Fruit size and leaf area are important indicators for plant health and are of interest for plant nutrient management, plant protection and harvest. In this research, an image-based method for measuring the fruit volume as well as the leaf area for cabbage is presented. For this purpose, a mask region-based convolutional neural network (Mask R-CNN) was trained to segment the cabbage fruit from the leaves and assign it to the corresponding plant. The results indicated that even with a single camera, the developed method can provide a calculation accuracy of fruit size of 92.6% and an accuracy of leaf area of 89.8% on individual plant level.
The cultivation of orchard meadows provides an ecological benefit for biodiversity, which is significantly higher than in intensively cultivated orchards. The goal of this research is to create a tree model to automatically determine possible pruning points for stand-alone trees within meadows. The algorithm which is presented here is capable of building a skeleton model based on a pre-segmented photogrammetric 3D point cloud. Good results were achieved in assigning the points to their leading branches and building a virtual tree model, reaching an overall accuracy of 95.19 %. This model provided the necessary information about the geometry of the tree for automated pruning.
Image-based yield detection in agriculture could raiseharvest efficiency and cultivation performance of farms. Following this goal, this research focuses on improving instance segmentation of field crops under varying environmental conditions. Five data sets of cabbage plants were recorded under varying lighting outdoor conditions. The images were acquired using a commercial mono camera. Additionally, depth information was generated out of the image stream with Structure-from-Motion (SfM). A Mask R-CNN was used to detect and segment the cabbage heads. The influence of depth information and different colour space representations were analysed. The results showed that depth combined with colour information leads to a segmentation accuracy increase of 7.1%. By describing colour information by colour spaces using light and saturation information combined with depth information, additional segmentation improvements of 16.5% could be reached. The CIELAB colour space combined with a depth information layer showed the best results achieving a mean average precision of 75.
A current problem in the design of small and lightweight autonomous agricultural robots is how to create sufficient traction on soil to pull an agricultural implement or load. One promising solution is offered by the interlock drive system, which penetrates spikes into the soil to create traction. The combination of soil penetrating spikes and a push-pull design offers new possibilities for vehicle control. By controlling the interlocking of the spikes and pushing and pulling them against the main frame, the vehicle can perform tight maneuvers. To validate this idea, we designed a robot, capable of creating high traction and performing headland turns. The navigation of the new robot system is performed by actively pushing the spikes, mounted on a slide into the soil, while the main frame is pushed back and pulled forward. The vehicle of 2-meter length was able to turn on the spot, and could follow a straight line, just using the spikes and the push-pull mechanism. The trajectory and the performed measurements suggest, that a vehicle which uses only spikes for traction and steering is fully capable of performing autonomous tasks in agriculture fields.