|
CSME 2024/02
Volume 45 No.1
:
23-32
|
|
Vision-Based Edge Detection Between Plant and Soil of Ningxia Vineyard
Wei Lia and Dong-yang Huangb
aZhejiang College of Security Technology, Wenzhou, Zhejiang 325016, Peoples R China
bSchool of Computer Science, Xi’an Polytechnic University, Xi’an, Shaanxi 710699, Peoples R China
|
Abstract:
Realizing mechanical automation is a necessary path for the scale and precision of vineyard production, and solving the automatic navigation problem of universal mobile platforms in unstructured environments is the key to realizing mechanical automation. In order to adapt to the unstructured road environment in the vineyard, this paper proposes a vision-based edge detection method between plant and soil of Ningxia vineyard. To study the composition of the road scene in visual perception, the road scene is mainly divided into plant elements and soil elements, and is segmented at pixel level by convolutional neural network to obtain semantic information and achieve end-to-end pixel level prediction, based on which, binarization is performed and the noise is reduced, and then an edge extraction model based on RGB color is established according to the visual features, the edge between plant and soil is extracted. The results show that the average accuracy of edge detection is 96.39%, and the average running time of edge detection is 0.2319s. The method can effectively realize edge detection and make the proposed edge detection technique applicable to a complex, dynamic and multi- variable scene, which can intervene and proactively adapt to unstructured environments, which is of great practical significance.
|
|
Keywords: edge detection, image processing, plant and soil, automatic navigation, vision.
|
Download PDF
|
|
*Corresponding author; e-mail:
|
©
2024
CSME , ISSN 0257-9731
|