棉花中异性纤维的多光谱检测

为有效检测棉花中与棉纤维形态、色泽极其相似的异性纤维杂质,提出一种利用多波段光谱信息融合成像检测的新方法。针对6种肉眼极难识别的异性纤维:无色塑料、黄麻、编织袋、白头发丝、白羊毛、猪鬃在7个离散波段光谱中的反射特性,获得不同波段光谱图像中异性纤维与棉纤维的图像特征差别,由此确定检测各异性纤维的最佳光谱波段。采用基于区域信息相关度权值小波分析算法将多个波段的图像进行融合,得到具有完整异性纤维特征信息的单幅杂质图像。实验结果表明,在多波段光谱融合图像中,异性纤维灰度、形态特征明显,可有效识别棉花中异性纤维杂质。

Detecting foreign fibers in cotton using a multi-spectral technique

An information fusion method using multi-spectral imaging was developed to detect foreign fibers having the same shape and color as the cotton. Six types of foreign fibers, plastic, jute, knitting, white hair, wool and bristle, which are difficult to identify by humans, were used in the experiments. The tests measured the reflectance characteristics of these foreign fibers at seven discrete wavelengths. The image features of the cotton fibers and the foreign fibers were then analyzed quantitatively to determine the optimum wavelength for detecting each type of foreign fiber. A single image with information about all the foreign fibers was prepared using an image fusion algorithm called partial correlation weight selected wavelet analysis. The tests show that the gray levels and shape features of the foreign fibers in the combined image using the multi-spectral information were obvious. This method provides an novel and effective way to identify foreign fibers in cotton.