Fault Diagnosis of Reliability Test for Low-Light-Level Vision Device Based on Structural Similarity Algorithm

At present, fault diagnosis in direct-view low-light-level photo-optical equipment is mainly conducted by manual detection, which is inefficient, error prone (faults will often be missed), and does not create a valid record. This paper proposes an automatic diagnostic method based on machine vision. In this method, a series of special adaptors are designed for developing a reliable connection between the industrial camera and the object to be tested, automatically collecting the images of the eyepiece field of view image of the tested product, and transmitting the monitoring image in real time. We used the Structural SIMilarity (SSIM) algorithm to calculate the similarity between the monitoring images and the template image in real time to automatically warn of abnormalities using a judgment threshold, which is determined in advance. When a failure occurs, the system issues an abnormal warning, generates a detection log, and stores the current monitoring images. Practice shows consistency of the results of our method with those of subjective judgment. Under stable illumination conditions, the accuracy of the diagnostic technique meets the actual requirements.