节点文献
全景图构建及信息加密在公共安全中的研究与应用
Research on Panoramic&Information Encryption and its Application in Public Security
【作者】 王帅;
【导师】 孙伟;
【作者基本信息】 中国矿业大学 , 控制理论与控制工程, 2014, 博士
【摘要】 面对高级别风险区域日益严峻的安全管控形势,全景安检及图像加密技术成为迫切需求。“全景图构建及信息加密在公共安全中的研究与应用”是将全景图构建理论与技术、信息安全理论与技术相结合,研究安检图像碎片全景构建、图像加密的理论与应用,具有重要的理论意义和实用价值。首先,安检全景图构建主要包括图像预处理、图像配准和图像融合,图像配准是全景图构建的基础和关键。分析安检图像的采集方法及安检碎片的特点,对图像预处理、图形配准、图像融合问题作了深入研究;针对SIFT描述子构建过程复杂、描述向量维度高、需要计算特征点附近邻域图像主方向等问题,本文提出一种改进的SIFT算法,利用圆形的旋转不变性把特征描述子的维数从128维降为48维,安检图像及标准测试图库比较试验可知,改进算法提高了特征点生成速度,减少了特征匹配的运算量;在特征点匹配方面,本文提出一种基于模长的匹配算法,步骤为限幅、归一化、计算相似匹配对模长的差值,通过限定差值确定图像间的匹配对,相比SIFT常用的遍历搜索NN算法,算法复杂度低,匹配效率高;在图像融合方面,本文提出一种基于亮度调制的非线性加权融合算法,有效抑制了融合鬼影及拼缝的产生。其次,阐述了图像加密环节的重要性及性能指标,分析了混沌基础理论、密码学基础理,并进行了两方面研究:(1)对一类传统的Logistic映射的混沌加密算法分析。分析算法原理,针对安检图像设计算法步骤,并进行算法安全性及算法效率测试,实验分析可知,该算法局部性能较好,但总体效果不佳,不适合安检图像加密。(2)提出一种多混沌系统图像置乱扩散加密算法。该加密算法分别通过Logistic映射和Kent映射产生置乱和扩散控制参数,并分别以Arnold映射和模加的方式实现图像置乱和灰度扩散。试验结果表明,改进的多混沌加密算法,密钥空间大且密钥敏感性强,能有效的抵挡穷举分析攻击;加密后像素灰度分布均匀、相邻像素点相关性很弱,能有效抵挡统计分析攻击;像素变化比率和灰度值平均改变程度在加密轮次大于等于2时同时达到理想值,抗差分攻击能力很强;较传统加密算法,加密效率高,实时性强。最后,针对公共安全非接触式安检中全景安检及图像加密问题,结合本文提出算法及方法,进行“全景安检及图像加密系统”设计,给出智能车非接触式采集方案,在此基础上实现了全景安检及图像加密,并进行了应用测试。设计的试验样机,具有重要的理论意义和应用前景。
【Abstract】 Due to the security situation been more and more serious, especially in High-levelrisk areas, panoramic and encryption technology has become an urgent requirement.Dissertation “Research on Panoramic&Information Encryption and its Application inPublic Security”, panorama theory and information security technology are combined,and the security scene mosaic and the image encryption theories are used in securityimage fragments. It has important theoretical significance and practical value.Firstly, the key steps in Panorama mosaic are mainly image preprocessing, imageregistration and image fusion. The second step is the base and key of mosaictechnology. The paper studies above steps about special application in security imagefragments. SIFT is a kind of feature point registration algorithm. It remains invariantfor scale transformation, rotating, brightness change, and good stability. So it iswidely used in image registration. But the feature description vectors in SIFT is128dimensions, which causes large computation complexity of corresponding matchingalgorithm has and lower efficiency. In this paper, an improved SIFT algorithm isproposed to reduce the dimension from128d to48d by using circular rotationinvariance. It improves the extraction speed of feature point and reduces thecomputation complexity of feature matching. To match the feature point, a matchingalgorithm based on modulus is used, which includes limiter, normalization, andmodulus calculation. And then the difference with the similarity matching modulus iscalculated so as to determine matching pairs through limiting difference. Comparedwith normal NN algorithm, the computation complexity of this method is decreasedand matching efficiency is improved. To Image fusion overlap region, a nonlinearweighted fusion algorithm is used, which can effectively eliminate the patchwork andghosting and ensures consistency of image brightness.Secondly, the importance and the performance indexes of image encryption areshown. The fundamental knowledge of cryptography and chaos theory as well aschaotic cryptography and its development are analyzed. Then the following two kindsof encryption algorithm are compared:(1) A traditional image encryption algorithmbased on Logistic chaotic system is given. By analysis algorithm principle, designalgorithm steps and test the safety and efficiency in security image, the analysis andtests show that the local performance is good whereas the overall effect is not suitable for encryption of the system.(2) Image encryption algorithm of scrambling anddiffusion based on multiple chaotic systems is proposed. The control parameters ofscrambling and diffusion are produced by Logistic and Kent mapping. The scramblingand diffusion of the gray value are realized by Arnold mapping and the modularaddition. The tests results indicate that this algorithm meets the basic security andreal-time requirements of the system. The experiment results show that, the improvedmulti-chaotic encryption algorithms has large key space and key sensitivity, whichcan effectively resist the exhaustive analysis of the attack; encrypted pixel graydistribution, weak correlation of adjacent pixels can effectively statistical analysisresist attack. Alternated2generations, the NPCR and UACI can achieve the desiredfor anti-differential attacks. Compared with the traditional encryption algorithms, theimproved multi-chaotic encryption algorithm has high encryption efficiency.At last, in order to solve the problems that panoramic inspection and imageencryption application in security protection,"panoramic image of security andencryption systems" was designed. It included hardware acquisition of smart car andpanoramic image of security and encryption software. The application testing showthat, experimental prototype have important theoretical significance and applicationvalue.
【Key words】 panoramic inspection; image fusion; image encryption; SIFT; multiplechaotic systems;