Detection of Farmland Obstacles Based on an Improved YOLOv5s Algorithm by Using CIoU and Anchor Box Scale Clustering

Abstract

It is necessary to detect multi-type farmland obstacles in real time and accurately for unmanned agricultural vehicles. An improved YOLOv5s algorithm based on the K-Means clustering algorithm and CIoU Loss function was proposed to improve detection precision and speed up real-time detection. The K-Means clustering algorithm was used in order to generate anchor box scales to accelerate the convergence speed of model training. The CIoU Loss function, combining the three geometric measures of overlap area, center distance and aspect ratio, was adopted to reduce the occurrence of missed and false detection and improve detection precision. The experimental results showed that the inference time of a single image was reduced by 75% with the improved YOLOv5s algorithm; compared with that of the Faster R-CNN algorithm, real-time performance was effectively improved. Furthermore, the mAP value of the improved algorithm was increased by 5.80% compared with that of the original YOLOv5s, which indicates that using the CIoU Loss function had an obvious effect on reducing the missed detection and false detection of the original YOLOv5s. Moreover, the detection of small target obstacles of the improved algorithm was better than that of the Faster R-CNN.

Multi-Objective Association Detection of Farmland Obstacles Based on Information Fusion of Millimeter Wave Radar and Camera

“…Firstly, the effect of the information fusion of the mmWave radar and the camera was tested in a non-agricultural environment, as shown in Figure 13 . In Figure 13 a, three obstacle targets such as people, houses, and trees were detected with the camera only, by using the improved YOLOv5s algorithm in the literature [ 9 ]. But after the data fusion, 8 targets sequences including position, longitudinal speed, and category were output: (−9.84, 13.21, 0.00, house), (−7.96, 17.57, 0.00, tree), (−5.32, 14.63, 0.00, house), (−0.21, 4.30, 0.00, person), (4.36, 13.51, 0.00, tree), (−4.78, 17.08, 0.00, nan), (−4.85, 12.83, 0.00, nan), (5.64, 13.97, 0.00, nan), as shown in Figure 13 b.…”

“…For the visual inspection module, this paper adopted improved YOLOv5s [ 9 ], automatically generated the anchor frame scale through the K-Means algorithm to speed up the convergence speed, and used the CIoU loss function to reduce false detection and missing detection to improve the accuracy. For the mmWave radar detection module, in order to reduce the amount of data during fusion input, a three-step filtering algorithm including empty target filtering, false target filtering, and non-threat target filtering was adopted, which was mainly realized through relative distance, effective target life cycle, and horizontal and vertical coordinate threshold.…”

“…(2) Decision level data fusion is designed after the effective target detection of the mmWave radar and camera, which gives play to their respective advantages and improves the reliability and robustness of detection. The improved YOLOv5s algorithm is used for vision detection of farmland obstacles, as shown in [ 9 ].…”

“…In agricultural fields, single or multiple sensors such as LIDAR, vision sensors, structured light sensors, ultrasonic sensors are usually used for obstacle detection [ 5 , 6 , 7 , 8 ]. In [ 9 ], the authors proposed an improved YOLOv5s algorithm for vision detection of farmland obstacles, which can improve detection precision and speed up the real-time detection.. In [ 10 ], the author introduced data acquisition, processing technology and post-processing technology to demonstrate the ability of detecting partially obscured targets from foliage with mmWave radar.…”

STAE‐YOLO: Intelligent detection algorithm for risk management of construction machinery intrusion on transmission lines based on visual perception

“…For the target detection task, the loss function of the bounding box regression (BBR) [34] is crucial. The baseline model we use is the complete-intersection over union (CIoU) loss [35], which considers three geometric measures, including the overlap between the predicted and target frames, the distance to the centroid, and the consistency of the aspect ratio. The CIoU formula is shown below:…”