视频星运动目标检测方法的对比和分析

视频卫星可提供大尺度运动场景的实时监控,而运动目标检测是其应用的核心问题。本文设计并实现了目前3种主流的运动目标检测方法,以路面交通监控视频和卫星视频为应用背景,详细探讨了3种方法在检测精度、实时性和运动目标完整性等方面的优缺点和适用条件。实验表明:混合高斯模型方法在多目标复杂背景下,表现出较强的适应性。粒子滤波法具有抗灰尘噪声干扰特性,运动目标边缘保持较好。

Comparison and Analysis of Moving Target Detection Methods for Video Star Data

Video satellites provide real-time monitoring of large-scale motion scenarios, and moving objects detection is a central issue in its application. This article designs and realizes the three main methods of moving target detection. The performance and applicable conditions are discussed in detail in these aspects such as the detection accuracy, real-time and the integrity of moving target and so on, with pavement traffic monitoring video and satellite video as the application background. The experimental results show that the Gaussian mixture model method puts up strong adaptability under the multi-objective and complex background, and Particle filter method has anti-dust noise interference characteristics and keep the edge better.