摘要
为了提高实时行人检测结果的准确性,避免出现较大的漏检率,分析了现阶段使用较广泛的实时行人检测算法:经典AdaBoost算法和非对称的GentleBoost算法。通过对比分析发现,非对称的GentleBoost算法能够针对实时行人检测筛选出特定的区域,降低训练样本的总数。同时,经过与传统的行人检测算法进行对比试验发现,所提出的算法能够在复杂的行人背景下进行分析,提高了实时行人检测的准确度和降低漏检率。
In order to improve the accuracy of real-time pedestrian detection results and avoid large missed detection rates,this paper analyzes the widely used real-time pedestrian detection algorithms at this stage:the classic AdaBoost algorithm and the asymmetric GentleBoost algorithm.Through comparative analysis,it is found that the asymmetric GentleBoost algorithm can screen specific areas for real-time pedestrian detection and reduce the total number of training samples.At the same time,through comparative experiments with traditional pedestrian detection algorithms,it is found that the algorithm proposed in this paper can analyze in the background of complex pedestrians,improving the accuracy and reducing missed detection rate of real-time pedestrian detection.
作者
张明
ZHANG Ming(Henan Vocational College of Water Conservancy and Environment,Zhengzhou 450000,China)
出处
《电子设计工程》
2021年第2期133-136,共4页
Electronic Design Engineering
