(中南大學(xué) 信息科學(xué)與工程學(xué)院，長沙 410083)
摘 要: 針對硫浮選泡沫圖像噪聲大、特征重要度差異顯著引起工況難以識別的問題，提出基于模糊支持向量機(jī)的硫浮選工況識別方法。通過融合樣本模糊隸屬度和特征信息增益，獲取圖像視覺特征的特征重要度；并結(jié)合特征重要度矩陣，改進(jìn)模糊支持向量機(jī)的核函數(shù)，進(jìn)而建立工況類別與圖像特征之間的關(guān)系模型，實(shí)現(xiàn)硫浮選工況識別。采用模糊隸屬度對噪聲賦予較小的權(quán)值，并結(jié)合模糊隸屬度來獲取特征重要度矩陣，可以減小噪聲樣本的影響，以揭示圖像特征重要度之間的差異，提高工況識別準(zhǔn)確性。鋅直接浸出冶煉硫浮選生產(chǎn)過程的實(shí)際測試數(shù)據(jù)驗(yàn)證了方法的有效性。
關(guān)鍵字: 硫浮選；特征重要度；模糊支持向量機(jī)；工況識別
（School of Information Science and Engineering, Central South University, Changsha 410083, China）
Abstract:Considering performance recognition problem caused by the high noise of froth images and the obvious difference of feature importance in sulfur flotation process, a performance recognition method for sulfur flotation process using fuzzy support vector machine was proposed. With the combination of fuzzy membership and feature information gain, the image feature importance was obtained, and the kernel function of fuzzy support vector machine was improved using the feature importance. Then, the model that reveals the relationship between performance and image feature was established to detect sulfur condition. As the fuzzy membership was used to define a small weight for the noise sample and acquire feature importance, which can reduce the effect of image noise points and reveal the difference of feature importance, the classification accuracy is effectively improved. The simulation results show the effectiveness by using actual running data from a sulfur flotation process of zinc direct leaching hydrometallurgy.
Key words: sulfur flotation; feature importance; fuzzy support vector machine; performance recognition
