CAJ | 학술논문

对视频帧中运动目标提取是计算机视觉研究的重点课题,对视频帧色差突变图像的背景检测常受到背景色差干扰,目标检测性能不好.提出一种基于视频序列的双背景建模的视频帧色差突变图像的背景检测和参量提取算法,背景建模加入了光照突变处理机制,进行色差补偿.计算视频帧差背景内的突变信息感知概率和统计概率,对视频图像进行滑动平均建模,采用模极大值法进行边缘检测,得到基于灰度方差的视频帧色差稳定性检测的判别函数.将当前帧与背景帧相减,建立滑动平均背景模型,提取其背景差异性特征参量.仿真结果表明,该算法的检测性能较好,当背景发生变化时,如光照突变、人群数目突然增大时,具有较好的背景检测性能,处理光照突变方面的图像平滑性较好.
대시빈정중운동목표제취시계산궤시각연구적중점과제,대시빈정색차돌변도상적배경검측상수도배경색차간우,목표검측성능불호.제출일충기우시빈서렬적쌍배경건모적시빈정색차돌변도상적배경검측화삼량제취산법,배경건모가입료광조돌변처리궤제,진행색차보상.계산시빈정차배경내적돌변신식감지개솔화통계개솔,대시빈도상진행활동평균건모,채용모겁대치법진행변연검측,득도기우회도방차적시빈정색차은정성검측적판별함수.장당전정여배경정상감,건립활동평균배경모형,제취기배경차이성특정삼량.방진결과표명,해산법적검측성능교호,당배경발생변화시,여광조돌변、인군수목돌연증대시,구유교호적배경검측성능,처리광조돌변방면적도상평활성교호.
The motion in a video frame object extraction is a key topic in computer vision research, the background color mutation detection video frame image is often influenced by background color interference, target detection performance is not good. Put forward a kind of background detection and parameter extraction of video frame color double background mod-eling video sequence mutation based on image, background modeling joined the light mutation mechanism, for color com-pensation. Calculation of video frame mutation probability and statistical probability information perception within a back-ground of moving average model, the video image edge detection, using the modulus maxima method, get the discriminant function of video frame color stability detection based on the gray variance. The current frame and the background frame subtraction, the establishment of the sliding average background model, extracting the characteristic parameters of back-ground difference. The simulation results show that the detection performance of this algorithm is better, when the back-ground changes, such as light mutation, population increases suddenly, with a background in better detection performance, processing lighting change aspects of image smoothing is better.