Research On Aluminium Alloy Atmospheric Corrosion Morphology Based On Image Recognition

Corrosion morphology image is one of the most important features to evaluate the material's corrosiveness. In the paper, the digital image processing methods were used in metal corrosion research. Using the modern mathematic theory, combined with the characters of corrosion in the atmosphere environment, the relations between the features of corrosion morphology images and the aluminium alloy samples'corrosion data were studied. The main conclusions are as follow:The corrosion morphology image acquisition systems were established which can be used both in the laboratory and filed. The suitable parameter to capture images and corresponding pre-processing methods were discussed and determined. The results showed that the median filter was an efficient method to eliminate the noise signal that produced during image capturing. While the fuzzy enhancement method was very effective to make the corrosion morphology features showed clearly.The wavelet image analysis method was used to study the corrosion morphology of LC4CS and LY12CZ aluminium alloy samples with period rain accelerated corrosion tests. The improved images were decomposed by muti-resolution wavelet transformation. Every subimages contained different information of a special scale and orientation. The energy values were obtained from each subimage. The canonical correlation analysis was used to gain the correlative coefficient between the energy values and the corrosion loss. The concept of image featureδwas introduced as evaluating indicator. Range analysis and variance analysis were used to study the effects of orthogonal experiment factors on acceleration. And the best combination of factor lever for accelerated corrosion test was also obtained.In the atmospheric corrosion exposure station, the image acquisition systems were established to capture the corrosion morphology of aluminium alloy samples and the samples with coating. The corrosion degree of aluminium alloy samples was qualitatively analyzed. The corrosion images were improved by median filter, and then the images were decomposed by muti-resolution wavelet transformation. The energy values were obtained from each subimage. According to the values, the corrosion degree of the aluminium alloy samples could be qualitatively estimated.