Identification and Depth Localization of Clustered Pod Pepper Based on Improved Faster R-CNN

Abstract

Traditionally height of end effector of pod pepper harvester is fixed, which induces it hardly adapt to growth height of clustered peppers. Firstly, aiming at the problems of small size and clustered growth of pepper fruits during identification task, an improved Faster R-CNN algorithm is proposed. On the one hand, strategies such as increasing the types and number of high-resolution anchors and using RoI Align instead of RoI Pooling are employed to improve the detection accuracy for tiny targets. On the other hand, ResNet+FPN instead of VGG16 and ResNet backbone structure is adopted as the low-level feature extractor, so extracting capability for small features can be enhanced effectively. Furthermore, to precisely locate the position of clustered peppers, a height calculation model combining the 2D image recognition results with its depth information is advanced. Comparative experiments show that the overall accuracy AP and AP50 of our method reach 75.79% and 87.30%, respectively. Compared with VGG16 feature extraction model, the two indicators are improved by 8.7% and 1.3%, respectively. The small target detection accuracy AP small is increased about 11.4%, with recall rate AR small increased up to 10.2%. The overall loss rate Loss is reduced by 4.7%, which manifests greatly improvement compared to YOLOv3 model. The detection time of a single frame reaches 42ms, which is slightly longer than that of YOLOv3 network, but it can still meet the real-time detection requirements of pepper harvester. In 3D location experiment, the average absolute height error of clustered peppers from the ground is 4.4mm, that accounts to the relative average error of 1.1%, thus suffices the adjustment error requirement of the end effector.

Research and Explainable Analysis of a Real-Time Passion Fruit Detection Model Based on FSOne-YOLOv7

“…To evaluate their performance, we compared SPPF and SimSPPF with the SPPCSPC employed in the original YOLOv7 model. A randomly sized input tensor with dimensions [8,1024,20,20] was generated, and it was passed through these three structures to obtain output tensors with dimensions [8,512,20,20]. The inference time required for 100 iterations of computation was measured and the experimental results are presented in Table 3.…”

“…For instance, Pan et al [19] achieved the automatic detection of sugarcane seedlings with an average accuracy of 93.67% using an enhanced Faster R-CNN model and a non-maximum suppression algorithm. Zhong et al [20] combined an improved Faster R-CNN model with depth information to locate clustered chili peppers, achieving an average precision (AP) of 87.30%. Moreover, Kumar and Kukreja [21] proposed a wheat leaf virus detection algorithm based on Mask R-CNN, achieving a remarkable detection accuracy of 97.16%.…”

Rapid detection of Yunnan Xiaomila based on lightweight YOLOv7 algorithm

“…However, it did not take into account factors such as model size and runtime speed. Zhong et al. (2022) proposed an improved fast R-CNN algorithm for the small size and cluster growth of pepper fruits in the detection process, which effectively improved the ability to extract small features.…”

“…However, it did not take into account factors such as model size and runtime speed. Zhong et al (2022) proposed an improved fast R-CNN algorithm for the small size and cluster growth of pepper fruits in the detection process, which effectively improved the ability to extract small features. Cong et al (2023) proposed an improved Mask RCNN with the Swin Transformer attention mechanism and exploited UNet3+ to improve the mask head and mask segmentation quality to efficiently segment sweet peppers of different categories under leaf occlusion.…”

MSGV-YOLOv7: A Lightweight Pineapple Detection Method

“…In contrast, twostage object detection algorithms, such as Faster R-CNN, involve generating candidate regions and subsequently employing convolutional neural networks for object detection. Zhong et al [27] enhanced the Faster R-CNN model by incorporating depth information to accurately locate clustered peppers, achieving an AP of 87.30%. Fangfang Gao et al [28] employed Faster R-CNN for apple detection, attaining an average precision of 0.879 for apples under various obstructions, including leaves, branches, wires, and other fruits.…”