摘要
以棉花中羊毛、白头发、塑料膜等杂质的检测为应用背景,提出一种利用多颜色空间特征融合方法。该方法构建了颜色特征评价函数,对近似目标在不同颜色空间的特征表现进行评估,从中抽取近似目标的若干最优特征;再利用区域信息相关度权值小波分析算法进行多特征融合,获取近似目标的图像。实验结果表明,融合图像比原始图像及单色空间图像具有较高的图像信息量值,近似目标视觉特征明显增强。此方法为提取与背景特征相近的近似目标提供一条新思路。
The detection method was developed for such low-contrast targets using multiple color space information fusion for detection of foreign fibers such as wool, white hair, or plastic in cotton. The method analyzes the different color features of the targets in different color spaces to seleet several optimal features using a constructed color feature evaluation function. After the different feature images are processed according to their characteristics using various image processing methods, the multiple feature images are fused using wavelet transforms to get a single image with clearly defined features. The enhanced features enable improved discrimination between the targets and the background than in the original image. Similar targets can then be separated from the background using image segmentation techniques. Therefore, the method provides an improved method for detection of similar targets that tend to blend with the background.
出处
《清华大学学报(自然科学版)》
EI
CAS
CSCD
北大核心
2006年第2期176-179,共4页
Journal of Tsinghua University(Science and Technology)
关键词
近似目标检测
多颜色空间
特征评估
特征融合
similar target identification
multiple color spaces
feature evaluation
features fusion