中圖分類號：TP202 文獻(xiàn)標(biāo)志碼：A DOI: 10.16157/j.issn.0258-7998.245833 中文引用格式： 馬方遠(yuǎn)，任杰夫，黃靜，等. 基于機(jī)器視覺的監(jiān)控視頻移動(dòng)目標(biāo)輪廓提取算法[J]. 電子技術(shù)應(yīng)用，2025，51(7)：78-82. 英文引用格式： Ma Fangyuan，Ren Jiefu，Huang Jing，et al. A machine vision based algorithm for extracting the contour of moving targets in surveillance videos[J]. Application of Electronic Technique，2025，51(7)：78-82.
A machine vision based algorithm for extracting the contour of moving targets in surveillance videos
Ma Fangyuan，Ren Jiefu，Huang Jing，Zhang Zhengchu
Beijing Guodiantong Network Technology Co.，Ltd.
Abstract： In surveillance videos, moving targets are easily affected by edge blurring and background noise interference, resulting in inaccurate contour extraction results. Therefore, a machine vision based algorithm for extracting the contour of moving targets in surveillance videos is proposed. Using machine vision technology to process images, combining with the image background model to establish pixel Gaussian distribution, and calculating the matching degree between each pixel value and the Gaussian distribution in the background model, pixels with matching degree below the threshold are regarded as foreground pixels, thus completing foreground segmentation. Combining chain code to encode the foreground target edge, identify the closed edge closest to the preset threshold, and use it as the rough positioning result of the target contour. Based on this, the statistical distribution parameters of the target area are fused into the grayscale optical flow image to obtain a grayscale differential image, and then extract the contour of the moving target. The experimental results show that under the application of the proposed method, the number of burrs on the target contour is always controlled below 50, and the contour extraction accuracy is high.
Key words : machine vision technology；surveillance video；moving targets；contour extraction；background model
