Design of A Small-sample Industrial Defect Detection System Based on Feature Enhancement

Under the background of the rapid development of globalization and automation, the industrial manufacturing industry is faced with increasingly severe product quality challenges, especially in the industrial defect detection under small sample conditions. This paper presents an industrial defect detection system method for small samples based on feature enhancement, aiming to improve the utilization rate of small sample data by combining various data enhancement techniques. Firstly, we systematically designed a set of small sample data enhancement scheme, using [1] rotation, translation and flip to enrich the training data set; secondly, we constructed a defect detection model based on convolutional neural network (CNN), which is to improve the learning ability of small sample data; finally, the results of the effectiveness of the model reached 92.5%, 90.0% and 91.2% in accuracy, recall rate and F1-score. This study provides new ideas in the field of industrial defect detection and lays a foundation for the development of small-sample learning in industrial applications. Future studies will focus on model optimization, dataset expansion, and its stability and reliability assessment under dynamic production conditions.