隐蔽光电目标的激光主动探测识别技术研究

发布时间：2018-05-08 14:31

本文选题：激光主动探测识别技术 + 猫眼效应　；参考：《中国科学院研究生院(长春光学精密机械与物理研究所)》2016年博士论文

【摘要】：随着针孔摄像头、微型摄像机等光电窃视侦察设备的普及,室内场所中进行的各类活动面临着日益严峻的信息泄露的威胁。泄露的敏感信息可能造成严重的军事、社会及商业损失。针对这一问题,可以利用激光主动探测识别技术对室内环境中隐藏的窃视侦察设备进行反侦察检测,排除潜在威胁。激光主动探测识别技术主要利用光电系统所具有的“猫眼效应”,通过目标回波信号的强度特征进行识别。然而,在室内环境中由于受较短的作用距离和高反射率干扰物等因素的影响,某些干扰物体也会对主动探测激光形成强度较高的反射信号。这些反射信号于某些情况下在强度特性上与真实目标所产生的回波信号十分接近。因此,仅仅依靠“猫眼效应”中的强度特征无法辨识真伪目标。本文深入研究了光电系统在激光主动探测下所具有的特征与特性,解决了在室内环境下应用识别技术的难点,并研制了一套手持操作式的微型相机检测设备。本论文针对真伪目标反射信号强度相近的问题,提出了一种局部纹理准则来描述“猫眼效应”中回波信号光截面的分布特征。并通过结合局部纹理准则和修改过的形态学准则,又提出了一种综合判定准则。该准则在复杂环境中的平均识别率为91%,与单独使用形态学准则和局部纹理准则相比,在平均识别率上分别提高了23%和21%。针对探测和识别过程实时性的问题,本文主要从算法优化上开展研究。本文提出了一种基于同心双窗模板匹配的连通域搜索算法,用于获取信号光斑在图像中的位置。该算法在搜索过程中引入形态学约束条件来排除在形态特征上明显异于真实目标的连通域,减少了区域标记的计算量,提升了算法速度。在研制设备上采用dsp+fpga的嵌入式系统架构,并对其进行了相应的优化。实验结果表明:上位机的平均计算时间为0.0804s,设备样机的平均用时为0.097s,达到实时性的要求。与最新的方法相比,本文算法在计算速度上是mspf算法并行模式的19.1~23.0倍;mspf算法串行模式的12.4~15.0倍;spcs算法的35.4~42.7倍;sfdc算法的63.7~76.9倍。针对便携式系统研制中的技术实用化问题,本文提出了一种基于目标稀疏性的自适应阈值分割算法,配合窄带滤光片设计,解决对视场内光电目标反射信号与背景信号之间的初步分割问题。该方法替代了传统的主被动图像差分策略,克服了主被动图像差分方法难以在手持操作式设备上应用的缺陷。本文基于在实验室环境中所进行的大量实验数据,针对本文提出的准则及相关算法的各方面的性能进行了分析和评价。上位机实验对算法的可行性、适应性、实时性和局限性等方面进行了研究。结果表明,在可行性方面,本文提出的算法在原理上的正确可行;在适应性方面,本文算法对外界光照强度和作用距离2.2~4.2m范围内具有鲁棒性。实时性方面已在前文中提到。局限性研究表明:当场景中包含高反射率的抛光球面干扰物或圆形金属干扰物体时,并且当他们的反射信号局部像素灰度值分布特性与真实目标类似时,本文算法失效。在对设备样机的试验中,本文设备成功地识别出完全隐藏在画框内的微型摄像机,该隐藏相机在不拆卸画框的情况下依靠人工检查难以发现,证明了设备在实际应用中的有效性和超越人工排查的优越性。本论文对隐蔽光电目标的激光主动探测识别技术在室内环境中的应用做出了开创性的工作;对用于识别技术的光电目标特征及度量准则方面做出了贡献;研制出了能够实际应用的原理样机,填补了国内该领域实用设备的空白;对光电目标主动探测识别技术的发展起到了促进作用。
[Abstract]:With the popularity of pinhole cameras, micro cameras and other photoelectric scouting equipment, all kinds of activities in indoor places are faced with the threat of increasingly severe information leakage. The sensitive information leaked may cause serious military, social and commercial losses. The hidden spy devices in the environment detect anti reconnaissance and eliminate potential threats. The laser active detection and recognition technology mainly uses the "cat eye effect" of the photoelectric system to identify the intensity characteristics of the target echo signal. However, in the indoor environment, the short range of action and the high reflectivity interferon are used in the indoor environment. With the effect of the element, some interfering objects will also form a high intensity reflection signal to the active detection laser. These reflected signals are very close to the echo signals produced by the real target in some circumstances. Therefore, the true and false targets can not be identified only by the strength characteristic of the "cat eye effect". This paper is deeply studied in this paper. The characteristics and characteristics of the photoelectric system under active laser detection have solved the difficulty of the application of recognition technology in the indoor environment, and developed a set of hand-held micro camera detection equipment. In this paper, a local texture criterion is proposed to describe the "cat eye effect" for the problem of the close intensity of the reflected signal of the true and false targets. By combining the local texture criterion and the modified morphological criterion, a synthetic criterion is proposed. The average recognition rate of this criterion in complex environment is 91%, compared with the use of morphological criteria and local texture criteria, the average recognition rate is increased by 23% and 21%, respectively. Aiming at the problem of real-time detection and recognition process, this paper mainly studies the