Fast and robust face recognition via coding residual map learning based adaptive masking

Robust face recognition (FR) is an active topic in computer vision and biometrics, while face occlusion is one of the most challenging problems for robust FR. Recently, the representation (or coding) based FR schemes with sparse coding coefficients and coding residual have demonstrated good robustness to face occlusion; however, the high complexity of l₁-minimization makes them less useful in practical applications. In this paper we propose a novel coding residual map learning scheme for fast and robust FR based on the fact that occluded pixels usually have higher coding residuals when representing an occluded face image over the non-occluded training samples. A dictionary is learned to code the training samples, and the distribution of coding residuals is computed. Consequently, a residual map is learned to detect the occlusions by adaptive thresholding. Finally the face image is identified by masking the detected occlusion pixels from face representation. Experiments on benchmark databases show that the proposed scheme has much lower time complexity but comparable FR accuracy with other popular approaches.     

General triangular midpoint subdivision

In this paper, we introduce triangular subdivision operators which are composed of a refinement operator and several averaging operators, where the refinement operator splits each triangle uniformly into four congruent triangles and in each averaging operation, every vertex will be replaced by a convex combination of itself and its neighboring vertices. These operators form an infinite class of triangular subdivision schemes including Loop's algorithm with a restricted parameter range and the midpoint schemes for triangular meshes. We analyze the smoothness of the resulting subdivision surfaces at their regular and extraordinary points by generalizing an established technique for analyzing midpoint subdivision on quadrilateral meshes. General triangular midpoint subdivision surfaces are smooth at all regular points and they are also smooth at extraordinary points under certain conditions. We show some general triangular subdivision surfaces and compare them with Loop subdivision surfaces.     

二调GIS资源的商业化应用研究

不管采用何种方式的商业化地图,其测量精度均无法与二调GIS地图精度相媲美。探讨了在二调GIS图即将被一普GIS图替代的情形下,将二调GIS图用于民用商业化市场的可行性和难点。最后得出,二调GIS资源进行商业化运作具有可行性,但同时也具有一定的难度。     

Euclidean Distance Mapping for computing microstructural gradients at interfaces in composite materials

This paper presents an image analysis procedure using Euclidean Distance Mapping to compute microstructural gradients at interfaces in composite materials. This method is capable of producing phase distribution plots at single pixel strip width very quickly and efficiently. Compared to conventional dilation-subtraction strip analysis, the new method is faster, more flexible and is not constrained by feature geometry and boundary conditions. This allows for truly random and unbiased sampling. The new method was applied to investigate microstructural gradients at the interfacial transition zone (ITZ) of an ordinary Portland cement concrete. The average results show strong gradients in anhydrous cement and detectable porosity at the ITZ, but this is highly variable from location to location. The overall ITZ characteristics depend on the amount of calcium hydroxide deposited on aggregate particles. The new method was able to measure the effect of these calcium hydroxide deposits on the porosity gradient, which has not been reported before.     

基于PCA的高维多目标优化可视化方法

高维多目标优化问题的高维解集由于目标和解的个数众多,对其可视化较为困难。针对上述问题,结合降维和非降维数据分析技术,提出一种高维多目标优化的可视化方法。该方法对高维多目标算法运行后的一组解集进行预处理,运用主成分分析方法分析数据特征,获取转换后的数据及其对应的贡献率。按照贡献率由大到小的顺序调整转换后的数据列顺序;利用主成分贡献率求解转换后数据的行间距离,运行分级聚类算法并对转换后的数据按行排序,重新组织数据,将最终的结果用热图显示。实验结果表明,该方法既能使用户明确转换后每个目标所占的贡献率,又能取得较满意的视觉效果,便于用户理解数据的整体分布并做出决策。     

多交叉混沌选择反向小生境遗传算法

为提高小生境遗传算法的全局以及局部搜索能力,提出一种多交叉混沌选择反向小生境遗传算法。利用分段线性混沌映射函数生成一组混沌数序列,在每次进行交叉操作前,依据序列中对应元素的数值大小选择不同的交叉算子进行操作,通过小生境遗传算法产生较优的子代种群。针对子代种群,应用反向搜索策略获得反向种群,在子代种群和反向种群中进行精英选择得到最终新种群,以进一步加强算法的局部寻优能力。仿真实验结果表明,该算法在最优解及均值方面好于小生境遗传算法,从而证明其可行性和优越性。     

基于资源三号卫星数据的数字正射影像图制作研究

资源三号卫星的成功发射填补了我国立体卫星测绘的空白,可以提供优于5m地面分辨率的遥感影像.论文提出基于资源三号卫星遥感数据的数字正射影像图（DOM）制作流程,重点对影像纠正、影像配准、影像融合、影像镶嵌等关键步骤进行研究,并通过实验验证方法的可行性.实验证明该方法具有较强的实用价值和广泛的应用前景.     

单幅图像多尺度小波深度提取算法

针对浅景深图像中平滑前景区域深度提取误差大的问题,基于像素点分类思想对深度值进行修正,提出一种基于多尺度小波线索的、可同时面向单幅浅景深图像和广角图像的深度图提取算法.首先使用小波分析法在多个尺度下提取图像深度信息;然后提出自适应分类法并根据尺度与深度变化规律对像素点做深度修正,得到深度图;最后结合区域生长与边缘分割算法对深度图进行区域优化.为了加快深度计算,还提出了快速zerocount法以及多尺度加速法来满足标清视频实时处理要求.实验结果证明,采用文中算法获得的深度图相对深度正确,前景和背景区域深度一致性好.     

基于三维点云模型的特征线提取算法

针对以往算法存在无法区分尖锐和非尖锐特征点、提取的特征点与视角有关、特征点未连线等问题, 提出一种基于高斯映射和曲率值分析的三维点云模型尖锐特征线提取算法。该算法先进行点云数据点的离散高斯映射, 并将映射点集聚类; 然后使用自适应迭代过程得到两个或多个面的相交线上曲率值和法向量发生突变的尖锐特征点, 这些点与视角无关; 最后, 用改进的特征折线生长算法, 将特征点连接, 得到光顺特征线。实验证明, 该算法具有良好的自适应性、抗噪性和准确性, 是一种有效的三维模型特征线提取算法。     

基于nRF24LE1和百度地图的校园警情定位系统设计

在手机报警受限的情况下,为了使求助者能向校园保卫处紧急报警并且准确定好自己所处的位置,本文研制了求助者随身携带的微型无线遥控器和相应的配套系统。当求助者按下紧急求助按钮时,立即向保卫处发出求助信号,同时在保卫处的监控界面上直观显示求助者所在的位置。测试表明,从求助者按下紧急求助按钮到保卫处页面刷新显示为40~50s,位置精度为20~40m。