Apple target recognition method in complex environment based on improved YOLOv4

Abstract

In order to solve the problem of accurate recognition of apples in complex environments, this article proposes an apple recognition method based on improved YOLOv4, which can accurately locate and recognize apples in a variety of complex environments. This method uses the lightweight EfficientNet‐B0 network as the feature extraction network for apple recognition and then combines with the PANet (Path Aggregation Network) network to fuse the features in the adjacent feature layers, which improves the recognition accuracy of fruit targets. In the experiment, the apples under bagging, nonbagging and night environment are identified. The results of the experiment show that the average precision of the improved YOLOv4 method is 93.42%, the recall rate is 87.64%, and the F1 value is 0.9035, which is about 2% lower than that of the original YOLOv4. However, the storage memory required for the model trained on the improved YOLOv4 is decreased by 87.8% compared with the original YOLOv4, which is only 29.8 MB, and the recognition speed is increased by 43%, reaching 63.20 frames per second. The recognition accuracy and speed are much higher than that of the two‐stage Faster‐RCNN. The recognition accuracy is equivalent to that of Mobilenetv3‐YOLOv4, but the model of our method is smaller and the recognition speed is faster. Practical Applications Achieving rapid and accurate recognition of apples in complex environments is one of the key technologies for automatic fruit picking by picking robots, and it is also an important method and means for realizing orchard yield estimation. In order to recognize apples in the orchard environment more quickly and effectively, a light‐weight YOLOv4 target recognition method based on deep learning is proposed. The method proposed in this article can significantly reduce the storage space and computing power requirements of the device. On the basis of maintaining high accuracy, it can achieve real‐time and accurate recognition of apples in complex environments, and has great application value and practical significance for picking robot.

Sugar Beet Damage Detection during Harvesting Using Different Convolutional Neural Network Models

“…Comparing with the methodologies used for broken corn detection at a conveyor belt of a corn harvester based on different YOLO v3 models [15], the proposed YOLO v4 According to the performance of YOLO v4, it achieved a good balance between precision, recall, F1-score and speed which could be considered as the best model for sugar beet damage detection during harvesting. This finding is in line with reported studies that the YOLO networks could achieve higher speed and better overall performance e.g., [31,33,36]. Furthermore, our finding is in agreement with [37] who reported that YOLO models achieved higher speed and F1-score compared with SVM, Faster R-CNN for apple surface defect detection.…”

“…In this study, the YOLO v4 model shows better performance compared to the other developed networks. This finding is in line with previous studies conducted for detection of citrus in an orchard [30], apples in a farming complex environment [31], pests [32] and tree trunks in a forest [33]. However, the Faster R-CNN NAS model shows lower performance in this research.…”

“…In this context, results of a reported study by [34] showed that the Faster R-CNN NAS model had higher precision than Faster R-CNN Inception v2 for kiwi fruit detection; however, Faster R-CNN Inception v2 showed better recall compared to the Faster R-CNN NAS model. detection of citrus in an orchard [30], apples in a farming complex environment [31], pests [32] and tree trunks in a forest [33]. However, the Faster R-CNN NAS model shows lower performance in this research.…”

“…In another study, pear fruit detection and counting using different YOLO v4 models (i.e., YOLO v4, YOLO v4-CSP, YOLO v4-tiny) resulted in achieved average precision of 98% [35]. Furthermore, [31] reported that YOLO v4 was able to detect apple fruit in a complex environment with recall and an average precision of around 93 and 88%, respectively, compared to Faster R-CNN with 90 and 83%, respectively. The use of R-FCN as a detector and ResNet101 as a feature extractor showed also high values of precision and recall of more than 98% in a study by [13] for potato surface defect detection.…”

Precision Detection of Dense Plums in Orchards Using the Improved YOLOv4 Model

“…The mAP of the improved model increased by 2.38∼4.81% through the analysis of detection effects under different lighting conditions, occlusion, and maturity. Ji et al (2021) proposed an apple detection method based on the improved YOLOv4, which could accurately locate and detect apples in various complex environments. Although the YOLO series networks have shown excellent performance in fruit recognition, it is difficult to detect small targets in deep feature maps due to the loss of spatial and detailed feature information.…”