Computerized image analysis in manufacturing industry anomaly detection using artificial intelligence techniques

Inspection for defects is crucial to guaranteeing the quality of industrial goods. The employment of an autonomous flaw inspection algorithm in real production is in great demand due to the shortcomings of the most popular human visual inspection approach, which includes high cost and low efficiency. However, due to discrepancy between data acquired in the actual production environment as well as pre-made datasets, only a small number of industrial production lines employ autonomous detection equipment. One industrial product that only rely on hand fault inspection is lace. Using the current image-based defect evaluation techniques, it is challenging to derive regular patterns from lace’s intricate and delicate texture. In order to discover anomalies utilizing computerized image analysis in the industrial business, this research proposes a revolutionary approach. Here, security camera footage from the industrial sector has been gathered and processed for noise reduction, smoothing, and normalization using computerized data. Then, using quantile vector reinforcement structural learning, the characteristics of the processed data are identified and chosen using a Gaussian radial Boltzmann neural network. For the video dataset and computerized data, experimental analysis is conducted in terms of training accuracy, average mean error, anomaly analysis, trust value analysis, and anomaly analysis. The proposed technique attained training accuracy of 98%, average mean error of 69%, anomaly analysis of 99%, and trust value analysis of 98%.