한국해양대학교

Detailed Information

Metadata Downloads

머신 비전과 임베디드 制御機 具現에 의한 知能形 3次元 形狀 測定 시스템 開發

Title
머신 비전과 임베디드 制御機 具現에 의한 知能形 3次元 形狀 測定 시스템 開發
Alternative Title
Development of Intelligent 3D Shape Measurement System based on Implementation of Machine Vision and Embedded Controller
Author(s)
김영탁
Publication Year
2006
Publisher
한국해양대학교 대학원
URI
http://kmou.dcollection.net/jsp/common/DcLoOrgPer.jsp?sItemId=000002174890
http://repository.kmou.ac.kr/handle/2014.oak/9013
Abstract
This thesis proposes a system to reconstruct and measure three dimensional image using the computer vision system, the embedded system, and the method of artificial intelligence, and implement mechanical structures, electronics hardware, and software appropriate for this system.

The method of reverse engineering is sequential processes: First, it reconstructs to three dimensional image about an object, then measures this image. Finally, it analyzes information that obtain measured result. More precisely, this method is the technique that get properties of geometric data on an object constructed or been in nature.

Recently, in most applications, for examples, quality evaluation of products, anthropolatry, computer aided design, reverse engineering, computer graphics, virtual environment and so on, apply three dimensional information of an object that obtain through contactless method using three dimensional vision system, rather than two dimensional analysis.

So, in this thesis suggest a system that can analyze characteristics of object, based three dimensional vision system using CCD camera and slit beam laser.

The images used to reconstruct three dimensional shape are inputted images that scanned an object using slit beam laser and projected from three dimensional world coordinate through eight CCD cameras arranged in two axises(Each axis has four CCD cameras.)

Then, it is to extract the parts come out the slit beam laser in the images using algorithms of image processing and to construct point clouds needed three dimensional reconstruction using intrinsic and extrinsic parameters of pin-hole camera model with calibration process.

In this thesis propose a robust binary filter that remove noise in image processing using neural network, intelligent method and process thinning of line image projected on object to reduce amount of computation in data.

Here use neural network as classification to classify feature data extracted in three dimensional information of object formed through post processing. Also, to control whole measuring system of three dimensional, PC and embedded processor are combined cooperatively, and are driven separately the part of control and of data processing. Here, the embedded controller ported RTOS is powerful application in multitasking environment for managing entities, rather than foreground and background system. And, the actuator for moving of measured frame is consist of the DC motor which is able to control feedback loop using encoder. But, it support that fuzzy controller based fuzzy theory in the part of motor control allows higher performance compared with PDI control, classical control. The experimental object used in this thesis is a foot of human, the characteristics obtained to reconstruct a shape of foot allow applications of medical science or manufacturing of shoes to apply.

Today, technology measuring size or shape of things three dimensionally using various tools in many fields develop rapidly, has various applications. Thus, it is expected that most consumer are satisfied about product when this technology is applied to shoes industry. However, this technology has need of the method of artificial intelligence, fuzzy or neural network to improve software, mechanic, electronics problem often caused development of product, so this allow more human oriented design and manufacture. Therefore, in this thesis the system proposed will be widely applied human foot measured in this, as well as all the industry and the academic world needed reverse engineering.
Appears in Collections:
전자통신공학과 > Thesis
Files in This Item:
000002174890.pdf Download

Items in DSpace are protected by copyright, with all rights reserved, unless otherwise indicated.

Browse