材质感知与识别及其在大洋矿物分类中的应用

项目名称： 材质感知与识别及其在大洋矿物分类中的应用

项目编号： No.61501417

项目类型： 青年科学基金项目

立项/批准年度： 2016

项目学科： 无线电电子学、电信技术

项目作者： 亓琳

作者单位： 中国海洋大学

项目金额： 21万元

中文摘要： 实物表面都由一定的材质构成，人类可以通过视觉信息识别材质种类（比如金属或塑料）并评估其属性。材质感知与识别是视觉科学与计算机视觉领域的重要组成部分，然而目前对材质感知机理和材质图像识别算法的研究与物体识别研究近年来取得的长足进步相比，却十分滞后。现有的材质识别系统使用的是适用于物体识别的图像特征，这些特征并不能很好地描述材质的属性。本项目在反向光学和统计外观模型理论的基础上，使用眼动仪进行大量心理物理学实验，研究感知数据与参数化的成像因素（光照、表面形状）和分解后的图像信息（颜色、轮廓、边缘、高光反射等）之间的关系，建立相应的感知模型。在感知数据的启发下对相应的图像信息进行分析，从而设计符合视觉感知的、能够描述材质属性的图像特征。由于大洋矿物样本的特殊性，材质是对其进行分类的重要参考指标，本项目研究的材质感知与识别方法，将用于大洋矿物样本的分类和相关属性的数值量化。

中文关键词： 图像信息；图像特征；图像特征提取；图像特征分析

英文摘要： Object surfaces are all made of certain materials. Humans can recognize various materials (such as metal or plastic) and estimate corresponding properties from visual information. Material perception and recognition is important component in the research field of vision science and computer vision. However, there is less investigation in the mechanism of material perception and the algorithms for material recognition, in contrast to the recent bloom in object recognition. Current material recognition systems utilize image features that were particularly designed for object recognition tasks, which may not work well for identifying materials. Based on the theory of inverse optics and statistical appearance model, the proposed project will conduct massive psychophysical experiments using the eye tracker. We will study how observers perform in perceiving materials with varying imaging factors (such as illumination and surface shape) and image information (such as color, contour, edge, specular highlights and so on). We will build perceptual models based on the relationship between the physical properties and the perceptual data. With the analysis of the corresponding image information, we will design perception inspired image features that are consistent with visual perception and capable to describe material properties. The specificity of the ocean minerals determines that the material properties are critical reference characteristics. The method of material perception and recognition proposed in this project will be applied in classification of ocean minerals and quantitative estimation of corresponding properties.

英文关键词： Image Information;Image Feature;Image Feature Extraction;Image Feature Analysis