| The apple industry is one of the major fruit industries in the world and has an important impact on agricultural economy and industrial development.There is a great relationship between the flowering intensity of apple flowers and fruit quantity and fruit quality,while the detection of apple flower growth status is very important for the control of flower thinning time in orchards.The detection of flowering intensity and growth state of apple flowers in apple orchards by deep learning can effectively solve the problems of incomplete and low efficiency of traditional methods for detecting flowering intensity in orchard production,as well as the problems of untimely and inaccurate manual identification of growth state for flower thinning decisions.Therefore,this paper investigates a deep learning-based method to identify and detect apple flower intensity and growth status,and designs a GUI interactive interface for apple flower identification and detection,which is of high practical value for the development of automatic flower thinning equipment.（1）The detection performance of different target detection networks on apple flower datasets was investigated.For the problem of small apple flower dataset,the own apple flower image dataset was built with different growth stages of a single flower and a cluster of blooming apple flowers,respectively.Data enhancement was performed on the original dataset to increase the number of datasets and optimize the dataset quality.The detection performance performance was studied by building different mainstream target detection based networks Faster R-CNN,YOLO v3,YOLO v4 and YOLO v5 models,and then migration learning and training on two apple flower datasets.By comparing the accuracy and detection time of different models for apple flower intensity and growth status detection,YOLOv5 was finally selected as the base network for apple blossom detection,and certain improvements were proposed for the base model.（2）An apple flower detection method based on the improved YOLOv5 model was proposed.Firstly,the K-means clustering method was improved,and the K-means++ algorithm was used to solve the problem that the randomly selected clustering centers cannot reach the global optimum when clustering;secondly,the cooperative attention mechanism was incorporated into the original YOLOv5 feature extraction network to obtain more shallow features to enhance the network performance;then,to improve the training speed and reduce the computation of the network model,the Ghost-Bottleneck was written into YOLOv5 to replace the Bottleneck module;finally,CIOU was adopted as the loss function of bounding box regression to improve the stability of the target box regression.（3）The performance of the improved YOLOv5 model for detecting apple flower flowering intensity and growth status was investigated.On the apple flower flowering intensity dataset,the recognition detection method based on the improved YOLOV5 model had the highest AP value,which was 2.53%,14.56%,5.08% and 2.42% higher compared to Faster R-CNN,YOLO v3,YOLO v4 and YOLOv5,respectively.On the apple flower growth state dataset,the method detected 94.90% of m AP values at different stages of apple flowers,which improved 1.98%,7.1%,5.42%,and 2.53% over Faster R-CNN,YOLO v3,YOLO v4,and YOLOv5,respectively.Compared with other detection methods,this method had higher accuracy of apple flower detection and was suitable for the detection of flowering intensity and growth status of apple flowers in orchards.（4）A GUI interactive interface for apple flower identification and detection was designed and implemented.The design of the GUI human-computer interaction interface was carried out by Py Qt5 and Py Charm.Finally,based on the generated apple flower detection model and algorithm,system debugging and testing were conducted,and the results showed that the designed apple flower detection interface was easy to operate,simple and intuitive.The apple flower identification detection method proposed in this study improves the performance of apple flower detection and can lay the foundation for the development of flower thinning equipment. |
PDF Full Text Request