Design of a Delta Robot Combined with CNN for Image Classification
碩士 === 國立臺灣海洋大學 === 電機工程學系 === 106 === The main purpose of this thesis is to combine a delta robot with deep learning techniques for image classification. A convolutional neural network (CNN) is designed for the object image classification, and the delta robot arm is controlled to grasp the target o...
| Main Authors: | , |
|---|---|
| Other Authors: | |
| Format: | Others |
| Language: | zh-TW |
| Published: |
2018
|
| Online Access: | http://ndltd.ncl.edu.tw/handle/zx4363 |
| Summary: | 碩士 === 國立臺灣海洋大學 === 電機工程學系 === 106 === The main purpose of this thesis is to combine a delta robot with deep learning techniques for image classification. A convolutional neural network (CNN) is designed for the object image classification, and the delta robot arm is controlled to grasp the target object, and move it to the designate position autonomously.
The system architecture is roughly divided into the video image processing and the delta robot control subsystems. The former is responsible for the object localization and recognition and the latter is responsible for the precise robot arm control.
The image captured is passed to a personal computer for image processing. In addition to the coordinates computation and geometric rectification, the processed image is sent into a convolutional neural network to achieve image recognition. And the resulting information is used by the delta robot control subsystem.
The delta robot is a parallel robot consisting of three motors and a connected base controlled by an Arduino microcontroller. The designed program will receive the coordinate values and convert it to the angles of three servo motors. A suction cup on the base can then grab the target and place it on the desired target coordinates. The target in the experiment will be handwritten with different Arabic numerals, allowing the convolutional neural network to classify the image for experimental purposes.
|
|---|