An adaptive method for Chinese license plate location

In the vehicle license plate recognition system, the accuracy of location directly decides the effect of latter recognition part. Since the Chinese license plates have complex colors and characters, former methods only can locate certain bounding rectangle of the plate but not the minimum one. Therefore interferences like plate frame and vehicle light may be included in the rectangle, and lead to difficulty in character recognition. This paper proposes a localization method using wavelet analysis and IHLS (improved HLS) color decomposition and Hough line detection. First, a wavelet multi-resolution analysis is implemented to locate the bounding rectangle. Then foreground (text) and plate background are separated by IHLS color decomposition. Through determining the overlap region between foreground and background, a minimum bounding rectangle is extracted from the image. Finally, an optional Hough transform can be used to revise the edge. Because in IHLS space, saturation is completely independent of brightness in theory, this method is much better than former ones under complex background, non-uniform illumination conditions and inclined license plate conditions. The experimental results tested on a large number of data confirm that.