weak structure relationship between a target and each component of surrounding environment is perceived to establish a model. Model maintenance corresponds to the target and two surrounding component sets, and a feature point and a feature descriptor are used for expressing the appearances of the components. In a tracking process, the component sets are combined with a movement model to generate a potential target center, then, the potential target center is clustered to reject noise to obtain an accurate target position, and a target size is updated. Under a weak structure tracking frame, a bottom-up way and a top-down way are introduced to carry out target context detection in order to enhance the prediction of a component position. Bottom-up detection provides consistent tracking information for each component through the estimation of the local movement of a pixel level. Top-down detection constructs a superpixel nuclear model to learn a difference between the target and a background on a level of individual, and guidance information is provided for target positioning and model update.

The invention discloses a target tracking method based on a three-branch neural network, and belongs to the technical field of computer vision. Visual target tracking belongs to video analysis and serves as an important branch in the field of computer vision, and the basic task of the visual target tracking is to predict the position, area and motion track of a target in a video sequence accordingto given position information of the target in an initial frame. The precision and speed of visual target tracking are low, and the visual target tracking is easily affected by shielding, backgroundconfusion, size change, severe appearance change, illumination change and the like. The invention provides a target tracking method based on a three-branch neural network. Compared with the traditional visual target tracking technology, the three-branch neural network is utilized to track the target, so that the target can be represented with high robustness, the obvious change of the appearance of the target can be handled, the background can be well distinguished, and meanwhile, the drifting of the algorithm can be effectively avoided. And the tracking speed is far higher than that of otheralgorithms.

The invention discloses an unmanned aerial vehicle visual target tracking method based on scale self-adaptive kernel correlation filtering, which comprises the following steps of selecting a trackingtarget, calculating to obtain the color and gradient initial probability density of a first frame of the tracking target, and training a classifier and detecting the central position of the target byusing the kernel correlation filtering algorithm for the first frame of data; establishing a one-dimensional kernel correlation filter from the second frame to detect the change of the target scale, and calculating kernel correlation filtering by using a convolution theorem; constructing a similarity function by utilizing the current target feature and the initial feature, if the similarity is smaller than a set threshold value, considering that the target identification is inaccurate or the target is lost, entering global search, otherwise, representing that the target is identified and tracked, and obtaining target position information; and sending the position information of the tracking target to an unmanned aerial vehicle flight control system in real time to control the position of the unmanned aerial vehicle. According to the method, the problem of fixed tracking scale of a kernel correlation filtering algorithm is optimized, and the tracking precision of target characteristicsis effectively improved.

The invention belongs to the technical field of visual target tracking, and discloses a rotor operation flying robot target tracking method and system, and the method comprises the steps: taking a Siamfc framework as the basis, introducing Resnet50 as a feature extraction network through offset learning, enabling the network to learn more semantic information, and coping with the appearance changeof a target; according to the tracking network, a target scale estimation module is newly added on the basis of a classification discriminator, an IOU of a target bounding box and a target real box can be predicted, the target bounding box is accurately predicted, iterative correction is carried out on the bounding box through reverse gradient, and thus the network can accurately predict the scale change of a target; output of different layers of the network is fused by utilizing Resnet50 multi-layer feature output and adopting a residual fusion strategy, so that the robustness of the algorithm is further improved, the network performance is improved, the discrimination capability of the network to a small target is guaranteed, and finally, accurate tracking of the target is realized.

The invention claims for protection of a spatiotemporal context visual tracking method with fusion of particle filtering. According to the method, first frame of image information is initialized, experimental parameters are set, and then a rectangular region in which a first frame of target is located is automatically selected; whether the target is blocked is determined by using a Bhattacharyya coefficient as a base; and when the target is blocked, a location and a motion track of the target at a follow-up image frame are estimated and predicted by introducing a particle filter algorithm, thereby realizing precise tracking of the target. Therefore, the method is suitable for visual target tracking under complicated background like illumination changing, target rotation, and background region interference; the robustness of target blocking is high and the real-time requirement is met.

The invention relates to the computer video processing technology, in particular to a visual target tracking method based on a credibility combination map model. The method comprises the following steps that (1) a training database is established; (2) features of the training database are extracted, and a two-dimensional disjunction unit classifier and a two-dimensional disjunction classifier are trained; (3) a credibility combination map of first-frame target objects is established; (4) features of a current-frame background frame are extracted; (5) a credibility graph is obtained; (6) a target is positioned, and a plurality of candidate windows are obtained; (7) a credibility combination map of the candidate windows is matched with the saved previous-frame credibility combination map, and optimal target location information is obtained; (8) an updating sample is obtained by means of combination map matching, and the classifiers, the credibility map model, the state of a tracker and the like are updated every five frames; (9) the step (4), the step (5), the step (6), the step (7) and the step (8) are repeated till a video is over. By means of the visual target tracking method based on the credibility combination map model, the problem of target drifting can be effectively restrained in the computer visual target tracking process, and therefore the stability of the tracker is improved.

