摘要
提出并分析了目标提取的稳定性问题，指出基于统计决策理论的目标提取算法性能不稳定的原因之一，在于缺乏时空可变的目标背景条件判别和最佳分割区域自调整能力，并且通常是基于绝对图像灰度或绝对灰度变化及其统计性质的。根据视觉原理，还提出了一个具有可控性的自适应序列图像目标提取模型和算法，给出了复杂背景中运动目标提取和跟踪的实验结果。
The stability of object extraction is analyzed.The reason why the performance of somealgorithms based on the statistical deckion theory appears to be unstable is because of(1)lack of both the assessment of object/background condition and the self-adjustment capabil-ty of optimal segmentation regions with varying time and space;(2)dependence of the algo-rithm on the absolute image grey level or its variation as well as their statistical properties。Based on the visual perception principle,a controllable model and an algorithm for adaptiveextraction of objects of interest in image sequence are proposed.Experimental results of ex-traction and tracking of moving objects in a complex background are given.
出处
《华中理工大学学报》
CSCD
北大核心
1994年第5期55-60,共6页
Journal of Huazhong University of Science and Technology
基金
国家自然科学基金
中国科学院机器人学实验室基金
关键词
机器视觉
图像分割
目标提取
跟踪
machine vision
image segmentation
object extraction
tracking
system sta-bility
