Thinning of pear and other fruit species is an important horticultural practice that is used to enhance fruit set and quality by removing excess floral buds. Though it is traditionally carried out after fruit set, an earlier reduction of the crop potential or crop load promotes a more efficient use of the available carbohydrate resources. As thinning is currently most often conducted through time-consuming manual labor, it comprises a large part of a growers production costs. Various thinning machines developed in recent years have clearly demonstrated that mechanization of the thinning process is both feasible and cost-effective. However, these machines still lack sufficient selectivity to take into account the specific fruit bearing capacity of individual trees. Furthermore, the current devices often cause damage to shoots, leaves and fruitlets which decreases tree vigor and makes the trees more susceptible to diseases such as fire blight (Erwinia amylovora) and cankers.The main objective of this dissertation was the development of a novel mechatronic device suited for early-season thinning of pear (cultivar Conference), offering a high degree of selectivity and causing minimal tree damage. To these ends, two sub-objectives were set: (1) the development of a sensor to detect and quantify the early-season crop potential; and (2) the investigation of a new non-contact way of thinning using pulses of compressed air.Pre-bloom recognition of green floral pear buds in a green orchard environment is hard to realize with standard cameras. Therefore, a dedicated multispectral sensor has been designed for this purpose. This was done by measuring the spectral properties of the floral pear buds in the laboratory with a hyperspectral camera in the visible and near infrared region of the spectrum. A stepwise variable selection algorithm was then applied to the recorded data to select the combination of six wavebands which are best-suited for discriminating between floral buds and their environment. Using canonical correlation analysis (CCA), a spectral discriminant model was built with these wavebands. Using this model an overall 95.14 % correct pixel classification under laboratory conditions was achieved.Based on the selected wavebands, a multispectral camera system was designed and subsequently tested in a commercial pear orchard during two flowering seasons. Measurements were conducted at nighttime under controlled illumination to limit the visibility of background objects. As the conditions in an orchard differ from those in the laboratory, a new spectral discrimination model was built that recognizes pixels originating from floral buds, again using CCA. A custom image analysis algorithm was then developed and used to translate the pixel classification into object recognition. This algorithm recognized 87 % of the unoccluded floral buds, making the multispectral sensor system well-suited to increase the efficiency of new and existing thinning devices. In order to design the pneumatic thinning technique, the fracture strength required to remove a floral bud has been measured on a laboratory test bench. The required strength was found to vary as a function of bud development and reaches a minimum around the stage Green cluster. Based on this information, a simple pneumatic setup with a single nozzle has been built and tested during a two-year orchard trial to determine the effects of air pressure, nozzle type, distance and phenology on the attainable removal efficiency. Thinning grades as high as 93.13 % and 74.52 % were achieved during a dry and a wet season, respectively. Furthermore, pneumatic thinning was observed to cause limited damage since floral buds were removed at their natural attachment point on the supporting branch. Based on these results, pneumatic removal of floral pear buds is considered feasible.In a next step, a more elaborate prototype setup has been designed which contains multiple nozzles than can be moved in front of the canopy while blowing. This setup more closely approximates a fully operational machine by targeting a region of a tree rather than a single bud. This setup was tested in the orchard during one growing season and a maximum success rate of 36.59 % was obtained. This is clearly lower than the maximal thinning rate observed during the single nozzle trials. Suggestions were made for further improvements of the setup which could increase the performance.Finally, the integration of the detection sensor and the pneumatic thinning technique has been elaborated. Such a mechatronic system can realize precision thinning by choosing the required settings of the pneumatic thinning device based on the measured floral bud distribution. A second bud detection and counting procedure after the thinning operation can be used to assess the effectiveness of the procedure and to adjust the machine settings if required. Initial tests of this concept have been conducted in the orchard and positive results were obtained.A conservative cost price estimation of the mechatronic device has been presented. This analysis indicated that pneumatic thinning can be an economically feasible alternative to the traditional hand thinning.
For a version with higher resolution images, please contact email@example.com or