In order to solve the problem of Apple recognition in complex environment, an apple target detection algorithm based on YOLOv5s pre training model is proposed. The apple orchard photos taken under different illumination and different angles were annotated and normalized, and the image was linearly smoothed by Gaussian filtering. The preprocessed data set is used to train the model, and the trained model is used to count the number, position and maturity of apples in the data set image. The results showed that the AP value of 0 (i.e. fully mature apple) was 0.974, indicating that the model had good detection effect on mature apple. The lightweight model is expected to be carried on the apple picking robot with embedded GPU to realize accurate identification and picking of mature apples and reduce the expensive labor cost.
Apple Recognition Algorithm Based on YOLOv5
23.10.2024
999805 byte
Aufsatz (Konferenz)
Elektronische Ressource
Englisch
Traffic sign recognition based on YOLOv5
SPIE | 2022
|Ship Remote Sensing Target Recognition Based on YOLOV5
Springer Verlag | 2024
|