Research on Captcha Recognition Based on TensorFlow and CNN Model
Aiming at the problems of low overall accuracy and insufficient generalization ability of segmentation and recognition methods applied to multi character captcha in traditional Machine Learning,an efficient and universal recognition method is proposed.It designs an end-to-end character captcha recognition process based on CNN model,and uses TensorFlow framework to implement data training and effectiveness verification of the process.This method can efficiently recognize character captcha with an average accuracy of over 95%.By inputting the entire image and directly outputting the overall recognition result,it has stronger universality.It uses CNN models to recognize multi character captcha has higher accuracy and versatility compared to traditional Machine Learning methods.
国家科技期刊平台
NSTL
万方数据
