Siamese Network Object Tracking Algorithm Combining Attention Mechanism and Correlation Filter Theory

Aiming to solve the problem of tracking drift during movement, which was caused by the lack of discriminability of the feature information and the failure of a fixed template to adapt to the change of object appearance, the paper proposes an object tracking algorithm combining attention mechanism and correlation filter theory based on the framework of full convolutional Siamese neural networks. Firstly, the apparent information is processed by using the attention mechanism thought, where the object and search area features are optimized according to the spatial attention and channel attention module. At the same time, the cross-attention module is introduced to process the template branch and search area branch, respectively, which makes full use of the diversified context information of the search area. Then, the background perception correlation filter model with scale adaptation and learning rate adjustment is adopted into the model construction, using as a layer in the network model to realize the object template update. Finally, the optimal object location is determined according to the confidence map with similarity calculation. Experimental results show that the designed method in the paper can promote the object tracking performance under various challenging environments effectively; the success rate increases by 16.2%, and the accuracy rate increases by 16%.

Real-Time Analysis of Athletes’ Physical Condition in Training Based on Video Monitoring Technology of Optical Flow Equation

This article is dedicated to the research of video motion segmentation algorithms based on optical flow equations. First, some mainstream segmentation algorithms are studied, and on this basis, a segmentation algorithm for spectral clustering analysis of athletes’ physical condition in training is proposed. After that, through the analysis of the existing methods, compared with some algorithms that only process a single frame in the video, this article analyzes the continuous multiple frames in the video and extracts the continuous multiple frames of the sampling points through the Lucas-Kanade optical flow method. We densely sampled feature points contain as much motion information as possible in the video and then express this motion information through trajectory description and finally achieve segmentation of moving targets through clustering of motion trajectories. At the same time, the basic concepts of image segmentation and video motion target segmentation are described, and the division standards of different video motion segmentation algorithms and their respective advantages and disadvantages are analyzed. The experiment determines the initial template by comparing the gray-scale variance of the image, uses the characteristic optical flow to estimate the search area of the initial template in the next frame, reduces the matching time, judges the template similarity according to the Hausdorff distance, and uses the adaptive weighted template update method for the templates with large deviations. The simulation results show that the algorithm can achieve long-term stable tracking of moving targets in the mine, and it can also achieve continuous tracking of partially occluded moving targets.

Target Tracking Algorithm Based on Adaptive Scale Detection Learning

In this paper, to better solve the problem of low tracking accuracy caused by the sudden change of target scale, we design and propose an adaptive scale mutation tracking algorithm using a deep learning network to detect the target first and then track it using the kernel correlation filtering method and verify the effectiveness of the model through experiments. The improvement point of this paper is to change the traditional kernel correlation filtering algorithm to detect and track at the same time and to combine deep learning with traditional kernel correlation filtering tracking to apply in the process of target tracking; the addition of deep learning network not only can learn more accurate feature representation but also can more effectively cope with the low resolution of video sequences, so that the algorithm in the case of scale mutation achieves more accurate target tracking in the case of scale mutation. To verify the effectiveness of this method in the case of scale mutation, four evaluation criteria, namely, average accuracy, cross-ratio accuracy, temporal robustness, and spatial robustness, are combined to demonstrate the effectiveness of the algorithm in the case of scale mutation. The experimental results verify that the joint detection strategy plays a good role in correcting the tracking drift caused by the subsequent abrupt change of the target scale and the effectiveness of the adaptive template update strategy. By adaptively changing the number of interval frames of neural network redetection to improve the tracking performance, the tracking speed is improved after the fusion of correlation filtering and neural network, and the combination of both is promoted for better application in target tracking tasks.

An Anchor-Free Siamese Network with Multi-Template Update for Object Tracking

Siamese trackers are widely used in various fields for their advantages of balancing speed and accuracy. Compared with the anchor-based method, the anchor-free-based approach can reach faster speeds without any drop in precision. Inspired by the Siamese network and anchor-free idea, an anchor-free Siamese network (AFSN) with multi-template updates for object tracking is proposed. To improve tracking performance, a dual-fusion method is adopted in which the multi-layer features and multiple prediction results are combined respectively. The low-level feature maps are concatenated with the high-level feature maps to make full use of both spatial and semantic information. To make the results as stable as possible, the final results are obtained by combining multiple prediction results. Aiming at the template update, a high-confidence multi-template update mechanism is used. The average peak to correlation energy is used to determine whether the template should be updated. We use the anchor-free network to implement object tracking in a per-pixel manner, which computes the object category and bounding boxes directly. Experimental results indicate that the average overlap and success rate of the proposed algorithm increase by about 5% and 10%, respectively, compared to the SiamRPN++ algorithm when running on the dataset of GOT-10k (Generic Object Tracking Benchmark).

Siamese Networks-Based People Tracking Using Template Update for 360-Degree Videos Using EAC Format

Rich information is provided by 360-degree videos. However, non-uniform geometric deformation caused by sphere-to-plane projection significantly decreases tracking accuracy of existing trackers, and the huge amount of data makes it difficult to achieve real-time tracking. Thus, this paper proposes a Siamese networks-based people tracker using template update for 360-degree equi-angular cubemap (EAC) format videos. Face stitching overcomes the problem of content discontinuity of the EAC format and avoids raising new geometric deformation in stitched images. Fully convolutional Siamese networks enable tracking at high speed. Mostly important, to be robust against combination of non-uniform geometric deformation of the EAC format and partial occlusions caused by zero padding in stitched images, this paper proposes a novel Bayes classifier-based timing detector of template update by referring to the linear discriminant feature and statistics of a score map generated by Siamese networks. Experimental results show that the proposed scheme significantly improves tracking accuracy of the fully convolutional Siamese networks SiamFC on the EAC format with operation beyond the frame acquisition rate. Moreover, the proposed score map-based timing detector of template update outperforms state-of-the-art score map-based timing detectors.