algorithm optimization. In this paper, a connected domain search algorithm based on concentric double window template matching is proposed to obtain the location of the signal spot in the image. The algorithm introduces morphologic constraints in the search process to exclude the obvious morphological features. The connected domain, which is different from the real target, reduces the calculation amount of regional markings and improves the speed of the algorithm. The embedded system architecture of dsp+fpga is adopted on the development equipment, and the corresponding optimization is carried out. The experimental results show that the average computing time of the upper computer is 0.0804s, the average use of the device prototype is 0.097s, and the requirement of real time is achieved. Compared with the latest method, the algorithm in this paper is the 19.1~23.0 times of the parallel mode of MSPF algorithm, the 12.4~15.0 times of the serial mode of the MSPF algorithm, the 35.4~42.7 times of the SPCs algorithm, and the 63.7~76.9 times of the SFDC algorithm. In this paper, an adaptive threshold based on the target sparsity is proposed for the problem of the practical technology in the development of the portable system. The value segmentation algorithm, combined with the narrow band filter design, solves the problem of the initial segmentation between the reflected signal and the background signal in the field of view. This method replaces the traditional passive image difference strategy and overcomes the defect that the main and passive image difference method is difficult to apply on the hand-held operation. This paper is based on the laboratory environment. A large number of experimental data carried out in this paper are analyzed and evaluated in accordance with the criteria proposed in this paper and the performance of all aspects of the related algorithms. The upper computer experiment has studied the feasibility, adaptability, real-time and limitation of the algorithm. The results show that the algorithm proposed in this paper is correct and feasible in principle. In terms of adaptability, this algorithm is robust within the range of 2.2~4.2m and the range of external illumination. Real time has been mentioned in the previous article. The limitation study shows that when the scene contains high reflectivity polished spherical interfering or circular metal interfering objects, and when the local pixel gray value distribution of their reflected signals is distributed. When the characteristics are similar to the real target, this algorithm fails. In the test of the device prototype, the device successfully identifies the miniature camera that is completely hidden in the frame. The hidden camera is difficult to find by manual inspection without removing the picture frame, which proves the effectiveness of the equipment in the practical application and beyond the manual investigation. This paper makes a pioneering work on the application of the active detection and recognition technology for the concealed photoelectric target in the indoor environment, and makes a contribution to the characteristics of the photoelectric target and the measurement criterion used in the identification technology, and has developed the principle sample machine that can be applied in practice, and fills the blank of the practical equipment in this field. It has promoted the development of photoelectric target active detection and recognition technology.

【学位授予单位】：中国科学院研究生院(长春光学精密机械与物理研究所)
【学位级别】：博士
【学位授予年份】：2016
【分类号】：TP391.41

【参考文献】