The appearance defects of precision hardware processing parts not only affect the appearance, but also due to the lack of precise coordination between parts during installation, causing damage to parts. At present, the quality testing of precision hardware processing devices has seriously lagged behind industry development and market demand. Due to the small size of the workpiece and the complexity of the detection surface, the detection accuracy is required. Most of them still use manual visual inspection methods to detect unstable quality, low detection efficiency, data cannot be traced, and cannot ensure the quality of batch production. Analyze the appearance defect type of precision hardware, study the testing technology of precision hardware, and realize the automatic intelligent testing of precision hardware. It is a problem that enterprises urgently need to solve. For precision hardware treatment, the cost of customization is high, the detection efficiency is low, and the test results cannot be tracked. This article designs a precision part -cylinder surface defect detection system based on convolutional neural networks. System requirements mainly include the following functions:

① High -quality image acquisition function: Design machine visual lighting mechanical system, reasonably select light sources and cameras, improve the quality of collection image, and lay the foundation for subsequent defect detection.

② Defective positioning function recognition and positioning of precision hardware processing parts of the cylindrical surface appearance defect, and indicate its position range in the image.

③ Dead type recognition function recognition precision hardware that the appearance defect type at different positions on the surface of the cylinder, and improves the efficiency of inspection workpieces defects.

④ The automatic detection function realizes the automatic detection of cylindrical surface defects during precision hardware processing, effectively reduces labor costs and improves the detection efficiency of workpiece defects.