基于概率分析的稀疏表示算法及在布料印染检测中的应用

为解决布料印染过程中，由于布料的褶皱或机器的故障而导致走布出现跑偏现象，本文提出了基于概率分析的稀疏表示分类算法(P-SRC)，将布料的检测问题转换为稀疏表示的图像分类问题；同时提出了自适应纹理窗口选择算法和自适应样本个数计算方法。先把初始无偏转的图像，通过旋转产生新的图像，再利用所有图像作为训练样本构造过完备字典，对采集到的测试图像，按照本文提出的算法进行稀疏表示并分类识别。根据测试样本所属的类可判断染布机在走布过程中布料是否有偏转以及偏转的角度。实验表明，该算法提高了图像的识别率。

Sparse representation algorithm based on probability analysis andapplication in fabric printing and dyeing inspection

In order to solve the phenomenon of cloth running deviation caused by cloth wrinkles or machine failure in the process of cloth dyeing and printing, this paper proposed the sparse representation algorithm based on probability analysis (P-SRC), transforming the cloth detecting problem into the sparse representation image classification problem. At the same time, an adaptive texture window selection algorithm and adaptive sample quantities calculating method are proposed. Rotate the original non-deflected images to generate new images, then use all the images as training samples to build over complete dictionary, to sparse represent and classify, identificate the collected testing samples according to the algorithm this paper proposed. Judging by the class that the test sample belongs to , whether the cloth deflection and the deflection angle during the dyeing machine working process can be detected. Experiment result shows that this algorithm improves the recognition rate of the image.