CAJ | 학술논문

针对传统的混合高斯背景建模算法实时性差及对光照等引起的环境突变适应性差的问题，提出一种改进算法。通过定义模型生存时间、连续匹配次数等参数定量约束模型的生存时间，改进传统算法中固定模型数量的限制，提高背景建模速度；通过定义突变因子、突变持续时间，检测光照突变。仿真实验表明，改进算法实现了模型数量的自适应增减，提高了建模效率，对由传统建模算法实时性差造成的误检及光照突变引起的误检能较好的解决，具有较强的鲁棒性。
침대전통적혼합고사배경건모산법실시성차급대광조등인기적배경돌변괄응성차적문제，제출일충개진산법。통과정의모형생존시간、련속필배차수등삼수정량약속모형적생존시간，개진전통산법중고정모형수량적한제，제고배경건모속도；통과정의돌변인자、돌변지속시간，검측광조돌변。방진실험표명，개진산법실현료모형수량적자괄응증감，제고료건모효솔，대유전통건모산법실시성차조성적오검급광조돌변인기적오검능교호적해결，구유교강적로봉성。
Traditional GMMbackground modelling algorithm is poor in real time property and poor in adaptability on sudden environment changes caused by lighting,etc.In view of these problems,this paper proposes an improved algorithm.It restricts the life-time of model by defining the parameter quantitatives of model life-time and the times of continuous matching,and improves the limit on fixed number of Gauss-ian model in traditional algorithm,and increases the background modelling speed.It detects sudden lighting changes by defining mutation fac-tor and the duration of mutation.Simulation experiment demonstrates that the improved algorithm achieves model number’s self-adaptive change,raises the background modelling efficiency.For the false detections caused by poor real-time performance in traditional modelling al-gorithm and by sudden changes in illumination,it can perfectly solve,and presents better robustness.