融合色调和局部特征的军人图片识别方法

摘要

Aiming at the problems of the bad robustness of the military uniform recognition under the conditions of the occlusion and illumination changes ,a high-speed recognition method of armyman’s photographs was proposed .Photographs were filtered using the high-speed facial recognition and the location algorithm ,and the upper part of the body of the target photographs was extracted .Then the photographs were classified by using the support vector machine based on the SIFT-BoW+ Hue fusion fea-ture .The experimental results show that the new method can get more than 90% accuracy under different conditions such as dif-ferent illuminations ,resolutions ,postures of man ,and even under the condition of the photograph covered partially .%针对目前军人图片识别方法在遮挡、光照变化情况下鲁棒性较差的问题，提出一种军人图片快速识别方法。在利用快速人脸识别和定位算法对图片进行过滤并提取出人物图片上半身区域的基础上，融合该区域SIFT-BoW和 Hue直方图特征并用支持向量机进行分类。实验结果表明，该方法在不同光照强度、分辨率、人物姿态及部分遮挡条件下仍能够达到90％以上的识别率。