Abstract
The system and performance of the developed data processing unit that detects and classifies the fabric defects as compared with the corresponding data of the normal region by using the statistical processing are studied.
The intensity distribution are sampled from the electro-optical output by scanning and the mean value of integration in each scanning times expressing the characters of the fabric without defects are obtained by the pre-processing part. Then the several parameters needed for detecting the defects are calculated according to these data by the digital computer.
The threshold values to separate the abnormal regions from normal are decided by way of K times of the standard deviation obtained from the histograms of the electro-optical output by the reflection or transparency from fabrics, also the values of K are judged by many experimental results.
The defect informations expressing the existing positions and sizes according to the threshold logic are detected by the analogue comparators and the peak holding circuits. These informations are changed into the binary coded signals by A-D converters and fed to the digital computer.
The digital computer prints out the several data fed from the data processing unit and the results calculated by these data.
