A perceptual Hash tracking algorithm based on modified Adaboost feature detection is proposed. Firstly,the modified Adaboost detection classifier is used to detect the image features, and extract a number of candidate targets conformable to the requirements for matching. Then, the hash sequence of the candidate target is calculated with perceptual hash algorithm. Finally, the similarity of hash sequences of between candidate targets and the template is decised by hamming distance, thus to determine the highest-similarity target as tracking target in the current frame. The experiments indicate that the proposed algorithm can greatly reduce the computing complexity, in the process of searching and matching resalted by the traditional template hash algorithm. Meanwhile, this algorithm is benefical to overcoming the drift problem in target tracking, re-lacationing and re-tracking after missing of target. This algorithm, high in efficiency, is applicable for long-time and real-time tracking.%提出基于改进Adaboost特征检测的感知哈希跟踪算法.首先,利用改进后的Adaboost检测分类器对图像特征进行检测,提取符合要求的若干个用于匹配的候选目标.其次,利用感知哈希算法,计算该候选目标的的感知哈希序列.最后,通过汉明距离,判定候选目标与模板的相似度,确定当前帧的跟踪目标.实验表明,所提算法能够大大减少传统感知哈希算法模板搜索及匹配过程中的计算量,同时对于克服目标跟踪过程中出现的漂移问题、目标丢失后重定位和恢复跟踪等问题有益,算法效率高,能够用于长时间的实时跟踪.
展开▼
