44results about How to "Tracking results are accurate" patented technology

The embodiment of the invention provides an object tracking method and a device, and the method comprises the steps: obtaining a plurality of frames of first images photographed by a first photographing device and the first photographing moment of each frame of first image, and the first image comprises a first object; obtaining a plurality of frames of second images shot by a second camera deviceand a second shooting moment of each frame of second image, wherein the second image comprises a second object; obtaining the distance between the first camera device and the second camera device; and judging whether the first object and the second object are the same object according to the multiple frames of first images, the first shooting moment of each frame of first image, the multiple frames of second images, the second shooting moment of each frame of second image and the distance. According to the object tracking method and the device provided by the embodiment of the invention, theproblem that the robustness of the existing scheme is poor due to the fact that the existing scheme is greatly influenced by the image shooting angle can be solved, and the tracking result is more accurate.

The invention discloses an object tracking method, an object tracking device, a storage medium and electronic equipment, and relates to the technical field of computer vision. The target tracking method includes: extracting the global features of the target area image; wherein, the target area image is the area image containing the target object predicted in the previous frame image of the current frame image; The global features of the current frame image are used for attention calibration, and the calibrated feature is used as the target calibration feature of the previous frame image; combined with the target calibration feature of the previous frame image, the target judgment feature of the current frame image is determined; the current frame image is determined. feature, determine the feature part that matches the target judgment feature from the global feature of the current frame image, and determine the image area corresponding to the feature part as the predicted area containing the target object in the current frame image, so as to realize the tracking of the target object . The present disclosure can improve the accuracy of target tracking.

The invention discloses a multi-target tracking method based on fusion detection technology. The method firstly models the surveillance video and pictures, outputs the category and location information of the concerned target stably and accurately, and tracks them. Specifically, first classify and locate the target in the picture through the target detector, and then use the tracker based on motion modeling to continuously predict the target trajectory, and then input the detection area and motion prediction area into the shape feature acquisition network to obtain their respective Based on the matching features of the target position and matching features, the data association between the two and the correction of the final tracking result box are completed. This method can detect target information accurately and quickly. At the same time, the motion prediction position and matching feature network make full use of the target motion and shape features, and the result of multi-target tracking is more accurate through matching and frame correction.

The invention provides a multi-body tracking method based on a human body part model, and relates to the technical field of video monitoring. It includes: acquiring image data, extracting the human body part model of each person and calculating color features, obtaining each person's part feature set, and assembling to obtain a list of multiple human body part feature sets. Then calculate the similarity between each person's part feature set obtained in the current frame and each person in the multi-body part feature set list obtained in the previous frame to obtain a matching matrix. Then, according to the matching matrix, calculate the similarity confidence of each person obtained in the current frame to each person obtained in the previous frame, and match the target obtained in the current frame with the target in the previous frame according to the similarity and similarity confidence to obtain the tracking result . This method can accurately match the detection target of the current frame to the position of the previous frame and discover a new target, effectively solving the problem of tracking failure due to feature differences caused by human body deformation in the process of multi-body tracking.

The invention discloses a target tracking method and system based on triple convolutional network and perceptual interference learning, which belongs to the field of target tracking research in image processing and machine vision. The method includes: inputting the video to be tracked into the triple convolutional network to obtain target tracking Results; the training of the triple convolutional network includes: constructing the triple convolutional network, obtaining the positive sample pair and the negative sample pair from the data set to obtain the training set; using the training set to train the triple convolutional network, two images of each sample pair in the training set Input the template branch and the detection branch respectively, or input the first frame branch and the detection branch respectively; the template branch and the first frame branch respectively extract the appearance model feature map, and compare the two appearance model feature maps with the feature map of the detection branch respectively. Cross-correlate to get two response maps; respectively calculate the loss of the two response maps for backpropagation, and thus obtain a trained triple convolutional network. The method of the invention has higher target tracking accuracy.

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 present invention provides a multi-target tracking algorithm based on multilayer restriction. The multi-target tracking algorithm comprises: employing a simple strategy to rapidly obtain tracking segments, employing vision information to perform correction and segmentation of all the tracking segments, converting a result to a graph model about the tracking segments for solution, employing high-level semantic information of the result obtained through solution to perform further correction, and obtaining a final tracking target track. The multilayer restriction multi-target tracking algorithm fully employs the vision feature of a tracking target to perform restriction layer by layer from shallow to deep so as to solve the problem that an accurate result is difficult to obtain because a current method is difficult to fully employ the vision feature and allow the tracking result to be accurate.