A 4-Axis Robot Controller for Detection Thrips and Orchid Midge in Dendrobium Orchid

Abstract

Thrips and Orchid midge are major pests of Dendrobium orchid throughout the year. Usually farmers found thrips after an outbreak had already occurred because of their small size. Signs of orchid midge outbreak were the vitrification and distortion on the orchid flower buds which were difficult to observe. The objective of this research was to design a detection equipment for thrips and orchid midge which could reduce error of the human labors. The prototype had two major parts. Part I was a Convolutional Neural Network (CNN) which had been used for images classifications. Part II was a 4-DOF robot arm which could be moved for insect pest detection through a digital camera. It was tested with 30 clumps of Dendrobium orchid grown in a greenhouse at Mueang district, Nakhon Pathom province and compared
its performance with a human labor. Results showed that the efficacy of the prototype in detecting thrips was 81.1%, for orchid midge was 88.1% and the average time of detection was 25.10 sec./clump. For human labor, the efficacy of detection for thrips was 75.8%, for orchid midge was 83.3% and the average time of detection was 53.37 sec./clump. Human labor worked better at the beginning as compared to the prototype, but in the latter time the prototype was better due to human labor fatigue.