Abstract
Detecting the flowering stage of tea chrysanthemum is a key mechanism of the selective chrysanthemum harvesting robot. However, under complex, unstructured scenarios, such as illumination variation, occlusion, and overlapping, detecting tea chrysanthemum at a specific flowering stage is a real challenge. This paper proposes a highly fused, lightweight detection model named the Fusion-YOLO (F-YOLO) model. First, cutout and mosaic input components are equipped, with which the fusion module can better understand the features of the chrysanthemum through slicing. In the backbone component, the Cross-Stage Partial DenseNet (CSPDenseNet) network is used as the main network, and feature fusion modules are added to maximize the gradient flow difference. Next, in the neck component, the Cross-Stage Partial ResNeXt (CSPResNeXt) network is taken as the main network to truncate the redundant gradient flow. Finally, in the head component, the multi-scale fusion network is adopted to aggregate the parameters of two different detection layers from different backbone layers. The results show that the F-YOLO model is superior to state-of-the-art technologies in terms of object detection, that this method can be deployed on a single mobile GPU, and that it will be one of key technologies to build a selective chrysanthemum harvesting robot system in the future.
Highlights
Numerous studies have shown that tea chrysanthemum can significantly inhibit the activity of carcinogenic substances, and boasts distinct anti-aging, cholagogic, antihypertensive, and other effects at the early flowering stage
Methods based on deep convolutional neural networks (CNNs) have made remarkable achievements in object detection tasks [4,5,6,7,8,9,10], under agricultural application scenarios, it is still difficult to build a lightweight network for a selective harvesting robot that can adapt to complex unstructured scenarios
We propose a lightweight CNN called Fusion-YOLO (F-YOLO), which can adapt to illumination variation, occlusion, and overlapping scenarios
Summary
Numerous studies have shown that tea chrysanthemum can significantly inhibit the activity of carcinogenic substances, and boasts distinct anti-aging, cholagogic, antihypertensive, and other effects at the early flowering stage. FPN uses multi-scale feature fusion to fuse the high-level and low-level feature maps, which has been adopted by YOLOv3 [19] and other single-stage object detection methods. Considering that object detection algorithms usually only contain the feature information of the image, instead of semantic information, Zhang et al [20] proposed detection with enriched semantics (DES) based on the single-shot multi-box detector (SSD) framework, in order to fuse the semantic information of the low-level and high-level features of an image. To realize the detection task, a variety of data enhancement methods were fused, and specific loss functions were used to train these modules By doing so, these function modules could better understand the features of chrysanthemum, and the performance of the lightweight network in complex unstructured environments could be improved in an end-to-end manner.
Sign-in/Register to view highlights & summary Sign-in/Register to access full text options 
Free) 
Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have
Schedule a call
