基于rbf網(wǎng)絡(luò)的典型機(jī)械零件.doc
約74頁DOC格式手機(jī)打開展開
基于rbf網(wǎng)絡(luò)的典型機(jī)械零件,摘要在科技和工業(yè)不斷發(fā)展的同時(shí)，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(jī)器視覺技術(shù)對廢舊機(jī)械零件進(jìn)行識(shí)別分類研究，以便于以后再次循環(huán)利用。機(jī)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(xué)科，其核心內(nèi)容是圖像的處理和識(shí)別。機(jī)器視覺是現(xiàn)代制造的一個(gè)重要組成部分，在現(xiàn)...
內(nèi)容介紹
此文檔由會(huì)員 違規(guī)屏蔽12 發(fā)布
摘要
在科技和工業(yè)不斷發(fā)展的同時(shí)，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(jī)器視覺技術(shù)對廢舊機(jī)械零件進(jìn)行識(shí)別分類研究，以便于以后再次循環(huán)利用。機(jī)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(xué)科，其核心內(nèi)容是圖像的處理和識(shí)別。機(jī)器視覺是現(xiàn)代制造的一個(gè)重要組成部分，在現(xiàn)代機(jī)械行業(yè)中，為了提高生產(chǎn)的柔性和自動(dòng)化程度，機(jī)器視覺得到了廣泛的應(yīng)用。例如，機(jī)械零件的檢測和識(shí)別、產(chǎn)品分類以及自動(dòng)化生產(chǎn)的裝配檢測等。因此，利用機(jī)器視覺對機(jī)械零件進(jìn)行識(shí)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(jī)械零件的識(shí)別方法。本文的主要工作內(nèi)容包括圖像識(shí)別理論和方法的介紹與分析、機(jī)械零件圖像的預(yù)處理、圖像的特征提取、機(jī)械零件識(shí)別和典型零件識(shí)別系統(tǒng)的設(shè)計(jì)。
在圖像識(shí)別理論及方法方面，本文主要對圖像匹配識(shí)別、統(tǒng)計(jì)模式識(shí)別、句法模式識(shí)別、模糊模式識(shí)別和人工神經(jīng)網(wǎng)絡(luò)模式識(shí)別等圖像識(shí)別方法進(jìn)行介紹和分析。
在圖像的預(yù)處理方面，本文著重對圖像去噪和圖像邊緣提取進(jìn)行研究。在對圖像進(jìn)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(jìn)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(jié)合的混合濾波方法具有較好的濾波效果。在機(jī)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(jìn)行邊緣提取，具有速度快、效果好的優(yōu)點(diǎn)。
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(jìn)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(zhuǎn)不變性和平移不變性等優(yōu)點(diǎn)。在識(shí)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(luò)對螺栓進(jìn)行識(shí)別研究。仿真實(shí)驗(yàn)表明RBF網(wǎng)絡(luò)技術(shù)應(yīng)用于零件識(shí)別，具有較好的識(shí)別效果，且識(shí)別速度快。
關(guān)鍵詞：特征提??；零件識(shí)別；預(yù)處理；幾何形狀特征；RBF網(wǎng)絡(luò)
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(jī)械零件圖像識(shí)別和檢測的研究現(xiàn)狀及發(fā)展趨勢 1
1.2.1 機(jī)械零件圖像識(shí)別和檢測的研究現(xiàn)狀 2
1.2.2 機(jī)械零件圖像識(shí)別和檢測的發(fā)展趨勢 3
1.3 機(jī)械零件圖像識(shí)別研究的內(nèi)容 4
1.4 選題的研究背景和意義 5
1.5 本文的主要研..
在科技和工業(yè)不斷發(fā)展的同時(shí)，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(jī)器視覺技術(shù)對廢舊機(jī)械零件進(jìn)行識(shí)別分類研究，以便于以后再次循環(huán)利用。機(jī)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(xué)科，其核心內(nèi)容是圖像的處理和識(shí)別。機(jī)器視覺是現(xiàn)代制造的一個(gè)重要組成部分，在現(xiàn)代機(jī)械行業(yè)中，為了提高生產(chǎn)的柔性和自動(dòng)化程度，機(jī)器視覺得到了廣泛的應(yīng)用。例如，機(jī)械零件的檢測和識(shí)別、產(chǎn)品分類以及自動(dòng)化生產(chǎn)的裝配檢測等。因此，利用機(jī)器視覺對機(jī)械零件進(jìn)行識(shí)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(jī)械零件的識(shí)別方法。本文的主要工作內(nèi)容包括圖像識(shí)別理論和方法的介紹與分析、機(jī)械零件圖像的預(yù)處理、圖像的特征提取、機(jī)械零件識(shí)別和典型零件識(shí)別系統(tǒng)的設(shè)計(jì)。
在圖像識(shí)別理論及方法方面，本文主要對圖像匹配識(shí)別、統(tǒng)計(jì)模式識(shí)別、句法模式識(shí)別、模糊模式識(shí)別和人工神經(jīng)網(wǎng)絡(luò)模式識(shí)別等圖像識(shí)別方法進(jìn)行介紹和分析。
在圖像的預(yù)處理方面，本文著重對圖像去噪和圖像邊緣提取進(jìn)行研究。在對圖像進(jìn)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(jìn)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(jié)合的混合濾波方法具有較好的濾波效果。在機(jī)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(jìn)行邊緣提取，具有速度快、效果好的優(yōu)點(diǎn)。
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(jìn)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(zhuǎn)不變性和平移不變性等優(yōu)點(diǎn)。在識(shí)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(luò)對螺栓進(jìn)行識(shí)別研究。仿真實(shí)驗(yàn)表明RBF網(wǎng)絡(luò)技術(shù)應(yīng)用于零件識(shí)別，具有較好的識(shí)別效果，且識(shí)別速度快。
關(guān)鍵詞：特征提??；零件識(shí)別；預(yù)處理；幾何形狀特征；RBF網(wǎng)絡(luò)
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(jī)械零件圖像識(shí)別和檢測的研究現(xiàn)狀及發(fā)展趨勢 1
1.2.1 機(jī)械零件圖像識(shí)別和檢測的研究現(xiàn)狀 2
1.2.2 機(jī)械零件圖像識(shí)別和檢測的發(fā)展趨勢 3
1.3 機(jī)械零件圖像識(shí)別研究的內(nèi)容 4
1.4 選題的研究背景和意義 5
1.5 本文的主要研..
