Extraction of a 3D Map of the Greenhouse Environment and Detection and Segmentation of Pots Using Stereo Vision

Creating a map of the greenhouse environment and determine the position of the pots on this map, which are the main obstacles in agricultural environments, especially greenhouses, is an essential step in automating agricultural operations. In this research, using stereovision, the map from the greenhouse environment was extracted and the pots in this map were detected and segmented. To reach this goal, ROS framework, nodes and network connections in this framework, was used. To evaluate the designed algorithm, the error rate is calculated using Euclidean distance between estimated locations and actual locations of pots. The results of this study showed that 100% of the pots were identified and positioned. The evaluation results showed that the mean errors in estimating the position of the pots was 0.056 and Root mean squared error (RMSE) was 0.0006. Also, the maximum error in estimating the position of the pots was 0.137m and the minimum error was 0.005m. The results showed that the designed algorithm has a high accuracy in estimating the position of the pots