基于细菌觅食算法的图像处理

发布时间：2018-02-13 20:46

本文关键词： 细菌觅食算法图像分割图像分类人脸识别参数优化　出处：《湖南工业大学》2017年硕士论文　论文类型：学位论文

【摘要】：随着计算机技术的不断发展以及其软硬件的更新换代,越来越多的人开始使用计算机对图像做各式各样的处理。图像处理技术也在迅速的发展,其应用范围也不断拓展,例如机器人视觉以及工业检测等。图像处理是一门引人注目和具有远大前景的学科,包括图像分割、图像增强、图像分类以及目标识别等多个技术领域。图像分割和图像分类这两方面内容如果处理的好,能对后续工作带来良好的结果;若是处理不好,这会对后续工作(如目标识别)带来不少的麻烦还有难点,然而许多常见的分割算法和分类算法可能只解决了部分的分割和分类问题,还余下大部分问题。基于仿生算法的分割和分类算法又可以进一步解决些问题。然而,传统的仿生算法也还是有些局限性,例如比较容易陷入局部最优,效率相对比较低下,搜寻精度不足等。所以本文提出了使用改进的细菌觅食算法来对图像进行分割和分类,主要研究工作如下:(1)提出了一种改进细菌觅食算法(IBFO)并应用于图像分割。针对传统细菌觅食算法在处理大量图像时效率较低和精度不高的问题,在算法的迁徙行为和趋化行为都做了动态的调整,根据所在阶段改变行为方式,并用改进算法对图像进行分割实验。通过实验验证,与基于仿生算法的聚类分析算法结果相比,并用测试函数收敛性、区域间灰度对比度、不均匀性和时间耗费指标来验证了改进细菌觅食算法的有效性。(2)提出了另一种改进细菌觅食算法(SIBFO)并应用目标图像分类。其主要利用改进细菌觅食算法对从视频中提取的行人、汽车以及宠物进行聚类分析完成图像中的目标分类。通过实验验证,与基于传统群体智能算法的目标分类结果相比,并用分类算法的指标查全率、查准率和综合这两个指标的加权指标来进行验证了改进细菌觅食算法的有效性。(3)将第二种改进细菌觅食算法(SIBFO)应用于支持向量机参数优化方法并进行人脸识别。根据支持向量机分类器的原理,将其与之前改进的细菌觅食算法结合起来以达到最优分类器的目的。然后使用结合后的分类器对人脸图像进行识别实验。通过实验验证,与基于传统仿生算法的支持向量机参数优化进行比较,在全局搜索方面、优化后的支持向量机的预测和误差分析方面以及对ORL人脸库和AR人脸库进行人脸识别的分类准确率方面考证了改进算法的优越性。
[Abstract]:With the continuous development of computer technology and the upgrading of its software and hardware, more and more people begin to use computers to do various kinds of image processing. For example, robot vision and industrial detection. Image processing is an attractive and promising subject, including image segmentation, image enhancement, Image segmentation and image classification can bring good results to the subsequent work if they are handled well. This will bring a lot of trouble and difficulties to the follow-up work (such as target recognition). However, many common segmentation algorithms and classification algorithms may only solve part of the segmentation and classification problems. However, the traditional bionic algorithm has some limitations, for example, it is easy to fall into local optimum, and the efficiency is relatively low. Therefore, an improved bacterial foraging algorithm is proposed to segment and classify images. The main research work is as follows: (1) an improved bacterial foraging algorithm (IBFOFOA) is proposed and applied to image segmentation. The migration behavior and chemotaxis behavior of the algorithm are dynamically adjusted. According to the stage, the behavior is changed, and the image segmentation experiment is carried out with the improved algorithm. The experimental results are compared with the results of the clustering analysis algorithm based on the bionic algorithm. The convergence of the function and the contrast between regions are tested. Inhomogeneity and time-wasting index to verify the effectiveness of the improved bacterial foraging algorithm. (2) another improved bacterial foraging algorithm (SIBFOO) is proposed and the target image is classified. The improved bacterial foraging algorithm is mainly used to extract pedestrians from the video. The target classification in the image is completed by clustering analysis of vehicles and pets. Compared with the target classification results based on the traditional swarm intelligence algorithm, the target recall rate of the classification algorithm is compared with that of the traditional swarm intelligence algorithm. The validity of the improved bacterial foraging algorithm is verified by the precision rate and the weighted index of these two indexes. The second improved bacterial foraging algorithm (SIBFOO) is applied to the parameter optimization method of support vector machine and face recognition is carried out. According to the principle of support vector machine classifier, It is combined with the previous improved bacterial foraging algorithm to achieve the purpose of the optimal classifier. Then the combined classifier is used to carry out the recognition experiment on the face image. Compared with the support vector machine parameter optimization based on the traditional bionic algorithm, in the aspect of global search, The prediction and error analysis of the optimized support vector machine (SVM) and the classification accuracy of ORL face database and AR face database are studied to prove the superiority of the improved algorithm.
【学位授予单位】：湖南工业大学
【学位级别】：硕士
【学位授予年份】：2017
【分类号】：TP391.41

【参考文献】