复杂背景下的视频前景检测方法研究

Research on Video Foreground Detection in Complicated Background

采用计算机视觉领域的相关算法，研究了如何有效地从监控视频中提取前景目标的问题．在应用中先将视频转换成逐帧图片并依据图片每一像素的灰度值将图形以二维矩阵形式数值化．在此基础上，采用帧差法、背景差法、混合高斯模型、Harris算法与FREAK算法等多种算法，利用Matlab和Python等工具建立不同背景下的相应模型．成功地从包含动态背景、镜头晃动等不同情境的监控视频中检测并提取出前景目标．同时依据不同前景目标的各维度特征信息，利用主成分分析、k-means聚类等机器学习方法构建了判断异常事件发生的算法模型，取得了一定效果.

In this paper, the algorithm in the field of computer vision is used to study how to effectively extract the foreground object from the surveillance video. We first convert the video into frame-by-frame pictures and digitize the graphics in a two-dimensional matrix based on the grayscale values of the pixels in the picture. On the basis, we use a variety of algorithms such as frame difference method, background subtraction method, Gaussian mixture model, Harris algorithm and FREAK algorithm, and use Matlab and Python to establish corresponding models under different backgrounds. We succeeded in detecting and extracting the foreground object from the monitoring video with different backgrounds such as dynamic background and lens shake. At the same time, an algorithm model for judging the occurrence of anomalous events is constructed based on the features of each dimension of different foreground targets by using principal other machine learning methods, and achieved component analysis, k-means clustering and some results.