IMPROVEMENT OF STIV TECHNIQUE BY USING DEEP LEARNING

Abstract

 With the flood disasters that have been occurring frequently in recent years, the sophistication of discharge observation has been promoted. Since the existing river monitoring cameras can be used, interest in image measurement methods is increasing. Among the image measurement methods, the STIV method is regarded as a powerful method because of its high measurement accuracy and robustness of measurement. However, depending on the shooting conditions such as rough weather, there are cases in which the conventional automatic analysis method produces an erroneous value, which is a challenge for establishing a real-time measurement system. Therefore, in this study, we tried to solve these problems by incorporating the deep learning method, which has been successful in the field of image analysis in recent years, into the STIV method. As a result, it was clarified that STIV method using deep learning is able to obtain correct results in many cases where the conventional method outputs anomalous values.