模板自适应的积分通道特征行人检测算法

Template Adaptively Integral Channel Feature Algorithm for Pedestrian Detection

本文对视频中的行人检测问题进行了研究。针对传统的积分通道特征算法在误测背景目标和丢失目标两方面存在缺陷,本文以积分通道特征的行人检测算法为核心,对原检测算法进行了优化并提出模板自适应的积分通道特征算法。通过使用预处理优化和自适应模板,该优化算法能够有效地检测到丢失的目标并且抑制了背景干扰,提高了检测的准确性。最后的实验数据表明该优化算法整体性能有所提高,尤其在提高检测丢失目标的准确率和降低误测背景目标的检测率上都相较于传统的积分通道特征算法有较大的提高。

This paper focuses on the pedestrian detection in video. With regard to the shortcomings of the traditional inte- gral channel feature algorithm, such as, the error detection of background targets and lost targets, the original detection al- gorithm is optimized and the template adaptive integral channel feature algorithm is proposed in this paper based on the pe- destrian detection algorithm of integral channel feature. This optimized algorithm can effectively detect the lost target and suppress the background interference to improve detection accuracy by applying the preprocessing optimization and adaptive template. The experiment results suggests that the overall performance of the optimized algorithm can be improved, espe- cially in improving the accuracy of detecting lost targets and reduce the error detection of background targets comparing with traditional integral channel feature algorithm.