摘要
针对一类面向家庭的目标监控机器人,利用改进的"Zhang"的标定方法实现对可移动旋转双目摄像头的内、外参数的估算,通过连续拍摄的图片获取视差图,利用前景检测获取前景目标,再运用直方图定位标记出移动目标物体。由于在家庭环境中,移动目标物体往往是待监测对象,辨识对象需要对包含目标物的图片进行人脸检测与识别。通过改进的局部二值模式算子与权重设计标记出脸部的轮廓,并检测相似度,实现对监测对象的归档管理与统计。实验测试结果表明,估算的外参数误差较小,移动目标物体在部分遮挡时亦可被检测出,人脸检测匹配的成功率较高。
For a type of home target monitoring robot,an improved "Zhang" calibration method is used to estimate the internal and external parameters of the removable and rotated binocular camera. The disparity map is obtained by means of the images taken successively. The foreground detection is used to acquire the foreground object,and then the moving target object is located and marked with histogram. In the home environment,since the moving target object often tends to the object under monitoring,it is necessary for identification object to detect and recognize the human face in the image containing the target object. With the LBP(local binary pattern)operator and weight design,the facial contour is marked and the face similarity is detected to realize the file management and statistics of the monitored object. The experimental results show that the estimated error of the external parameters is small,the moving target object can be detected in the condition of partial occlusion,and the success rate of face detection and matching is high.
出处
《现代电子技术》
北大核心
2017年第23期51-54,58,共5页
Modern Electronics Technique
基金
国家级大学生创新创业训练计划项目(201513631031)
