视频图像运动目标鬼影校正算法研究

Research on Moving Object Ghost Correction Algorithm in Video Image

运动目标检测准确度低,完整性差,且易受鬼影干扰。为解决上述问题并提高运动目标检测的背景适应度,将自适应VIBE算法与HOG匹配鬼影相结合,提出一种运动目标校正算法,即HAVB算法。算法首先采用采用ADVIBE算法对预处理后的图像分离目标背景,通过自动更新匹配与距离阈值,提高背景动态适应性;然后采用HOG计算匹配与素值置零的方式,检测并抑制鬼影影响,提升目标检测的完整度;接着通过提取阴影概率分布逐点匹配算法,检测并消除阴影区域,同时基于形态学操作与阈值降噪抑制误差噪声提高目标检测的准确性。不同图像校正基线算法仿真对比实验结果表明,在AEID健美操图像数据集上,较传统算法相比,提出的HAVB算法校正后的图像目标检测准确度高,稳定性强,且具有较高的背景动态适应性:P、R和F指标参数平均提高了6.6%、6.1%和6.2%,背景提取时间平均降低了47.9%。综上所述,构建的HAVB图像鬼影校正算法模型有效的提高目标检测的完整度与准确率,为目标识别与跟踪打下良好基础,具有重要的研究意义。

Motion target detection suffers from low accuracy,poor integrity,and is prone to ghosting interference.In order to solve these problems and improve the background fitness of moving target detection,this paper combines the adaptive VIBE algorithm with HOG matching ghost,and proposes a moving target correction algorithm,namely the HAVB algorithm.Firstly,the ADVIBE algorithm is used to separate the target background from the preprocessed image,and the background dynamic adaptability is improved by automatically updating the matching and distance thresholds;then,the HOG algorithm is used to calculate the matching and set the value to zero,so as to detect and suppress the ghost influence and improve the integrity of target detection.Then the shadow region is detected and eliminated by extracting the shadow probability distribution point by point matching algorithm,and the error noise is suppressed based on morphological operation and threshold noise reduction to improve the accuracy of target detection.The experimental results of different image correction baseline algorithms show that,on the AEID aerobics image data set,compared with the traditional algorithm,the proposed HAVB algorithm has high accuracy,strong stability and high background dynamic adaptability:the P,R and F index parameters are increased by 6.6%,6.1% and 6.2% on average.The background extraction time is reduced by 47.9% on average.In summary,the HAVB image ghost correction algorithm model constructed in this paper effectively improves the integrity and accuracy of target detection,and lays a good foundation for target recognition and tracking,which has important research significance.