摘要
基于图像的车辆匹配是图像匹配在智能交通领域内的具体应用。为了实现车辆对象的快速匹配,文中提出一种多维特征融合的车辆对象同一性匹配方法。该方法分为两个阶段进行。第一阶段先对车辆对象进行初步筛选,提取车辆对象的颜色特征和车型进行快速匹配,计算特征向量的欧氏距离,排除最不可能相似的车辆对象;第二阶段根据摄像头物理条件及视频编码格式对车辆的局部特征、感知哈希特征进行加权多比较,进行车辆的同一性精细化匹配。这样第一阶段匹配完成后,第二阶段简化为只在车辆大类内匹配,缩小了匹配范围。实验结果表明,该方法能够有效缩小车辆匹配范围,匹配到最有可能和目标车辆是同一车辆的准确率较高。
Vehicles based on image matching is image matching in the specific application in the field of intelligent transportation. In order to achieve quick matching for vehicle object,a method of vehicle object identity matching based on multidimensional feature combination is proposed. This method can be divided into two stages. First,the vehicle object is carried on the preliminary selection,extracting color features and models for fast matching,calculating the Euclidean distance of feature vector to exclude the impossible similar vehicles.Then,according to the physical condition of camera and video coding format,the local features of the vehicle and perceptual hash features are compared to complete the vehicle＇s identity matching. After completion of the first phase matching,the second phase is simplified to only within the vehicle types match,narrowing the scope of the match. The experimental results showthat this method can effectively reduce the vehicle matching range,and match to the most likely to be the same as target vehicle with high accuracy.
出处
《计算机技术与发展》
2016年第4期167-171,176,共6页
Computer Technology and Development
基金
国家自然科学基金资助项目(61363043)
关键词
特征融合
同一性
图像匹配
局部特征
感知哈希特征
feature combination
identity
image matching
local feature
perceptual hash feature
