摘要
在科技和工業(yè)不斷發(fā)展的同時，地球的資源也在極度的消耗之中，如何節(jié)約和再利用資源也越來越受到重視。在這種環(huán)境下，本文利用機器視覺技術對廢舊機械零件進行識別分類研究，以便于以后再次循環(huán)利用。機器視覺是一門通過研究圖像或視頻數(shù)據(jù)來觀察周圍世界的學科，其核心內容是圖像的處理和識別。機器視覺是現(xiàn)代制造的一個重要組成部分，在現(xiàn)代機械行業(yè)中，為了提高生產(chǎn)的柔性和自動化程度，機器視覺得到了廣泛的應用。例如，機械零件的檢測和識別、產(chǎn)品分類以及自動化生產(chǎn)的裝配檢測等。因此，利用機器視覺對機械零件進行識別研究具有重要的意義。
本文主要以廢舊螺栓為研究對象，根據(jù)螺栓的幾何形狀特征提出了一種適合螺栓等典型機械零件的識別方法。本文的主要工作內容包括圖像識別理論和方法的介紹與分析、機械零件圖像的預處理、圖像的特征提取、機械零件識別和典型零件識別系統(tǒng)的設計。
在圖像識別理論及方法方面，本文主要對圖像匹配識別、統(tǒng)計模式識別、句法模式識別、模糊模式識別和人工神經(jīng)網(wǎng)絡模式識別等圖像識別方法進行介紹和分析。
在圖像的預處理方面，本文著重對圖像去噪和圖像邊緣提取進行研究。在對圖像進行去噪的過程中，本文主要對均值濾波、中值濾波和小波濾波進行研究。通過對幾種方法的比較，發(fā)現(xiàn)經(jīng)過中值濾波和均值濾波相結合的混合濾波方法具有較好的濾波效果。在機械零件圖像的邊緣提取過程中，經(jīng)過對現(xiàn)有幾種常見邊緣檢測算子的比較研究，發(fā)現(xiàn)利用Sobel算子進行邊緣提取，具有速度快、效果好的優(yōu)點。
在特征的提取方面，本文著重根據(jù)螺栓的幾何形狀特征對螺栓進行特征提取的研究。本文提出的螺栓的特征向量具有尺寸不變性、旋轉不變性和平移不變性等優(yōu)點。在識別的算法方面，本文根據(jù)提取出來的螺栓的特征向量利用RBF網(wǎng)絡對螺栓進行識別研究。仿真實驗表明RBF網(wǎng)絡技術應用于零件識別，具有較好的識別效果，且識別速度快。

關鍵詞：特征提取；零件識別；預處理；幾何形狀特征；RBF網(wǎng)絡


Abstract
With the technology and industry developing, the earth's resources are being consumed extremely. So how to save and reuse of resources are getting more and more attention. Under such conditions, we use the machine vision technology to identify classification of wasted machine parts for recycling in the future again. Machine vision is a discipline of studying image or video data to observe the world around. Its core content is the image processing and recognition. Its core content is the image processing and recognition. Machine vision is an important component of modern manufacturing. In order to raise flexibility and automation in production, machine vision has been widely used. Such as, detection and identification of mechanical parts, product classification, assembly testing in automated production. Using machine vision to study identification of mechanical parts is significant.
In this paper, mainly used wasted bolts for study object, according to the geometry characteristics of the bolt, we propose a suitable method for identification of bolts and other typical mechanical parts. Recognition of the simple machines, the main content of the work with mechanical parts image preprocessing, image feature extraction, part identification and part identification system design simple. In this paper, the main work includes analysis and introduces the theory and method of image recognition, the image preprocessing of machine parts, mechanical parts identification, the design of typical parts identification system.
In theory and methods of image recognition, this article mainly introduces and analyzes the image matching recognition, statistical pattern recognition, syntactic pattern recognition, fuzzy pattern recognition, artificial neural network pattern recognition and other image recognition.
In image preprocessing, the paper focuses on studying the image denoising and the image edge extraction. In the process of image denoising, the paper mainly studies the mean filter, median filter and wavelet filter. Through the comparison of several methods, we find a hybrid filter method which combined by median filter and mean filter has better filtering effect. In the image edge extraction of mechanical parts , after comparing several common edge detection operator, we find using Sobel operator for edge detection is fast and effective.
In the feature extraction, the paper mainly studies extracting the bolt feature based on the geometry characteristics of the bolt. The bolt characteristic vector which this paper presented has certain advantages. Such as , scale invariant, rotational invariance and translation invariance. In the recognition algorithm, based on the extracted feature vector of the bolt, the RBF network is used in the bolt identification. Simulation results show that the RBF network technology used in parts identification, with good recognition effect, and quick recognition rate.

Key words: feature extraction；parts identification；pretreatment；geometry characteristics；
RBF network

目錄
摘要 I
Abstract II
第1章 緒論 1
1.1 引言 1
1.2 機械零件圖像識別和檢測的研究現(xiàn)狀及發(fā)展趨勢 1
1.2.1 機械零件圖像識別和檢測的研究現(xiàn)狀 2
1.2.2 機械零件圖像識別和檢測的發(fā)展趨勢 3
1.3 機械零件圖像識別研究的內容 4
1.4 選題的研究背景和意義 5
1.5 本文的主要研..