| Abstract [eng] |
In this paper, two methods have been offered to detect defects in texture surfaces. These methods could be practically used for an automated visual inspection and quality control in a process of serial production to avoid the financial problems caused by the selling decrements. The discrete Haar wavelet transform (HT) has been applied in the failure detection algorithms, because Haar wavelet is better then other wavelet basis functions such as Daubechies, Symlets, Coinflets, for the inspection of defects. The wavelet transform decomposition can be performed by passing the digital image through the low- pass and high- pass filters to generate the different frequency components whose spectral coefficients store all information about the original image. Besides, the experiments were carried out to show that at least one of the obtained components consists of enchanced wavelet coefficients in the image with the different kind of failure. This property of discrete wavelet transform gives possibility of the detecting of defects in texture surfaces. The empirical rule and the calculation of the relative differences between the corresponding maxima of wavelet coefficients are used in the detection procedure too. The proposed methods were analyzed using the digital images of the intact and defected surfaces of the paving blocks. The experimental results revealed that all images with every type of manufacturing defects such as cracks, pinholes and spots, were classified correctly, but about 15- 20 percent of the intact images were classified falsely as the defected ones. Sometimes the situation when a part of the undamaged paving blocks are rejected is required because of the perfect quality of the production. The results showed that the proposed methods detect defects successfully, so the obtained information about the inspection of defects in the texture images could be meaningful for the localization procedure. The basics principles about the localization of the defects are described in this paper too. |
