Deep reinforcement learning enables adaptive-image augmentation for automated optical inspection of plant rust

Front Plant Sci. 2023 Jul 7:14:1142957. doi: 10.3389/fpls.2023.1142957. eCollection 2023.

Abstract

This study proposes an adaptive image augmentation scheme using deep reinforcement learning (DRL) to improve the performance of a deep learning-based automated optical inspection system. The study addresses the challenge of inconsistency in the performance of single image augmentation methods. It introduces a DRL algorithm, DQN, to select the most suitable augmentation method for each image. The proposed approach extracts geometric and pixel indicators to form states, and uses DeepLab-v3+ model to verify the augmented images and generate rewards. Image augmentation methods are treated as actions, and the DQN algorithm selects the best methods based on the images and segmentation model. The study demonstrates that the proposed framework outperforms any single image augmentation method and achieves better segmentation performance than other semantic segmentation models. The framework has practical implications for developing more accurate and robust automated optical inspection systems, critical for ensuring product quality in various industries. Future research can explore the generalizability and scalability of the proposed framework to other domains and applications. The code for this application is uploaded at https://github.com/lynnkobe/Adaptive-Image-Augmentation.git.

Keywords: adaptive image augmentation; automated optical inspection; deep Q-learning; deep reinforcement learning; semantic segmentation.

Grants and funding

This work was supported by the Guangzhou Government Project (Grant No. 62216235), the National Natural Science Foundation of China (Grant No. 622260-1), the Natural Science Foundation of Guangdong Province (Grant Nos. 2022A1515240061 and 2023A1515012975), the Foshan Industrial Field Science and Technology Research Project (Grant No. 2020001006827), and the Characteristic and Innovative Project for Guangdong Regular Universities (Grant No. 2021KTSCX005). The authors thank the Researchers Supporting Project Number (RSP2023R35), King Saud University, Riyadh, Saudi Arabia, and Hainan Provincial Natural Science Foundation of China (No. 123QN182) and Tianjin University-Hainan University Independent Innovation Fund Cooperation Project (NO. HDTD202301, 2023XSU-0035).