One of the drastically growing and emerging research in recent days is video surveillance. Object detection and tracking is a challenging task in the surveillance video. Detecting and identifying an abnormal event in a video is a security application used in various emerging industries like military, airport, remote surveillance and etc. Tracking an object in a video based on the time is a challenging task in video processing. In earlier research works, video processing tools have been developed for video processing. Software tools detect the position of the objects in the image and segment the objects. The objects are detected using boundaries. But the boundary-based object detection is not highly accurate. Considering this problem, this paper motivated to improve the accuracy of the object detection and classification. To do this, this paper proposed a novel Feature Incorporation Model (FIM) for improving the object detection and activity recognition. FIM used spatio temporal feature process using Convolutional Neural Network and Multiclass Support Vector Machine for classification. The proposed approach is experimented and tested over UCF and KTH benchmark datasets. The performance of the FIM is evaluated by comparing with the detection and classification accuracy with the state-of-the-art methods. Finally, it is identified that FIM outperforms than the other approaches.
Volume 11 | Issue 6
Pages: 282-292