An optimization-based deep learning for multi object tracking with visual and spatial model
Multiple object tracking (MOT) in video frames is employed for locating the objects, and it is considered a difficult task because of the occlusion of dissimilar targets. Moreover, MOT has been used frequently due to its extensive applications, namely, autonomous driving and surveillance. Further, conventional techniques exploited to track multiple objects reduced the accuracy rate. Thus, a new model termed the Flamingo Hiking Optimization Algorithm with Deep Recurrent Neural Network (FHOA_DRNN) has been devised for MOT. First, the video is fed as input to the extraction process for mining the frames from the video. Then, object segmentation is performed by exploiting Deep Fuzzy Clustering (DFC). Next, the object is detected using ShuffleNet, which is tuned using Jaya Political Search Optimizer (Jaya-PSO). Next, MOT is executed by spatial and visual tracking methods. Here, the visual tracking is presented using the neighborhood search and second derivative model, where the spatial tracking is developed by a Deep Recurrent Neural Network (DRNN). Here, DRNN is tuned by utilizing the Flamingo Hiking Optimization Algorithm (FHOA). Lastly, integration of both tracking models is done for tracking the object. Furthermore, the FHOA_DRNN computed a maximum specificity, accuracy, MOTP, and sensitivity of 93.542%, 94.022%, 91.665%, and 94.858%.
Authors
- T. Manikandan (ORCID: https://orcid.org/0000-0002-8416-279X)
- R. Puviarasi
- T.S. Sindhu
- J Thirumalai
Institutions
- Anna University, Chennai (IN)
- Saveetha University (IN)
Publication Details
- Journal
- Communication in Statistics- Theory and Methods
- Published
- 2026-11-20
- DOI
- https://doi.org/10.1080/03610926.2026.2736786
- Primary Topic
- Video Surveillance and Tracking Methods
- Type
- article
- Field-Weighted Citation Impact
- 0.00