烧成带状态的准确识别是回转窑烧结过程中最为关键的一环。给出了一种新颖的基于火焰图像的烧成状态识别方法。首先,基于一种新颖的设计方法得到的压缩Gabor滤波器组作为预处理阶段,增强具有不同纹理特性的物料区域和火焰区域的可分性。然后,对预处理后的火焰图像采用主成分分析寻找特征火焰图像,通过关联每幅火焰图像与特征火焰图像提取火焰图像的全局特征向量,最终经由概率神经元网络分类器对特征向量进行分类识别。实验结果表明了该方法的有效性。
Accurate recognition of burning state is considered to be the critical issue in sintering process control of rotary kiln.In this study,we propose a new method for burning state recognition with the goal of achieving more reliable state recognition.Firstly,A Gabor filter is employed as the pre-processing step to distinguish material and flame zones with distinguishable texture properties.In our study,a new approach is proposed to design an optimal compact filter bank.Then,from the filtered flame image database,eigen-flame images are found using principal component analysis,and global features are extracted through correlating each flame image with the eigen-flame images.Finally,the feature vectors are classified using probabilistic neural network pattern classifier.The proposed new method is validated through extensive experimental studies.