基于细胞神经网络和并行压缩感知的 图像加密算法

基于细胞神经网络(CNN)和并行压缩感知(CS)提出了一种高安全性的非可视化图像加密算法, 旨在提高现有加密算法的信息传输效率以及减少存储空间。首先明文图像的小波系数经过阈值处理和索引置乱 后,利用受控的部分哈达玛矩阵对其进行并行压缩,接着执行费雪耶兹行列置乱和加模操作,然后再将部分加 密图像分割并通过最低有效位(LSB)嵌入算法随机地隐藏到剩余加密图像的 alpha 通道中生成最终的类噪声密 文图像,具有超混沌特性的 CNN 所产生的伪随机序列用于构造置乱、扩散以及受控测量矩阵。最后,通过一 系列的安全性分析表明,该算法具有很高的传输效率和安全性。     

基于大数据的产品族本体造型意象挖掘方法研究

为提升产品造型意象成族的准确性,增强感性工学研究中意象词汇提取的规范性和知识重用性。 首先从同族、泛族、异族的概念切入,对目标产品族的造型意象本体进行了定义;再借助 word2vec 工具和主 成分分析法(PCA)分别完成了对产品族意象词汇的关联性联想和降维提取,从而构建了意象词汇挖掘机制,通 过此机制可以辅助设计师利用网络大数据资源更高效、准确地进行目标产品意象的挖掘,一定程度上解决了传 统感性工学在意象挖掘方法中模糊性。最后结合意象词汇与造型特征的映射关系,利用 Protege 工具构建了产 品族造型意象的本体模型,对目标产品的造型意象知识进行了逻辑化表征,为下一代产品继承并发展原产品族 造型意象提供参考。以马自达 MX-5 车系的造型意象为例构建了本体模型,并进行了子代产品前脸的概念设计。     

图像去雾算法研究综述

图像去雾技术的目的是为了去掉图像中雾的影响,从而获得高质量的图像。本文主要从图像增强、图像复原和深度学习的角度归纳总结了图像去雾方法的研究状况,对暗通道先验等经典算法以及新活跃在去雾领域的几种深度学习去雾算法做了进一步的分析,并对各类算法的性能进行了总结,最后针对各类图像去雾方法指出了存在的问题及未来的展望。     

半稠密光流法的图像跟踪匹配算法

针对传统光流法图像特征缺失、边界及遮挡等处容易导致目标跟踪质量降低或丢失的问题,提出一种半稠密光流法实现图像特征的稳定跟踪。首先计算出图像中像素梯度变化较大的像素区域;其次利用时变图像灰度的时空梯度函数来计算像素的速度矢量,进而实现像素区域的跟踪;最后将状态向量作为剔除跟踪失败的依据,保留跟踪质量优良的像素区域。结果表明,该算法能有效地提高图像特征跟踪能力,保证图像中重要信息不丢失,同时保证运算速率。     

基于钻孔图像的结构面分布分维数计算方法

在工程实践中,对岩体结构面分布的分维数进行实测往往存在一定的难度,但是采用钻孔摄像技术既能够准确获取布设的钻孔内各项地质信息,而且形成的钻孔图像能够直观地反映钻孔中岩体结构面的特征参数,因此提出了一种基于钻孔图像的岩体结构面分布分维数计算方法。该方法的原理是:首先,通过钻孔图像获取研究区域内各钻孔内的岩体结构面的特征参数(包括倾角、倾向、隙宽、间距);然后,根据各钻孔之间的连通判断准则,判断区域内各钻孔结构面的连通性,并估算出研究区域内结构面的迹长值,同时参照钻孔的孔位及深度信息,来界定研究区域模型的范围;最后,结合Monte-Carlo法,建立岩体结构面分布的三维网络模型,根据分维数计算方法,求解模型各研究面的分形维数,并将3个钻孔的整体连通模型和各钻孔模型所得到的分维数开展对比分析。结果表明,该方法的计算结果与其他评价方法的结论一致。     

