基于人眼视觉的航空图像检索仿真

Aviation Image Retrieval and Simulation Based on Human Visual and Sensation

研究航空图像准确检索的问题。航空图像特定区域中的目标一般呈随机性分布,而且干扰目标与识别目标交错分布。传统的图像检索算法的思路都是以目标为线索,根据特定目标的像素特征或者关联特征,来完成图像的定位。同时,干扰排除也以特征对比为主,较为机械,并且目标众多环境下,导致排除过程极其复杂,效率与定位准确性都很低。为了避免上述缺陷,提出了一种基于人眼视觉感知的航空图像目标检索方法。方法利用局部二进制方法进行航空图像的特征提取,再将上述不同种类的特征作为人眼视觉感觉航空图像检索的基础数据,然后建立人眼视觉模型,进行航空图像检索。实验结果表明,利用改进算法进行航空图像检索,可以极大地提高检索的查准率和查全率,也为不同领域的航空图像应用提供基础。

In this paper, accurate retrieval of aviation image was researched. The target in aviation image specific areas generally show random distribution, interference objectives and recognized target are distributed in cross. The traditional image retrieval algorithms are based on the idea of goals for clues, according to the specific character- istics of the target pixel or related features, to complete the image positioning. Meanwhile, the disturbance rejection is also based on features contrast, more automated, under the circumstances of many targets, resuh in the extreme complex of exclusion process, efficiency and positioning accuracy is very low. To avoid these shortcomings, we proposed an aviation image object retrieval approach based on human visual perception. The method used local binary method to extract aviation image features, and the characteristics of the different types of human visual sense was used as essential data of aviation image retrieval, then create the human visual sensation model for aviation image retrieval. Experimental results show that the improved algorithm for image retrieval aviation can greatly improve the retrieval precision and recall, and for applications in different fields to provide the basis for aviation images.