This study presents an improved CES-YOLOv8 network for strawberry maturity detection, enhancing the accuracy and robustness of automated harvesting robots. The research addresses the challenges of detecting strawberries in complex environments, such as varying lighting conditions, occlusions, and angles. The improved model incorporates the ConvNeXt V2 module to enhance feature capture and the ECA attention mechanism to improve feature representation. Additionally, the SIoU loss function is used to enhance the Intersection over Union (IoU) calculation, improving the localization accuracy of the model's predictions. Experimental results show that the improved CES-YOLOv8 model achieves an accuracy of 88.20%, a recall rate of 89.80%, an mAP50 of 92.10%, and an F1 score of 88.99%, representing improvements of 4.8%, 2.9%, 2.05%, and 3.88% respectively, compared to the original YOLOv8 network. The model's performance is validated through various metrics and compared with other target detection networks, demonstrating its superior accuracy and efficiency in detecting strawberry ripeness. The study concludes that the improved CES-YOLOv8 model provides a technical support for efficient and precise automated harvesting, contributing to the advancement of smart agriculture and sustainable agricultural development.This study presents an improved CES-YOLOv8 network for strawberry maturity detection, enhancing the accuracy and robustness of automated harvesting robots. The research addresses the challenges of detecting strawberries in complex environments, such as varying lighting conditions, occlusions, and angles. The improved model incorporates the ConvNeXt V2 module to enhance feature capture and the ECA attention mechanism to improve feature representation. Additionally, the SIoU loss function is used to enhance the Intersection over Union (IoU) calculation, improving the localization accuracy of the model's predictions. Experimental results show that the improved CES-YOLOv8 model achieves an accuracy of 88.20%, a recall rate of 89.80%, an mAP50 of 92.10%, and an F1 score of 88.99%, representing improvements of 4.8%, 2.9%, 2.05%, and 3.88% respectively, compared to the original YOLOv8 network. The model's performance is validated through various metrics and compared with other target detection networks, demonstrating its superior accuracy and efficiency in detecting strawberry ripeness. The study concludes that the improved CES-YOLOv8 model provides a technical support for efficient and precise automated harvesting, contributing to the advancement of smart agriculture and sustainable agricultural development.