This study proposes an improved CES-YOLOv8 model for strawberry maturity detection, enhancing the accuracy and efficiency of automated harvesting robots. The model is based on the YOLOv8 framework, with modifications to the backbone network, attention mechanism, and loss function. The backbone network replaces some C2f modules with ConvNeXt V2 to improve feature extraction, while the ECA attention mechanism enhances feature representation. The SIoU loss function is used to improve the intersection over union (IoU) and enable faster localization of prediction boxes. The model was trained on a dataset of 546 strawberry images collected under various lighting conditions, occlusion levels, and angles. Data augmentation techniques were applied to increase the dataset size to 2722 images. The model achieved an accuracy of 88.20%, recall of 89.80%, mAP50 of 92.10%, and F1 score of 88.99%, representing improvements over the original YOLOv8 network. The model also maintains a high frame rate of 184.52 FPS, ensuring real-time performance. The improved CES-YOLOv8 model demonstrates superior performance in detecting strawberries under various conditions, including occlusions and varying lighting. The study highlights the effectiveness of the model in enhancing the accuracy and efficiency of automated harvesting robots, providing technical support for smart agriculture. The algorithm is adaptable and can be extended to other fruit crops. The results indicate that the improved model significantly enhances the accuracy of strawberry maturity detection while maintaining real-time processing capabilities. The study also discusses the limitations of the model, including the need for further verification of its generalization to other fruits and crops, and the need to explore its adaptability under diverse conditions. Overall, the improved CES-YOLOv8 model provides a robust solution for strawberry maturity detection in smart agriculture.This study proposes an improved CES-YOLOv8 model for strawberry maturity detection, enhancing the accuracy and efficiency of automated harvesting robots. The model is based on the YOLOv8 framework, with modifications to the backbone network, attention mechanism, and loss function. The backbone network replaces some C2f modules with ConvNeXt V2 to improve feature extraction, while the ECA attention mechanism enhances feature representation. The SIoU loss function is used to improve the intersection over union (IoU) and enable faster localization of prediction boxes. The model was trained on a dataset of 546 strawberry images collected under various lighting conditions, occlusion levels, and angles. Data augmentation techniques were applied to increase the dataset size to 2722 images. The model achieved an accuracy of 88.20%, recall of 89.80%, mAP50 of 92.10%, and F1 score of 88.99%, representing improvements over the original YOLOv8 network. The model also maintains a high frame rate of 184.52 FPS, ensuring real-time performance. The improved CES-YOLOv8 model demonstrates superior performance in detecting strawberries under various conditions, including occlusions and varying lighting. The study highlights the effectiveness of the model in enhancing the accuracy and efficiency of automated harvesting robots, providing technical support for smart agriculture. The algorithm is adaptable and can be extended to other fruit crops. The results indicate that the improved model significantly enhances the accuracy of strawberry maturity detection while maintaining real-time processing capabilities. The study also discusses the limitations of the model, including the need for further verification of its generalization to other fruits and crops, and the need to explore its adaptability under diverse conditions. Overall, the improved CES-YOLOv8 model provides a robust solution for strawberry maturity detection in smart agriculture.