Research And Improvement Of Template Matching Tracking Algorithm

Moving target tracking is one of the most important parts of intelligent monitoring system, whether in the military or livelihood areas are equipped with a range of irreplaceable value. Many types of them, performance characteristics, and that in process of tracking, depending on which method we choose, the performance of tracking is definitely very different. Among them, the template matching tracking algorithm has long been a research focus on one project staff, which contains accurate positioning, high efficiency and convenient and practical features of the algorithm is based tracking method, has been researched and applied abroad. Color-based template matching algorithm is a significant branch of those methods, we can not effectively target tracking weakness against its changes and track targets or dramatic changes in the gray light intensity in the case of tracking the target size changes evident in the case, focusing on studied adaptive mean shift tracking algorithm and hash-based enhanced tracking algorithm fully and effectively solves both problems.Adaptive mean-shift tracking algorithm,once you have determined it is unable to adaptive changes in real time and make the target size for the mean-shift algorithm search window change, resulting in moving objects or background noise outside the window into the window range, reducing the consequences of accuracy, the introduction of the image geometric moments concept. By tracking the target of zero-order moment, the first moment and the second moment for processing and calculation to obtain the search window size and location of the target size of the transformed and updated bandwidth vector kernel function, so this algorithm can adaptively update search window’s size, its background environment adaptability, feasibility, stability improved significantly.The template matching method based on hash boost. The orthodox template matching algorithm, for instance, sum of squared difference algorithm and normalized cross correlation algorithm, is totally hinged on images’gray-scale, the image gray dramatic changes or changes in light intensity under circumstances likely to be significantly affected the image of the high-frequency information, leading to the phenomenon of migration tracking accuracy to reduce puzzle, this paper proposes a tracking template matching hash enhancement algorithms, the specific process is:Get tracking targets low-frequency information to obtain perception hash feature, reinforced structure contour tracking target, and the extracted feature is binarized tracking target generated hash sequence; after matching hash sequence of the target track as a template, by comparing Hamming distance in each frame to find the most similar to the target in order to achieve the effect of tracking; using drawer principle for reducing cost time to compare Hamming distance, speed matching speed; template will update after matching complete, to guarantee the tracking’s continuity. Enhanced tracking algorithm based on a hash matching accuracy remains high, tracking speed is faster than need, in environment of real-time target gray-scale or brightness changes.