摘要
目标检测问题是计算机视觉领域最普遍和关键的问题之一.基于级联结构的AdaBoost算法目前被认为是较有效的检测算法,但是其在低FRR端的性能仍需改进.文章提出了一种针对目标检测问题的改进AdaBoost算法———AD AdaBoost.AD AdaBoost采用了新的参数求解方法,弱分类器的加权参数不但与错误率有关,还与其对正样本的识别能力有关.该算法能够有效地降低分类器在低FRR端的FAR,使其更适用于目标检测问题.新旧算法在复杂背景中文字检测的实验结果对比证实了新算法在性能上的改进.
Object detection is one of the most popular and important issues in the domain of cornputer vision. AdaBoost algorithm based on cascade structure can solve the problem effectively, however it has its own shortcoming. This paper proposes a revised type of AdaBoost algorithm, AD AdaBoost. AD AdaBoost adopts a new method to acquire parameters. The weighted parameters of weak classifiers are determined not only by the error rates, but also by their abilities to recognize the positive samples. The algorithm can decrease the classifiers＇ false alarm rates in the low false rejection rate end, so it is more adaptive to the object detection based on cascade structure. The experiment results prove the improvement achieved by the new algorithm.
出处
《计算机学报》
EI
CSCD
北大核心
2007年第1期103-109,共7页
Chinese Journal of Computers
基金
国家自然科学基金(60472002)资助.
