基于孤立像素点去除的车牌ROI区域快速定位方法研究

车牌识别（LPR）作为当前较为广泛研究的热门课题之一，已经日趋成熟和完善，但传统车牌提取算法中仍然存在对单个像素点过分依赖、识别率不高等问题。提出一种基于孤立像素点去除的车牌感兴趣区域提取方法，在Sobel算子边缘检测算法的基础上，采用基于行列比例和周围相似度两种方法，对单像素点进行处理，有效去除影响感兴趣区域中的孤立点，突出车牌特征并提高识别精度。实验结果表明，该方法较传统方法识别速率平均提高约5%，识别精度平均提高约15%。

Research on the Method of License ROI Plate Fast Location Based on Isolated Pixel Removal

License plate recognition is widely used as the current hot topic of research, which has been matured and improved, but the traditional li-cense plate extraction algorithm still exist over-reliance on a single pixel, low recognition rate. Presents a license plate extraction method based on region of interest isolated pixel removal, the Sobel operator-based edge detection algorithm, using the similarity of the two meth-ods of processing a single pixel proportion based on the ranks and around the effective removal effect of the region of interest isolated point, and the projecting plate to improve the recognition accuracy characteristics. Experimental results show that the method used by ap-proximately 5%on average compared with the conventional method of identifying the rate of recognition accuracy average of approximate-ly 15%.