In this study, it was analyzed whether the brightness levels of object images taken with industrial cameras affect the image recognition with the function developed in marble classification. Within the scope of this research, studies have been performed on the success rate of the classification process of marble plates extracted from underground in different regions of Turkey. Marble plates are used as a reference for the object image to be processed and the related experiments are performed. The effects of the camera and the light source factors on the images are investigated. The effects of camera and light source factors on the images have been examined. It calculates the period when the camera and brightness factors are active by changing the colour of the graphs and the brightness values. In the graphs examined, it is observed that image colour and brightness values deteriorate similarly after a certain period of time. Brightness values are tried to be normalized by the methods that are already known in the literature, but the desired results are not obtained. Having balanced the average graph with the experimental graph curve the improvement factor was obtained which would be corrected after a certain period of time. A model is calculated to normalize the brightness of images with an improvement factor. Image enhancement is provided by applying the function on the images. Fisher's feature extraction method is used for the extraction of object features. In order to make comparisons reliable in the same setup, Bayes Classification and Stimulating Artificial Neural Fuzzy Classifier (ANFIS) is used. In previous studies, with the help of applied normalization and classification methods, the overall classification success rate of marble surfaces was 98.2%. In this proposed study, the same classification methods were used to make an accurate comparison with the developed standardization method, with an overall classification success rate of 99.2%. That is, the proposed standardization method can be used to achieve further success rates in different classification studies as well as more successful classifications in marble surfaces.
Anahtar Kelimeler: Marble classification, image recognition, light filter development.
|
