针对电容层析成像反问题流型识别较难的问题,提出了一种新的ECT流型辨识算法——差分演化优化极端学习机算法,进而提出了基于自适应差分演化优化极端学习机(SaDE-EML)的ECT辨识算法.在论述极端学习机算法的基础上,结合差分演化算法对极端学习机算法进行优化,自适应差分演化算法中的关键参数,通过训练得到各类流型的分类器的参数,构造分类器进行精准与快速分类.实验结果表明:该算法能有效克服极端学习机算法的缺点并提高了局部与全局收敛能力,通过与BP、SVM算法比较,该算法具有竞争力,并为电容层析成像流型辨识的研究提供了新算法.

To solve the flow pattern identification difficulty in electrical capacitance tomography( ECT),a gaussian mixture model( GMM) flow pattern identification algorithm for electrical capacitance tomography system is presented. On the basis of Gaussian mixture model( GMM) and the principle of EM algorithm,the Kmeans algorithm should be united in wedlock. Then we get the parameter of gaussian mixture model( GMM) through training of electrical capacitance tomography flow pattern,establish a classifier of gaussian mixture model( GMM) to achieve the goal of faster identification of five flow patterns. The experimental result shows that the algorithm has higher identification accuracy than the algorithm of neural network and support vector machine and decision tree. The algorithm provides a new idea for the research on the electrical capacitance tomography identification algorithm.

On account that density estimation of Mean-Shift algorithm exists the shortcoming of losing the target under the influence of speed,light and other factors,the paper proposes a tracking methods combining target color histogram and mean-shift iterative algorithm on the basis of the basic principle of Mean-Shift algorithm.This paper applies the algorithm to plate tracking and researches the feasibility and efficiency of this algorithm under the influence of obstructions and light.The simulation and experimental results show that the algorithm has the advantages of faster tracking speed,higher recognition accuracy and stability than the traditional Lucas-Kanade algorithm and Camshift algorithm.The algorithm provides a new thought for the research on the plate tracking algorithm.

In order to solve the problem of difficulty in plate texture classification,the paper proposes the plate texture classification algorithm based on Kmeans-GMM model. Based on the Gaussian mixture model( GMM) and the principle of parameter estimation algorithm,this paper adopts the gray level co-occurrence matrix to extract plate texture feature,and estimates the parameters about the Gaussian mixture model( GMM) combining with Kmeans algorithm through training to realize the plate texture classification. The experimental results show that the algorithm has the advantages of faster speed and higher accuracy of recognition than the traditional neural network algorithm and SVM algorithm. The algorithm provides a new thought for the research on the plate texture classification algorithm.