Unified Graph-Based Multicue Feature Fusion for Robust Visual Tracking

Walia, Gurjit Singh; Ahuja, Himanshu; Kumar, Ashish; Bansal, Nipun; Sharma, Kapil

doi:10.1109/tcyb.2019.2920289

Cited by 19 publications

(6 citation statements)

References 44 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…The skeletal data consists of multiple joint point position coordinates, which are related to the reference coordinate system and usually differ in various real scenes. Even in the same shooting scenario, the joint point coordinates of people with the same pose may differ due to differences in height and size, sensor angle, and distance from the sensor [ 18 ]. Therefore, the original skeletal sequences need to be normalized to reduce the effects of skeletal scale and shooting angle.…”

Section: Design Of Behaviour Detection and Recognition System For Hig...mentioning

confidence: 99%

Behaviour Detection and Recognition of College Basketball Players Based on Multimodal Sequence Matching and Deep Neural Networks

Zhang

2022

Computational Intelligence and Neuroscience

View full text Add to dashboard Cite

This study fuses multimodal sequence matching with a deep neural network algorithm for college basketball player behavior detection and recognition to conduct in-depth research and analysis, analyzing the basic components of basketball technical action videos by studying the practical application of technical actions in professional games and teaching videos from self-published authors of short video platforms. The characteristics of the dataset are also analyzed through literature research related to the basketball action dataset. On the established basketball technical action dataset, combined with the SSD target detection algorithm, the video images with cropping of human motion regions to reduce the size of image frames, generating a basketball technical action dataset based on cropped frames, reduces the amount of network training and improves the efficiency of subsequent action recognition training. In this study, by analyzing the characteristics of basic camera motion, a univariate global motion model is proposed to introduce a quadratic term to accurately express the shaking transformation, while the horizontal and vertical motion are independently represented to reduce the model complexity. Comparative experimental results show that the proposed model achieves a good balance between complexity and accuracy of global motion representation. It is suitable for global motion modeling in behavior recognition applications, laying the foundation for global and local motion estimation. On this basis, the visual feature change pattern of the key area of the scene (basketball area) is combined with the group behavior recognition based on motion patterns and the success-failure classification based on key visual information to achieve basketball semantic event recognition. The experimental results at NCAA show that the fusion of global and local motion patterns can effectively improve group behavior recognition performance. The semantic event recognition algorithm combining motion patterns and video key visual information achieves the best performance.

show abstract

Section: Design Of Behaviour Detection and Recognition System For Hig...mentioning

confidence: 99%

Behaviour Detection and Recognition of College Basketball Players Based on Multimodal Sequence Matching and Deep Neural Networks

Zhang

2022

Computational Intelligence and Neuroscience

View full text Add to dashboard Cite

show abstract

“…(4) Model updater Usually model updaters include online classifiers, incremental subspace learning algorithms, and real-time template change updates. Ensuring that the appearance update of the target and the background can be accurately described without causing the model's ability to describe the target is also a major problem in computer vision TT [14].…”

Section: Tt Componentsmentioning

confidence: 99%

Algorithm Research of Neural Network for Feature Target Tracking

2022

View full text Add to dashboard Cite

In recent years, with the explosive growth of research in the field of computer vision, TT has an important application in visual recognition tasks. It can assist target detection and improve the speed of recognition, which has certain theoretical value and research significance. However, in practical application scenarios, target tracking (TT) still faces the problems of inaccurate tracking, poor robustness, and low overall system speed caused by scene changes. Since the TT algorithm based on convolutional neural network (CNN) was proposed, it has attracted the attention of a large number of researchers with the advantages of both speed and accuracy. The multi-layer convolution features extracted from the input image through the CNN have a good appearance representation ability for the target under a variety of complex interference factors. Therefore, this paper uses the CNN to build a feature TT algorithm model, and experiments prove that the accuracy of the model can be improved by increasing the order of magnitude and deepening the ResNet backbone network design. After comparing the tracking effects of the CNN algorithm and other algorithms, it shows that the CNN algorithm has the highest tracking accuracy and success rate.

show abstract

“…Classification results 30 demonstrates that the multicue information unification by non-linear graph method is more precise than linear graph methods. Cross diffusion approach proposed by Walia 38 , et al is employed for feature fusion in the proposed framework. This method is better than previous methods 29,30 and improves the detection accuracy because of the iterative normalisation of similarity matrix and updated sparse representation.…”

Section: Multi-cue Feature Unificationmentioning

confidence: 99%

An Efficient Multistage Fusion Approach for Smartphone Security Analysis

2021

Self Cite

View full text Add to dashboard Cite

Android smartphone ecosystem is inundated with innumerable applications mainly developed by third party contenders leading to high vulnerability of these devices. In addition, proliferation of smartphone usage along with their potential applications in diverse field entice malware community to develop new malwares to attack these devices. In order to overcome these issues, an android malware detection framework is proposed wherein an efficient multistage fusion approach is introduced. For this, a robust unified feature vector is created by fusion of transformed feature matrices corresponding to multi-cue using non-linear graph based cross-diffusion. Unified feature is further subjected to multiple classifiers to obtain their classification scores. Classifier scores are further optimally fused employing Dezert-Smarandache Theory (DSmT). Strength of suggested model is assessed both qualitatively and quantitatively by ten-fold cross-validation on the benchmarked datasets. On an average of outcome, we achieved detection accuracy of 98.97% and F-measure of 0.9936.

show abstract

Unified Graph-Based Multicue Feature Fusion for Robust Visual Tracking

Cited by 19 publications

References 44 publications

Behaviour Detection and Recognition of College Basketball Players Based on Multimodal Sequence Matching and Deep Neural Networks

Behaviour Detection and Recognition of College Basketball Players Based on Multimodal Sequence Matching and Deep Neural Networks

Algorithm Research of Neural Network for Feature Target Tracking

An Efficient Multistage Fusion Approach for Smartphone Security Analysis

Contact Info

Product

Resources

About