Improvising Enhanced Laplacian Thresholding Technique For Efficient Moving Object Detection In Video Surveillance

摘要

Video Object Detection is more demanding in various video surveillance application i.e., public and private domains. The video object detection, identify similarity of objects and object parts between consecutive frames of video. The moving object detection, analyze the video frames with its foreground and back ground image objects. The foreground image objects are usually considered to be moving ones and the background objects are considered to be static. Recently several contributions in the video object detection had been made, but lacks detection accuracy and unaddressed background object detection in the video frames. In this paper, a novel framework of object detection for video surveillance called Improvised Enhanced Laplacian Threshold (IELT) technique. The improvisation of ELT is done with Gaussian-based Neighbourhood Intensity Proportion (GNIP). The IELT technique initiates the process of video object segmentation, object tracking and finally object detection. In video object segmentation, the input video frames are segmented with the help of Median Filter-based Enhanced Laplacian Thresholding to improve the video quality. In object tracking, Color Histogram-based Particle Filter is applied to the segmented objects by computing the likelihood function, particle posterior and particle prior function based on the Bayes Sequential Estimation model. Finally, the object detection is performed with improvisation of enhanced Laplacian threshold by analyzing neighbourhood Intensity proportion of moving object contour. IELT with Gaussian distribution of neighbourhood proportion improves video object detection accuracy and identify background moving object detection. Experimental evaluation is done on IELT with performance metrics such as time taken for object segmentation, object tracking and detection accuracy, and peak signal-to-noise ratio of moving video object frames. The data set sample of different videos extracted from Internet Archive 501(c) (3), a non-profit organization on effective video object detection for video surveillance. Experimental analysis shows that the GNIP framework is able to reduce the object segmentation time by 52% and improve the video object detection accuracy by 12% compared to the state-of-the-art works.