摘要
针对外部环境的复杂多变性导致基于传统多光谱聚合通道特征(MACF)的行人检测算法漏检率高的问题,在提取红外通道梯度方向直方图(HOG)特征时,利用局部区域的邻域像素强度差异估计和信息熵分析,构建新的熵加权强度差异直方图(EWHID)特征,来加强行人目标边缘轮廓部分对整体特征的贡献程度。针对昼夜不同环境条件下多光谱行人特征存在差异导致误检率较高的问题,利用HSV颜色空间直方图信息建立的检测器对输入图像进行昼夜分类,用训练好的日间和夜间Real Adaboost分类器对图像进行昼夜分时检测。仿真实验结果表明:相比于MACF行人检测算法误检窗口减少,检测精度提升。
In order to solve the problem of high missing rate of pedestrian detection algorithm based on traditonal multispectral aggregate channel features(MACF)due to variability and complexity of external environment,when extracting the infrared channel feature histogram of oriented gradient(HOG)feature,use local pixel intensity difference estimation and information entropy analysis,a new entropy weighted histogram of intensity difference(EWHID)feature is constructed to enhance the contribution of the edge contour of pedestrian target to the overall feature.In order to solve the problem of high false detection rate due to the difference of multispectral pedestrian features in day and night time,input image is classified into daytime and nighttime by using the detector established by histogram information of HSV color space,and the image is detected by using the trained Real AdaBoost classifier in daytime and nighttime.Simulation results show that compared with MACF algorithm,the proposed method reduce false detection window and improve detection accuracy.
作者
余子航
于凤芹
YU Zihang;YU Fengqin(School of Internet of Things Engineering,Jiangnan University,Wuxi 214122,China)
出处
《传感器与微系统》
CSCD
北大核心
2022年第3期145-149,共5页
Transducer and Microsystem Technologies
基金
国家自然科学基金资助项目(61573168)
中央高校基本科研业务费专项资金资助项目(JUSRP51733B)。
关键词
行人检测
多光谱聚合通道特征
信息熵
昼夜分时检测
pedestrian detection
multispectral aggregate channel features(MACF)
information entropy
detection in day and night time
