發(fā)布時間：2019-06-06 06:20

【摘要】：隨著生物神經(jīng)網(wǎng)絡(luò)系統(tǒng)的研究和智能化信息處理技術(shù)的發(fā)展,人工神經(jīng)網(wǎng)絡(luò)成為科學(xué)研究者們探索的熱點(diǎn).其中遞歸神經(jīng)網(wǎng)絡(luò)被廣泛地應(yīng)用在圖像處理、聯(lián)想記憶、平行計算、信號處理、模式識別等領(lǐng)域,因此關(guān)于遞歸神經(jīng)網(wǎng)絡(luò)的實(shí)際應(yīng)用研究很有現(xiàn)實(shí)意義. 本文提出了一種基于Cohen-Grossberg神經(jīng)網(wǎng)絡(luò)的圖像解密消噪的方法,彩色圖像在標(biāo)準(zhǔn)的RGB三色空間中表達(dá),圖像加密方式采用Arnold變換.Cohen-Grossberg神經(jīng)網(wǎng)絡(luò)將加密后的圖像數(shù)字矩陣作為網(wǎng)絡(luò)的平衡點(diǎn)進(jìn)行存儲,以實(shí)現(xiàn)解密前消除噪聲的功能.消除噪聲的加密圖像數(shù)字矩陣通過執(zhí)行正確的Arnold變換迭代次數(shù)實(shí)現(xiàn)解密.仿真實(shí)例驗(yàn)證了提出方法的有效性,實(shí)現(xiàn)了消除傳輸噪聲的功能. 另外本文還用訓(xùn)練單層前向神經(jīng)網(wǎng)絡(luò)的方法來實(shí)現(xiàn)Hopfield神經(jīng)網(wǎng)絡(luò)記憶存儲的功能,并將此Hopfield神經(jīng)網(wǎng)絡(luò)應(yīng)用在產(chǎn)品質(zhì)量分類中.仿真實(shí)例表明Hopfield遞歸神經(jīng)網(wǎng)絡(luò)的分類效果比較理想,可以為產(chǎn)品優(yōu)化、市場決策提供有效信息.
[Abstract]:With the research of biological neural network system and the development of intelligent information processing technology, artificial neural network has become the focus of scientific researchers. Recurrent neural network is widely used in image processing, associative memory, parallel computing, signal processing, pattern recognition and other fields, so the practical application of recurrent neural network is of great practical significance. In this paper, a method of image decryption and denoising based on Cohen-Grossberg neural network is proposed. The color image is expressed in the standard RGB trichromatic space. Arnold transform is used in image encryption. Cohen-Grossberg neural network stores the encrypted image digital matrix as the balance point of the network in order to eliminate noise before decryption. The encrypted image digital matrix which eliminates noise is decrypted by performing the correct number of iterations of Arnold transform. The simulation example verifies the effectiveness of the proposed method and realizes the function of eliminating transmission noise. In addition, this paper also uses the method of training single-layer forward neural network to realize the memory storage function of Hopfield neural network, and applies the Hopfield neural network to product quality classification. The simulation example shows that the classification effect of Hopfield recurrent neural network is ideal, and it can provide effective information for product optimization and market decision.
【學(xué)位授予單位】：天津大學(xué)
【學(xué)位級別】：碩士
【學(xué)位授予年份】：2012
【分類號】：TP333

【參考文獻(xiàn)】

相關(guān)期刊論文 前1條

1 廖曉昕,肖冬梅;具有變時滯的Hopfield型神經(jīng)網(wǎng)絡(luò)的全局指數(shù)穩(wěn)定性[J];電子學(xué)報;2000年04期

，

本文編號：2494134