Design and Implementation of a Multi-function Farm Machine based on Machine Vision in Nursery

• Main Authors: Lin, Kuan-Min, 林冠旻
• Other Authors: Chang, Chung-Liang
• Format: Others
• Language: zh-TW
• Published: 2018
• Online Access: http://ndltd.ncl.edu.tw/handle/7n4fm9

碩士 === 國立屏東科技大學 === 生物機電工程系所 === 106 === Commonly speaking, a highly effective agriculture machinery can always provide a higher productivity in farming. However, most modern machineries are designed with a single function and high in cost. Thus, a farmer will usually need a few more different machineries to carry out the farm works. Besides, each of those farm machines might call for a different professional skillset to operate and which is both time-consuming and cost-consuming to switch machines. This thesis proposes a multifunctional intelligence farm machine, built in a machine vision platform to identify target crops and weeds, and to measure soil moisture and locate weeds. Furthermore, the built in platform is also equipped with a programmed multitasking system and a fuzzy logic algorithm, connecting a few external soil humidity sensors to collect data from the experiment site and planning the working progress, to decide the movement of the claw to perform weeding or drive the pump to watering. The image processing system used to identify objects in the thesis adopts HSV layer value extrusion and then applies a suitable threshold. The experiment results have proven that this intelligence farm machine is able to identify different target crops at the actual outdoor field, with an average success rate of 95% and above. Then the system core platform will give a strategic progress. The designed mechanism will conduct the following steps, with an accuracy rate of 80%. Lastly, the intelligence farm machine proposed in this thesis will work with another guiding vehicle in order to be multitasked in the farm.