Suspected Vehicle Detection for Driving Without License Plate Using Symmelets and Edge Connectivity

Abstract

This paper proposes a novel suspected vehicle detection (SVD) system for detecting vehicles that are travelling on roads without a license plate. We start with detecting vehicles in a still image by utilizing a symmelet-based approach which allows us to determine a vehicle's region of interests (ROIs). A symmelet is a pair of an interest point and its corresponding symmetrical one. We modify the nonsymmetrical SURF descriptor into a symmetrical one in which no additional time complexity is added and no motion feature is required. This method allows for different symmelets to be efficiently extracted from road scenes. The set of symmelets can be used to locate the desired vehicle's ROI with the use of a projection technique. We then examine the existence of a license plate within this ROI, with an edge connectivity scheme that highlights possible character regions for plate detection. This SVD system provides two advantages; the background of the image does not need to be subtracted from analysis and the system does not require the use of a GPU. It is extremely efficient for real-time intelligent transport system (ITS) applications.