淺究MATLAB神經(jīng)網(wǎng)絡(luò)工具箱中的BP人工神經(jīng)網(wǎng)絡(luò)

全文37頁(yè) 約14000字 論述翔實(shí)

摘 要
為了客觀評(píng)估豬肉各項(xiàng)指標(biāo)和豬肉等級(jí)，本論文采用MATLAB神經(jīng)網(wǎng)絡(luò)工具箱中的BP人工神經(jīng)網(wǎng)絡(luò)，根據(jù)豬胴體圖象特征參數(shù)和活體豬圖象特征參數(shù)建立BP神經(jīng)網(wǎng)絡(luò)模型。本文對(duì)樣本進(jìn)行了訓(xùn)練，由訓(xùn)練后的BP神經(jīng)網(wǎng)絡(luò)模型進(jìn)行仿真。將仿真結(jié)果與人工評(píng)估結(jié)果進(jìn)行對(duì)比，并對(duì)采用不同的BP神經(jīng)網(wǎng)絡(luò)隱含層的傳遞函數(shù)和隱含層神經(jīng)元數(shù)目對(duì)仿真結(jié)果及訓(xùn)練目標(biāo)誤差的影響進(jìn)行比較。結(jié)果表明，BP人工神經(jīng)網(wǎng)絡(luò)模型可以評(píng)估豬肉各項(xiàng)指標(biāo)和等級(jí)識(shí)別。在豬肉胴體圖象特征指標(biāo)下評(píng)價(jià)豬肉等級(jí)準(zhǔn)確率達(dá)到98%，在活體豬圖象特征參數(shù)評(píng)價(jià)豬肉等級(jí)準(zhǔn)確率達(dá)到80%。說明豬肉胴體圖象特征比活體豬圖象特征參數(shù)更能代表豬肉質(zhì)量品質(zhì)也符合客觀現(xiàn)實(shí)。同時(shí)也表明MATLAB神經(jīng)網(wǎng)絡(luò)工具箱中的BP人工神經(jīng)網(wǎng)絡(luò)可以應(yīng)用在豬肉等級(jí)識(shí)別上。

關(guān)鍵詞：BP神經(jīng)網(wǎng)絡(luò),MATLAB,豬肉等級(jí)

SUMMARY
In order to eva luate various pork indexes and pork grade impersonally,this paper adopted BP artificial nerve network in the tool box of MATLAB nerve network,and estabished the BP nerve network model on the basis of analysis of the pig body’s characteristic parameters and live pig image parameters.The samples were trained,then the trained BP nerve network model was emulated.The emulated result and manual eva luated result were compared.Further more,effects on the emulated result and trained target error were compared.
The result expresses, the BP artificial nerve network model can take the gauge of pork various index signs and eva luate the pork grade. The accurate rate in grade on the basis of pig body’s characteristic parameters is 98%,while the accurate rate in grade on the basis of live pig image parameters is 80%.As a result, live pig image parameters is more practical to show the quality of pork.It also showed that BP artificial nerve network in the tool box of MATLAB nerve network can be applied in pork grade idendification.

Keywords:BP nerve network; MATLAB ;pork grade

目 錄
第一章 緒 論 1
1.1 神經(jīng)網(wǎng)絡(luò)的發(fā)展 1
1.2 人工神經(jīng)網(wǎng)絡(luò)的原理概述 1
1.3 問題的提出和課題的意義 2
1.4 課題的研究?jī)?nèi)容和目標(biāo) 3
第二章 可行性研究 4
2.1 背景 4
2.2 可行性研究前提 4
2.3 技術(shù)方面的可行性 5
第三章 系統(tǒng)需求分析 8
3.1 系統(tǒng)概述 8
3.2 功能概述 8
3.3 數(shù)據(jù)輸入 8
3.4 系統(tǒng)輸出 9
3.5 硬件、軟件、運(yùn)行環(huán)境 9
第四章 系統(tǒng)概要設(shè)計(jì) 10
4.1 數(shù)據(jù)類型 10
4.2 輸入數(shù)據(jù)預(yù)處理 11
4.3 數(shù)據(jù)輸入輸出方法 11
4.4 處理流程和數(shù)據(jù)流程 12
4.5 MATLAB中有關(guān)神經(jīng)網(wǎng)絡(luò)的重要函數(shù) 13
4.6 BP網(wǎng)絡(luò)的設(shè)計(jì)分析 13
第五章 詳細(xì)設(shè)計(jì) 15
5.1 根據(jù)胴體豬肉圖象進(jìn)行等級(jí)仿真的實(shí)現(xiàn) 15
5.2 根據(jù)活體豬圖象特征值進(jìn)行仿真的實(shí)現(xiàn) 19
5.3 不同隱含層傳遞函數(shù)和隱含層神經(jīng)元個(gè)數(shù)對(duì)BP網(wǎng)絡(luò)影響研究 27
第六章 用戶手冊(cè) 32
6.1 數(shù)據(jù)輸入方式 32
6.2 網(wǎng)絡(luò)訓(xùn)練和網(wǎng)絡(luò)仿真 32
6.3 數(shù)據(jù)輸出 32
第七章 總 結(jié) 33
致 謝 34
附 錄 參考書目 35

部分參考文獻(xiàn)
[14] MATLAB 7.0實(shí)用指南，蘇金明, 王永利 編著，電子工業(yè)出版社，2004
[15] MATLAB 6數(shù)學(xué)建?；A(chǔ)教程，云舟工作室 編著，人民郵電出版社，2001
[16] MATLAB神經(jīng)網(wǎng)絡(luò)設(shè)計(jì)，陳聞新 周露 王丹力 熊曉英 編著，科學(xué)出版社，2000.9
[17]A cerebellar neural network model for adaptative control of saccades implemented with Matlab;Rodriguez Campos, Francisco A. (Biomedical Engineering, University of Connecticut); Enderle, John Source: Bioengineering, Proceedings of the Northeast Conference, 2003, p 5-10
[18] Neural network modeling for dynamic pulsed GTAW process with wire filler based on MATLAB;Zhao, Dongbin (Department of Mechanical Engineering, Tsinghua University); Chen, Shanben; Wu, Lin; Chen, Qiang Source: China Welding (English Edition), v 10, n 2, November, 2001, p 88-93