基于镜像源方法的室内声场脉冲响应仿真

针对基于室内声场脉冲响应描述声场音质特性时其精确度不高的问题,以经典镜像源方法为核心,讨论了基于变换频域和反射系数的改进算法.以矩形空间为具体实例,给定接收点和声源位置,并利用计算机仿真室内声场脉冲响应.在不同参数条件下,对各个算法进行了分析,估计了能量衰减曲线和房间混响时间.结果表明,基于变换频域和反射系数的改进算法具有更高的准确性和可靠性.     

基于新型Canny算法雷达距离图像边缘检测算法

针对激光雷达距离像的噪声滤波和边缘检测问题,提出了一种结合改进环圈滤波算法和自适应Canny算法的距离像边缘检测算法.通过改进环圈滤波对距离像的噪声进行抑制,在滤除噪声的同时保留了图像的细节特征信息.从梯度幅值计算、非极大值抑制和阀值自动选择3个方面对传统Canny算法进行改进,克服了噪声影响和边缘检测模糊等缺点.实验结果表明,该算法能够可靠的对距离图像进行边缘检测,并且检测结果的信息熵和标准差指标数据优于传统的Sobel算法和Laplace算法,能够较好地满足激光雷达距离像边缘检测的实际需要.     

Fast sensing method in compressive sensing with low complexity

The random sensing algorithms are hard for hardware implementation while the deterministic sensing algorithms have difficulty in acquiring large signals in the sensing systems of compressive sensing. This paper proposes a fast sensing method for compressive sensing with low complexity. The input signal is firstly permuted by an m-sequence controlled interleaving device. Then the permuted signal is transformed by the fast Walsh-Hadamard transform and down sampled to generate the measurements. Theoretical analysis indicates that the entries of the corresponding sensing matrices are asymptotically normally distributed. Simulation results show that the sensing performance of the corresponding matrices is almost the same as that of completely random sensing operators with a shorter computational time cost. The proposed method has good sensing performance and is easier for hardware implementation, which is meaningful in practice.     

Visible light localization algorithm based on multi-measurements constraints

In the indoor visible light transmission environment with complex noise, traditional received signal strength indication(RSSI)localization algorithm and imaging localization algorithm are unable to accurately determine the reason of the inaccurate localization model and inadequate constraints of position coordinates. In order to solve the problems, a visible light localization algorithm based on multi-measurements constraints is presented in the paper. The algorithm first constructs a joint localization state space model based on RSSI localization and the imaging localization algorithm. Then, states, measurements and their equations are constructed based on the space model. Finally, the equations are resolved using the unscented Kalman filter(UKF). Compared with RSSI localization and imaging localization, simulation results show that the RMSE of the proposed algorithm could be closer to the Cramer Rao bound. Besides, the localization probability of the proposed algorithm is higher than that of RSSI localization and imaging localization algorithm.     

New soft morphological parameterized symmetric logarithmic image processing filtering algorithm

Synthetic Aperture Radar (SAR) imaging has the inherent random speckle noise, which seriously affects the quality of the SAR image, and it is very important to study the suppression of the speckle noise. In view of the fact that the traditional spot noise suppression method could not remove speckle noise while protecting the image texture information, a novel filtering algorithm is proposed based on the soft morphological and parameterized symmetric logarithmic image processing model. The proposed algorithm combines the flexibility of the order-statistic soft morphological operations and the characteristic of the parameterized symmetric logarithmic image processing model which establishes a parameterized symmetric structure processing part of the image, which can suppress speckle noise while protecting the texture information on the image. In order to validate the despecking performance of the proposed algorithm, we compare the effect of the filter with some of the existing filtering algorithms, and use some of the image evaluation parameters of the suppress speckle noise image to evaluate the performance of the SMPSLIP filtering algorithm. Experimental results show the effectiveness of this algorithm.