复杂车辆图像中的车牌定位

2512879 2010-12-10 | rar | 1043 | 次下载 | 免费

资料介绍

针对复杂背景的车牌定位问题，提出了一种综合形态、颜色、投影等多种特征的车牌定位算法。基于车牌区域边缘信息丰富的特点，首先利用边缘检测和数学形态学方法定位候选车牌区域，并消除噪声干扰；然后根据质心所在位置对候选区域进行由低到高的排序，并通过颜色识别作进一步筛选，以排除车灯等区域的干扰；最后采用投影法及谷值分析对截取出的缺损车牌进行补全，从而得到准确的车牌位置。实验结果表明，算法的效率与精度与车辆图像的清晰度、曝光度及车牌大小等因素有关，与传统车牌定位算法相比，效率和精度分别提高了15%和20%以上。

Abstract:: A vehicle license-plate locating algorithm is proposed based on the morphology， color and projection features of license-plates. The algorithm uses edge detection and mathematical morphology to locate candidate regions and filter noise， basing on the fact that the license-plate region contains abundant edges. Further screening is then performed by sorting the centroids and identifying the colors of the candidate regions. Finally， restore the incomplete license-plate image using projection method and valley-value analysis， so as to obtain the accurate license-plate location. Experimental results show that the efficiency and accuracy of the algorithm is related to the definition and exposure of the vehicle images and the sizes of the license-plates. Compared to the traditional algorithm， the efficiency and accuracy has been improved 15% and 20% separately.