DefinePK hosts the largest index of Pakistani journals, research articles, news headlines, and videos. It also offers chapter-level book search.
Title: Double threshold based method for person re -identification using significant colour matching in the spatial correspondence regions
Authors: M. K. Vidhyalakshmi, E. Poovammal, V. Masilamani, Vidhyacharan Bhaskar
Journal: ARPN Journal of Engineering and Applied Sciences
Publisher: Khyber Medical College, Peshawar
Country: Pakistan
Year: 2020
Volume: 15
Issue: 10
Language: English
Person re-identification is a technique which has an objective to identify a person captured in different views in a community space monitored by network of numerous non-overlapping surveillance cameras. For better view over a region, surveillance camera is usually placed at a height more than person’s height. With this set up , getting the facial features of the person becomes very difficult. Moreover, the cameras used will be of less resolution and other conditions including view angle, lighting conditions, occlusion and pose change are amongst the challenges which prevents to get essential facial features from the person’s image. In this paper we have proposed double threshold-based person re-identification using significant colour matching in the spatial correspondence regions, which uses clothing colour as key feature to re-identify a person. Experiments are conducted on images in underground re-identification (GRID) dataset and results of the Cumulative Matching Characteristic (CMC) curve in evaluation with existing methods establish the efficiency of our method for person re-identification. With this proposed method we have achieved the first rank accuracy as 40.1% with the GRID database.
Loading PDF...
Loading Statistics...