基于并行演化计算的图像特征提取研究

Research on Image Feature Extraction Based on Parallel Evolutionary Computation

图像特征提取已成为困扰智能化视觉信息处理的瓶颈。为了解决复杂背景下钢板表面缺陷的机器视觉检测问题,缩短视觉系统的开发周期,提高其实用性,在并行计算环境和视频仿真平台的基础上,探索机器视觉可塑性及形成机制,提出了将并行计算、视频仿真、演化计算相互融合解决钢板表面缺陷特征提取问题的方法,实现了演化计算与并行层次处理的特征选择、特征提取方法,为智能化的视觉信息处理开辟新的思路。实验证明,改进方案不仅具有可行性,而且能提高缺陷检测的准确性、实时性。

Feature extraction has become the bottleneck of intelligent visual information processing. In order to solve the problems of steel surface defects detection and feature extraction under complex image background, shorten the design cycle for vision system, and increase its utility, a parallel computation platform and a video simulation platform were designed. Based on that, a hybrid method integrated with parallel computing, video simulation, and evolutionary computing, was proposed to solve the problem of steel surface defects detection and feature extraction. New ideas were stimulated for intelligent visual information processing. Experiments show that, the method not only is viable, but also can improve the accuracy and immediacy during defects detections.