earticle

영어

In strawberry cultivation, maturity classification plays an important role in ensuring the efficiency and quality of harvesting. In this study, we propose an Improved Faster R-CNN model to address these challenges, using MobileNetV3-Large as the backbone network to achieve a lightweight model, and introducing RoI Align to improve the spatial accuracy of the feature map. Experiments are conducted using the KGCV_Strawberry dataset, with precision, recall, F1 score, and mean average precision (mAP) measured for performance evaluation. The experimental results show that the proposed model achieves an average precision of 71.35%, recall of 71.07%, and F1 score of 71.21% across all classes. In particular, the proposed model achieves 63% performance on mAP0.5 and 58% performance on mAP0.5:0.95, which is comparable to existing ResNet-based models while achieving faster inference speed. The proposed model achieves a processing speed of 27.6543 ms, which is about 2 ms faster than existing ResNet-based models. This indicates that the goal of creating a lightweight model with improved image processing capability was achieved with minimal performance degradation. This research is expected to contribute to the development of automated strawberry cultivation systems in greenhouse environments and has the potential to be applied to various agricultural environments in the future.