Search

• YOLOv8-GP based on key point detection is proposed. • C2f in the backbone network is replaced by C2f-Faster-EMA. • Multi-scale feature fusion is achieved by using BiFPN in Neck network. • While reducing the calculation, the AP of grape is improved by 3.3 %. • The positioning error of the picking point is kept within 30 pixels. Precise positioning of fruit and picking point is crucial for harvesting table grapes using automated picking robots in an unstructured agricultural environment. Most studies employ multi-step methods for locating picking points based on fruit detection, leading to slow detection speed, cumbersome models, and algorithmic fragmentation. This study proposes an improved YOLOv8-GP (YOLOv8-Grape and picking point) model based on YOLOv8n-Pose to solve the problem of simultaneous detection of grape clusters and picking points. YOLOv8-GP is an end-to-end network that integrates object detection and key point detection. Specifically, the Bottleneck in C2f is replaced with FasterNet Block that incorporates EMA (Efficient Multi-Scale Attention), resulting in C2f-Faster-EMA. BiFPN is applied to substitute the original PAN as Neck network. The FasterNet Block, designed based on partial convolution (PConv), reduces redundant computation and memory access, thereby extracting spatial features more efficiently. The EMA attention mechanism achieves performance gains with lower computational overhead. Furthermore, BiFPN is employed to enhance the effect of feature fusion. Experimental results demonstrate that YOLOv8-GP achieves AP of 89.7 % for grape cluster detection and a Euclidean distance error of less than 30 pixels for picking point detection. Additionally, the number of Params is reduced by 47.73 %, and the model complexity GFlops is 6.1G. In summary, YOLOv8-GP offers excellent detection performance, while the reduced number of parameters and model complexity contribute to lower deployment costs and easier implementation on mobile robots. [ABSTRACT FROM AUTHOR]

• A prediction point Hough transform with a maximum error of 0.5°is proposed. • Based on the G-B and 2G-B-R grayscale factors, a new grayscale factor is proposed. • The least square method is used to optimize the traditional Hough transform. Accurate extraction of navigation path is very important for automated navigation of agricultural robots. Based on the machine vision system, this paper proposes a new algorithm for the fitting of navigation path of greenhouse cucumber-picking robots. Aiming at the problems that the traditional Hough transform has a large amount of calculation and the least square method has a low precision, the prediction point Hough transform algorithm is proposed to extract the navigation path. The prediction-point Hough transform contains 4 parts: intercept area of interest, image segmentation, navigation point extraction, navigation path fitting. In this paper, only the final 160-pixel rows of image captured by the camera are taken as the region of interest. In the image segmentation stage, this paper proposes a new graying factor. For navigation path extraction, a regression equation is used to determine the prediction point, and finally the proposed prediction point Hough transform is used to fit the navigation path. The experimental results show that the proposed grayscale factor can well segment cucumber plants and soil, and the segmentation effect is better than 2G-B-R and G-B grayscale factors. The proposed prediction point Hough transform fits the navigation paths with an average error less than 0.5°, which is 10.25° lower than the average error of the least-square method. Also, the computation time of the proposed Hough transform is 17.92 ms. Compared with the traditional Hough transform, it saves 35.20 ms. [ABSTRACT FROM AUTHOR]

• A median point Hough transform with a maximum error of 0.5°was proposed. • Based on the 2G-B-R factor, a new grayscale factor is proposed to realize the segmentation of crops and soil. • Get the coordinate of the navigation points through the relative coordinate center of the pixel. The extraction of navigation paths is essential for the autonomous navigation of an agricultural robot. As most of the current navigation path extraction algorithms are time-consuming, difficult and susceptible to interference, the median point Hough transformation algorithm was proposed in this paper based on an optical system to fit the navigation paths of a greenhouse tomato-cucumber spraying robot quickly and accurately. The algorithm consists of four parts: interception of the target area, image segmentation, navigation points extraction and navigation path fitting. Image segmentation was based on the improved grayscale factor and Otsu method to achieve the segmentation of soil and plants. The navigation points then were extracted based on the relative coordinate center of the white pixel points of the binary image. Finally, the navigation path was fitted by the median point Hough transformation method. The experimental results show that the maximum error of the navigation path fitted by the improved Hough transformation was 0.5° and the time taken was 7.13 ms, which is 45. 24 ms faster than the traditional Hough transform and on average 5.1° more accurate than the least squares method. [ABSTRACT FROM AUTHOR]