Automatic object recognition using deep learning for legacy waste treatment

レガシー廃棄物の処理に向けた深層学習モデルを用いた自動物体識別技術

吉田 幸彦

Yoshida, Yukihiko

JAEA is addressing the back-end issues with the steadfast promotion of sustainable measures. The R&D plans are being rationally pursued by considering priorities based on indicators such as bottleneck issues in waste streams, relevance to WAC settings, and effectiveness of cost reduction. In addition, the future vision (JAEA 2050+) has been formulated to promote cross-disciplinary R&D through a new approach that cannot be reached by conventional methods, and active incorporation of information technologies, such as AI technologies. Currently, we are developing intelligent sensing that combines sensing and information processing technologies to realize automatic sorting technology, and non-destructive evaluation technology using high-energy X-ray CT, for legacy wastes. The core technology common to these technologies is image evaluation technology using deep learning models, which was confirmed to perform very well in the evaluation of waste object recognition.