BI-US/17-18-012: Pruning by Numbers: Integrating tree physiology with growth simulation and 3D reconstruction to optimize apple tree pruning

Tree pruning is one of the most important measures to ensure high yield performance by maintaining a balance between vegetative and reproductive growth [1]. This task is at each year before the growing season carried out manually by tree growers and requires deep understanding of tree physiology to predict a tree response to the pruning and years of practice. To increase the understanding of tree reactions to the pruning the appropriate computer models have been developed very early. First such model can be found in 1996, when the virtual reality application for apple tree growth after the pruning has been presented [2]. Since then a lot of simulators have been developed for various types of fruit trees (e.g., peach [3], cherry [4], and walnut [5]), but apple trees by far retained the most attention of researchers. Recent software simulators took full advantage of graphical processing units, which enables realistic real-time tree visualization. A good example of new generation simulators is IMapple [6], which incorporates a precise functional-structural tree growth model for Golden Delicious apple trees, based on long-time measurements with a photorealistic tree visualization employing 3D manifold watertight meshes for representing the tree geometry, together enabling interactive tree growth simulation including flowering and fruiting. Interactive simulation and faithful 3D geometric models offers also EduAPPLE, interactive teaching tool for apple tree crown formation [7], where user train a one year old apple tree in a form of unbranched whip towards the desired tree form using different tree training techniques. The growth model in EduAPPLE incorporates only the most basic tree growing rules, common to all cultivars, while growing condition are modelled with random variables in order to maximize tool generality. However, in order to improve users pruning skills this is not enough. To do that the teaching tool has to take more active role and start to suggest the user which branches have to be removed in order to achieve best results. The first step in that direction represents the work of [8], where the pruning is presented as a combinatorial optimization problem of performing the cuts on a virtual tree model in order to achieve the best light distribution inside the tree crown, which is a good start, but not sufficient to actually be used as recommendation system. Recent advances in 3D geometric modeling have shown an unprecedented precision in detailed reconstruction [9], inverse modeling [10], and physics-based responses of vegetation [11]. The advance of moderns GPU and recent algorithms for vegetation modeling [12] allowed realistic and interactive modeling and simulation of plants at scales and geometric details that were not possible before. The purpose of this bilateral project is to combine the approaches of [6-12] in order to develop new pruning optimization software that would not only be used as a teaching tool but also a pruning recommendation system for seasoned fruit growers, who want to improve their pruning techniques. To achieve that we have to integrate tree physiology knowledge collected over long-time observations into EduAPPLE and develop a pruning recommendation system based on criteria resulting from this new knowledge. The expected results would provide novel approach in the training of apple tree pruning to all those, who would like to acquire pruning knowledge in order to ensure local fruit source from their gardens, as well as the virtual assistant to all those who want to optimize apple tree pruning in their orchards.