The invention provides a multi-thread visual target tracking method based on STC and block re-detection, and the method comprises the following steps: S1, reading a first frame image, and determininga tracking target; S2, establishing a spatial context model for the first frame of image by adopting an STC algorithm; S3, performing blocking operation on a rectangular template area where a trackingtarget in the first frame image is located, and training an SVM classifier; s4, reading a next frame of image, learning from a previous frame of image to obtain a spatial context model, and calculating target neighborhood context priori; s5, updating the space-time context model of the current frame image; s6, obtaining a confidence map of the current frame image; s7, judging the shielded degreeof the tracking target in the current frame image according to the confidence probability; s8, selecting a corresponding processing strategy according to the judged shielding condition of the trackingtarget; and S9, circularly executing the steps S4-S8 until the current video or image sequence is completely processed. According to the invention, the target tracking reliability and the target tracking efficiency can be improved.

The invention relates to the field of visual target tracking, and discloses a single visual target tracking algorithm and system based on a deep neural network. The basic principle of the method is asfollows: a tracking target is specified by a first frame in an image sequence, target features and to-be-searched region features are extracted by adopting the same convolutional network in subsequent frames, convolution and foreground-background distinguishing networks are performed to obtain a target position, and the width and height of a target frame are obtained through regression, so that aregion frame where the target is located is obtained; and when the confidence value of the tracking target is lower than a certain degree, considering that problems such as target loss may occur, andperforming re-search by adopting a re-search strategy to ensure the tracking effect of the target. For targets of different sizes, the sizes of the targets are input into a size adjusting module, sothat the sample cutting size is dynamically adjusted according to the target sizes. Therefore, the template can adapt to targets with different sizes and different motion characteristics, and the tracking performance is improved.

The embodiment of the invention relates to the technical field of visual target tracking, and discloses a KCF tracking target loss detection method and system based on foreground detection. The KCF tracking target loss detection method based on foreground detection comprises: extracting a foreground image by adopting a foreground detection algorithm; extracting a target detection frame by using aKCF tracking algorithm; obtaining a foreground region corresponding to the target detection frame, calculating the ratio of the target contour area in the foreground area to the total area of the foreground area; judging whether a KCF tracking target is lost or not by setting a ratio threshold. According to the KCF target tracking method and device, the problems that the KCF algorithm cannot be perceived after the target is lost and the KCF tracker takes the background information as the target to continue tracking are solved, and accurate judgment when the tracking target is shielded in a large area or leaves a lens is realized on the basis of not influencing the real-time performance of KCF target tracking.

The invention discloses a robust visual target tracking method suitable for long-range tracking. The method includes: extracting positive and negative samples according to an initial frame image of avideo sequence and position information of a target in an initial frame, performing feature extraction on sample image blocks to obtain low-dimensional eigenvectors, using the linear support vector machine technique to initialize a target appearance model; performing logistic regression on the obtained support vector machine model, and estimating a target position for the target appearance model under a particle filtering framework; combining the median flow tracking algorithm and the current particle filtering algorithm to perform collaborative tracking, using the incremental reduction technique to update the appearance model online in the tracking process, and combining the original appearance model and new samples to update the appearance model online till the last frame is updated. Inthis way, the visual target tracking of the robust is achieved. The parallel complementation of the two-way tracking method with different mechanisms is achieved, and the problem of spatial redundancycaused by continuously generating new information in the tracking process can be solved.

The invention discloses a visual target tracking method and a visual target tracking system oriented to an image sequence. The visual target tracking method comprises the following steps of training aconvolutional regression model for target tracking by use of a given initialized image and a to-be-tracked target rectangular frame; predicating the position of a target by use of the convolutional regression model obtained by training; further predicting the size of the target on the basis of a prediction result of the target position; and updating the convolutional regression model according tothe position and the size of the target obtained by tracking. According to the visual target tracking method and the visual target tracking system oriented to the image sequence, technologies such astraining of a target overall regression model, training of a target texture regression model, prediction of the target position, prediction of the target size, updating of a tracking model and the like are involved, the interference of various environmental factors in a tracking scene can be fully overcome to implement accurate prediction for the target position and size, and relatively higher commercial values and research significances are possessed.

The embodiment of the invention discloses a model training method and device, a terminal and a storage medium. The method comprises: acquiring a template image and a test image; calling the first object recognition model to process the characteristics of the tracking object in the template image to obtain a first reference response, and calling the second object recognition model to process the characteristics of the tracking object in the template image to obtain a first reference response; calling the first object recognition model to process the characteristics of the tracking object in thetest image to obtain a first test response, and calling the second object recognition model to process the characteristics of the tracking object in the test image to obtain a second test response; tracking the first test response to obtain a tracking response of the tracked object; and updating the first object recognition model based on the difference information between the first reference response and the second reference response, the difference information between the first test response and the second test response and the difference information between the tracking tag and the tracking response. According to the embodiment of the invention, the accuracy of visual target tracking can be improved.

The invention belongs to the technical field of computer vision, and provides a moving target visual tracking method based on multi-source information fusion. Aiming at a moving target visual tracking task in a scene with rapid movement and severe illumination, the invention firstly manufactures a moving target tracking data set based on an event camera, and meanwhile, provides a visual target tracking algorithm based on cross-domain attention to accurately track a visual target based on the data set. According to the invention, the respective advantages of the frame image and the event data can be utilized and combined; the frame image can provide rich texture information; and the event data can still provide clear object edge information in a challenging scene. By respectively setting the weights of the two kinds of domain information in different scenes, the advantages of the two kinds of sensors can be effectively fused, so that the target tracking problem under complex conditions is solved.