基于灰色自适应小目标检测汇编.docVIP

基于灰色自适应小目标检测 吴国平1吴亦奇2 裘咏霄1 杜志顺1 1) 中国地质大学（武汉）机械与电子信息学院，武汉，430074 2) 中国地质大学（武汉）计算机科学与技术学院，武汉，430074 摘要 目标检测是图像处理领域非常关键的一个环节，目标检测效果直接影响后续处理的成败。本文以实际拍摄的复合板图像为研究对象，以灰色系统理论为基础，构建图像处理全零灰色关联模型，对图像进行处理，压制图像背景干扰，强化图像小目标区域信息。用提取的灰关联图像信息，研究灰关联灰度直方图特性，构建差分方程模型，分析其差分灰关联度特征曲线，实现阈值的动态自适应分割，实现复合板小目标缺陷的检测。采用本算法，对相同相近噪声背景下同类130帧单缺陷或多缺陷（最多四个）复合板图像进行了缺陷检测，完全正确检测126帧，有漏检缺陷的图像1帧，误检3帧，正确率为96.9%。 关键词 复合板缺陷，图像小目标检测，灰色关联，自适应，分割阈值 Adaptive Detection for Small Target Based on Grey System Theory Wu Guoping1, Du Zhishun,Wu Yiqi2, Qiu Yongxiao1 1) Faculty of Mechanical Electronic Information ,China University of Geosciences ， Wuhan 430074 (E-mail: wugoping@yahoo.cn) 2)）Faculty of Science and Technology ,China University of Geosciences , Wuhan 430074 Abstract—Target detection is a crucial procedure in the area of image processing and has a direct effect on the subsequent processing. This paper, taking the actual images of composition plates as research objects, based on grey system theory, composes image processing grey relation model to process images, suppress image background interference and amplify small target area information of images. Using the extracted grey relational image information can study the properties of grey relational gray-scale histogram, compose equation model, analyze the characteristic curves of the difference grey relational grade, implement the dynamic adaptive of the and detect the small target flaws of composite plates. By using this algorithm, after detecting flaws of 130 frames of composite plates with one or more flaws (no more than 5) under the same or similar noise background, the result shows that 126 frames of those are completely right, one frame is omitted and 3 frames are falsely detected. The accuracy rate is 96.9%. Key word— composite plate flaw, small target detection of image, grey relation, Adaptive, threshold segmentation 1.引言 复合板表面的缺陷检测问题属于小目标检测问题，小目标检测是图像领域重要的研究方向，是目标检测领域研究的一个热点，也是难点。小目标由于其所占像素点少，容易受干扰，要找到一种有效而适应能力强的检测方法十分困难。对目标进