Authors: Pawan Kumar Mishra; G.P. Saroha
Addresses: Department of Computer Science and Engineering, Uttarakhand Technical University, Dehradun, Uttarakhand, India ' University Computer Center, Maharshi Dayanand University, Rohtak, Haryana, India
Abstract: A framework has been designed for detection and classification of multiple moving vehicles. Background subtraction is used for detection of multiple moving objects like vehicles using Gaussian mixture model (MOG). Classification for multiple moving vehicles using K-nearest neighbour is done based on different features in this research. The method used in this research also improves the value of accuracy and occlusion rate for multiple moving vehicles in video frames. In this paper, we also learn a single detector for different types of multiple moving vehicles such as buses, trucks, and cars. This detector uses a special kind of function that is known as occlusion metric function. The main goal of this research is to build a function that is used to calculate the performance of detector between number of false positives and hit rate in heavy traffic (high activity) and small traffic (low activity) region.
Keywords: detection; classification; occlusion; accuracy; false positive; hit rate.
International Journal of Computational Vision and Robotics, 2020 Vol.10 No.2, pp.167 - 184
Received: 08 Nov 2018
Accepted: 19 Apr 2019
Published online: 04 Mar 2020 *