Tomato ripeness detection method based on FasterNet block and attention mechanism

In modern agriculture, accurate detection of tomato maturity is crucial for efficient harvesting and grading. Traditional detection methods rely on manual experience, which is time-consuming, inefficient, and prone to subjective interference, making them unsuitable for large-scale production. To address this, this study proposes a tomato maturity detection model based on an improved YOLOv11n, incorporating the C3k2-Faster-EMA module to enhance the model's feature extraction capability and detection efficiency. In addition, the SimAM attention mechanism is introduced, enabling the model to intelligently focus on key features of the tomatoes, thereby improving its ability to recognize tomatoes at different maturity stages and enhancing detection accuracy. Furthermore, the generalized intersection over union loss function is employed to introduce a target box overlap metric, optimizing the object localization process and improving the precision of fruit positioning. Experimental results on the tomato maturity dataset show that the proposed method performs excellently in tomato maturity detection, achieving an mAP of 86.0% and an accuracy of 85.4%. Compared to the baseline model, the number of parameters is reduced by 11.2%, while the frames-per-second detection speed is increased by 23.1%, with significant improvements in stability. This provides reliable technical support for intelligent harvesting and grading, with broad application prospects.