Towards an automatic vehicle access control system: License plate location

An automatic vehicle access control system (AVACS) can be divided into three steps: vehicle location, vehicle license plate (VLP) location, and VLP recognition. This paper presents a new method for VLP location based on the horizontal gradient, morphological operations, connected components analysis, and statistical measures. First, the horizontal gradient is acquired and a mean filter is applied on it. Morphological operations are then used to darken non-VLP high-valued regions (saliences). At this step, we work on an integer-valued image rather than on the binary one, improving salience detection. Finally, the image is binarized, then a connected component analysis is performed, and statistical measures are used to decide among the VLP candidates. Experiments show that, in a database of 722 images, our method correctly locates the VLP in 95% of the cases, outperforming previous approaches.