论文部分内容阅读
本文对现有产生纹理“Tuned”模板的方法进行了研究与分析,指出这种方法的缺点:解的随机性,井提出了改进意见,在模板的调整中引入变异操作,在分析的基础上,提出产生“Tuned”模板的新方法—遗传算法,使寻找最佳模板的随机性减小。通过航空影像的试验,与现有方法作对比,表明新方法的纹理分类效果有明显改善。此外,还对“Tuned”模板对纹理方向的自适应性和模板元素的对称性作了实验分析,得到一些有实用价值的结果。
In this paper, we study and analyze the existing methods of generating texture “Tuned ” template, pointing out the shortcomings of this method: the randomness of solutions, the improvement suggestions, the introduction of mutation operation in the template adjustment, Based on this, a new method of producing “Tuned” template - genetic algorithm is proposed, which reduces the randomness of finding the best template. Compared with the existing methods through the aerial image test, it shows that the texture classification effect of the new method has obviously improved. In addition, the experimental results on the adaptiveness of the “Tuned ” template to the texture direction and the symmetry of the template elements are also obtained. Some practical results are obtained.