A Hybrid Cracked Tiers Detection System Based on Adaptive Correlation Features Selection and Deep Belief Neural Networks

Tire defects are crucial for safe driving. Specialized experts or expensive tools such as stereo depth cameras and gages usually used to investigate these defects. In image processing, feature extraction, reduction, classification presented three challenging symmetric ways affect the performance of machine learning models. This paper proposes a hybrid system cracked tire detection based on adaptive selection correlation features deep belief neural networks. The proposed has steps: selection, classification. First, oriented gradient histogram extracts from images. Second, selects important with threshold value adapted nature last step is predict category networks technique. model tested evaluated using real images normal tires. experimental results show that solution performs better than current studies in effectively classifying defect Deep Belief Neural Networks’ provided accuracy (88.90%) Networks (81.6%) Convolution (85.59%).