YOLOv4-Lite–Based Urban Plantation Tree Detection and Positioning With High-Resolution Remote Sensing Imagery

Abstract

Automatic tree identification and position using high-resolution remote sensing images are critical for ecological garden planning, management, and large-scale environmental quality detection. However, existing single-tree detection methods have a high rate of misdetection in forests not only due to the similarity of background and crown colors but also because light and shadow caused abnormal crown shapes, resulting in a high rate of misdetections and missed detection. This article uses urban plantations as the primary research sample. In conjunction with the most recent deep learning method for object detection, a single-tree detection method based on the lite fourth edition of you only look once (YOLOv4-Lite) was proposed. YOLOv4’s object detection framework has been simplified, and the MobileNetv3 convolutional neural network is used as the primary feature extractor to reduce the number of parameters. Data enhancement is performed for categories with fewer single-tree samples, and the loss function is optimized using focal loss. The YOLOv4-Lite method is used to detect single trees on campus, in an orchard, and an economic plantation. Not only is the YOLOv4-Lite method compared to traditional methods such as the local maximum value method and the watershed method, where it outperforms them by nearly 46.1%, but also to novel methods such as the Chan-Vese model and the template matching method, where it outperforms them by nearly 26.4%. The experimental results for single-tree detection demonstrate that the YOLOv4-Lite method improves accuracy and robustness by nearly 36.2%. Our work establishes a reference for the application of YOLOv4-Lite in additional agricultural and plantation products.

Maize plant detection using UAV-based RGB imaging and YOLOv5

“…Intersection over Union (IoU) is the ratio of intersection and union of the prediction box and ground truth box. It determines the prediction case and evaluates the distance between the predicted box and the target box ( Zhang et al., 2022 ). An IoU of greater than 0.5 corresponds to TP, otherwise, to FN.…”

“…Given the advantages, the YOLO algorithm has been applied in a range of object detection applications requiring both simplicity and efficiency, particularly for plant detection tasks. For example, urban plantation tree detection with high-resolution remote sensing imagery based on YOLOv4-Lite ( Zheng and Wu, 2022 ), real-time strawberry detection based on YOLOv4 ( Zhang et al., 2022 ), crop diseases detection based on YOLOv5 ( Zhao et al., 2023 ), and wheat spike detection in UAV images based on YOLOv5 ( Zhao et al., 2021 ). Recently, variant versions of YOLOv5, notably the nano (n) and small (s) versions, referred to as YOLOv5n and YOLOv5s, respectively, have become attractive, considering the real-time performance requirements of YOLOv5 applied to UAVs or field robots.…”

Rapid Target Detection of Fruit Trees Using UAV Imaging and Improved Light YOLOv4 Algorithm

“…Chen et al [27] used the improved YOLOv3 model for spruce counting based on UAV images, and achieved fast and accurate counting of spruce numbers by adding a dense connection module and an over the module to the trunk extraction network Darknet-53. Zheng et al [28] proposed an improved YOLOv4-tiny-based single-wood detection method, which finally achieved a performance optimization of nearly 46.1% compared to traditional methods such as the local maximum method and the watershed algorithm; it also outperformed novel methods such as the Chan-Vese model and the template matching method by nearly 26.4% compared to them. The above study shows that deep learning is highly robust for tree canopy detection.…”

A Thermal Infrared Pedestrian-Detection Method for Edge Computing Devices

“…Hence, the MobileNet [20] series and the ShuffleNet [21] series were presented, which are represented by a lightweight base model [22]. Zheng et al [23] utilized MobileNetv3 and depthwise separable convolution to achieve parameter compression and inference acceleration of Yolov4 (i.e., Yolov4-lite). Wang et al [24] achieved SSD parameter compression using MoblieNetV2 in infrared image pedestrian detection.…”