SHEET RUBBER QUALITY CLASSIFICATION USING IMAGE PROCESSING AND THE K-MEANS ALGORITHM

Abstract

The research on sheet rubber quality classification using image processing and K-Means algorithm was aimed to 1) study about the algorithm of classifying quality of sheet rubber using  image-processing method 2) classify the quality of sheet rubber by using image-processing method and 3) determine the quality classification efficiency of sheet rubber by image processing method. The research methods were as follows: The first step was data collection, the second step was the analysis of need and processes of system, the third step was to design algorithm for quality classification of sheet rubber using image processing, the fourth step was the system development, the fifth step was about testing the system to correct any defects, the sixth step related to evaluate the working performance and the seventh step related to the summary of research results. Statistics for data analysis were average and percentage.

            The research results were revealed that 1) the algorithm for classifying the quality of sheet rubber were as follows: 1.1) To design a method of photographing sheet rubber comprised 3 formats including control light format, room light, and outdoors. 1.2) Pre-processing images – images were cropped for creating dataset employing 150 cropped images of both raw and smoked rubber sheet in form of light patterns with size 100 x 100 pixels. But 100 images were used in the clustering process and only 50 images were used to create test data per type
1.3) To analyze the color brightness level employing clustering with K-Means algorithm. 2) The quality of sheet rubber classified by image processing has been divided into 2 groups: The first group was normal sheet rubber and the moldy sheet rubber. Overall, the evaluation of accuracy was 96.57 percent, of precision was 99.95 percent, of recall was 93.19 percent as well as of F1-Score was 96.45 percent, respectively.