Path Recognition for Agricultural Robot Vision Navigation under Weed Environment
In this paper, a path recognition method for agricultural robot vision navigation under weed environment is proposed. The vision navigation is based on color images sampled by a high speed camera. First, the crop and weed information is extracted using an
- PDF / 1,065,550 Bytes
- 7 Pages / 439.363 x 666.131 pts Page_size
- 2 Downloads / 199 Views
Abstract. In this paper, a path recognition method for agricultural robot vision navigation under weed environment is proposed. The vision navigation is based on color images sampled by a high speed camera. First, the crop and weed information is extracted using an appropriate color feature model to separate the green crop from background; then the image is thresholded, and the noise caused by weed is filtered by deleting small-area objects in the image; the navigation path is extracted through Hough transformation. Experiments are carried out in corn seedling field, and results show that the method can recognize navigation path correctly under weed environment. Keywords: vision navigation, agricultural robot, weed noise filtration, path recognition.
1
Introduction
Autonomous agricultural robot has many applications such as environment information collection, seeding, fertilizing, spraying, etc. The goal of autonomous robot navigation is to control the trajectory of the robot and keep it along the driving path [1]. Computer vision based navigation has been widely researched recent years due to its advantages such as wide detection range, rich target information, good cost performance and flexibility [2-4]. The key to vision navigation is identifying crop automatically from the images and recognizing the navigation path [5]. Usually, the central line of the crop is extracted as the baseline for agricultural robot vision navigation. But the weed in the field will interfere with the crop information because it often has the same green color as the crop, and causes false navigation path recognition results [6]. Astrand has researched on weeding robot, and used two separate vision system, one is used for navigation and the other is used for weed recognition [7]. Zhao Bo uses neural network algorithm to classify the field environment [8]; the algorithm can classify the weed environment, but it’s too time-consuming, and how to recognize the navigation path after environment classification is not specified. Xue Jinlin adopts a variable field-of-view method to guide an agricultural robot [9], which can improve the guiding performance at the end of a crop row. Dong Fuhong has built a row guidance system for an autonomous robot for white asparagus harvesting [10], but the system is specific to asparagus, thus lack of generosity. D. Li and Y. Chen (Eds.): CCTA 2013, Part I, IFIP AICT 419, pp. 242–248, 2014. © IFIP International Federation for Information Processing 2014
Path Recognition for Agricultural Robot Vision Navigation under Weed Environment
243
In this paper, we propose a path recognition method for agricultural robot vision navigation under weed environment. The areas of the connected components in the image are calculated, and the noise caused by weed is filtered by deleting small-area objects in the image. Experimental results verify the correctness and performance of the proposed method.
2
Path Recognition Method
To recognize the path for agricultural robot vision navigation under weed environment, first
Data Loading...
