SN Computer Science Computer Graphics and Computer-Aided Design
Improved Vehicle Detection Accuracy and Processing Time for Video Based ITS Applications
Abstract
The increase in daily traffic volume needs a more effective, intelligent, and sophisticated traffic management and control strategy. Video-based traffic monitoring, which incorporates computer vision algorithms, has become one of the most viable and widely utilised intelligent transportation systems (ITS)-based solutions in recent years. The two primary objectives in existing video-based traffic monitoring are to achieve higher accuracy in detecting vehicle and reducing the processing cost. In order to achieve these objectives, a recurrent architecture for parallel vehicle detection scheme (RAP-VDS) with the following two modules: multilevel parallel spatial color information processing (MSCIP) and reduction of redundant temporal color information (R2TCI) are proposed in this work. The MSCIP module uses spatial colour information to increase detection accuracy, whereas R2TCI reduces processing time by eliminating repeated frames over time. The vehicles detected using specified virtual segment within the video frames using RAP-VDS is presented. The results show that irrespective of the computer vision methods used for vehicle detection, by incorporating RAP-VDS there is an improvement in accuracy and processing time.