CAJ | 학술논문

运动目标检测是运动行为理解的前提，也是安防系统研究的热点、难点问题。在分析现有检测算法的基础上，针对背景更新模型不准确、分割阈值难以选取等问题，提出了一套自适应背景差分运动目标检测算法。算法包括：基于像素相关区域灰度曲率特征的背景更新模型，基于直方图统计的动态阈值，改进型区域生长的运动目标标识。实验表明该算法能较好解决光照变化所引起的背景更新以及不同环境下阈值选取等问题。
운동목표검측시운동행위리해적전제，야시안방계통연구적열점、난점문제。재분석현유검측산법적기출상，침대배경경신모형불준학、분할역치난이선취등문제，제출료일투자괄응배경차분운동목표검측산법。산법포괄：기우상소상관구역회도곡솔특정적배경경신모형，기우직방도통계적동태역치，개진형구역생장적운동목표표식。실험표명해산법능교호해결광조변화소인기적배경경신이급불동배경하역치선취등문제。
Moving target detection is the basis of the analysis of object behavior. It is currently a hot and difficult problem in security monitoring system. Several common algorithms were analyzed including ad- vantages and disadvantages, an adaptive moving object detection algorithm based on background differ- ence was proposed to solve the traditional algorithm of which the background updating model is not accu- rate and the threshold value is difficult to select. Key algorithms include： the background updating model based on the feature of gray-scale curvature in the interrelated region, the dynamic threshold based on the histogram statistics under the Gaussian noise model, and the improved target identification algorithm based on region growing. Experimental results show that the algorithm can overcome background updating caused by illumination change and select threshold in different environments.