基于视频侦查的特征识别技术研究

Research on feature recognition technology based on video detection

提出一种基于超像素分割的行人识别算法。根据图像填充方法进行前景区域填充,恢复图像行人区域完整性,再进行图像的超像素块分割。提取超像素块的像素点特征,采用KNN算法建立超像素块聚类中心直方图作为特种量,得到图像在目标图像的投影区域,寻找最相似超像素块,计算最相似超像素块的最近邻距离作为当前行人与目标行人间距离,采用相似度排名和统计得到行人的识别结果。通过实例验证表明:超像素分割行人识别算法识别度相对较高,且在复杂环境下的识别精确度远高于传统算法。

This paper proposes a pedestrian recognition algorithm based on hyper pixel segmentation.According to the image filling method,the foreground area is filled to restore the integrity of the pedestrian area of the image,and then the image is segmented by super pixel block.Pixel features of ultra fast pixel is extracted,and clustering center pixel block histogram is established as a special quantity by using KNN algorithm calculation to get the projection area of the image in the target image,and find the most similar super fast pixel.Calculating the nearest neighbor distance of similar pixel block as the distance between current pedestrians and targeted pedestrians,and using similarity ranking and statistics to obtain the recognition results of pedestrians.The example shows that the hyperpixel segmentation pedestrian recognition algorithm proposed has a relatively high recognition degree and the recognition accuracy is much higher than that of the traditional algorithm in complex environment.