Aided Application of Neural Network-Fuzzy in Human Vision Refraction Imaging

With the rapid advancement of auto-refractor technology, refraction services have become readily available in most optometry stores. While auto-refractors offer speed and convenience, the measurements they provide are subject to a notable margin of error influenced by psychological and physical factors. Consequently, repetitive testing is required to obtain more accurate mean error values. Therefore, it is crucial to explore research on measurement methods that can complement auto-refractor measurements and validate refraction results. This will help ensure more reliable and precise outcomes in the field of refraction. The Liou eye model is employed to simulate the human eye and facilitate the development of optometric techniques. In this study, CODE V optical design software was utilized to calculate the visual imaging generated by personalized, individual human eye models. Subsequently, an innovative algorithm combining fuzzy logic and backpropagation neural networks was proposed for category recognition. The experimental results using this new approach achieved a recognition accuracy rate exceeding 98%, thereby enhancing the precision of optometry and enabling dual confirmation through computer-aided optometry.