舰载运动平台红外面阵图像复原技术研究与实现

发布时间：2019-04-02 13:45

【摘要】：舰载红外跟踪系统由于舰船受到海风海浪等扰动,致使红外面阵图像产生旋转和模糊现象,给指控中心的操作人员以及图像后续处理中弱小目标的检测、识别与跟踪都带来了一定的影响。另外,鉴于图像数据越来越大和图像处理算法越来越复杂等趋势,对图像处理平台的性能也提出了更高的要求。本文基于以上三个方面的问题,进行了红外图像消旋复原、红外图像去模糊复原、高性能信号处理板研制以及复原算法并行实现的研究。本文主要进行了以下研究工作:1)、推导出了舰载运动平台红外图像旋转角度。通过分析舰载红外跟踪系统以及图像发生旋转的原因,从空间几何角度定义了红外图像的旋转角度。并依据艏摇角、纵摇角、横摇角、方位角和俯仰角五个角度信息,在刚体绕定点旋转的物理学知识上,利用四元数法结合空间解析几何知识推导出了红外图像旋转角度。最后在旋转角度基础上进行图像消旋,复原出稳定的图像,缓解了红外图像的晃动程度,实现了电子稳像。2)、推导出了线性运动退化函数的运动长度和运动方向。分析和定义了舰载运动平台下,红外面阵图像序列的空间几何关系,在四元数和空间解析几何基础上,得到红外图像视场的视轴角度差和退化函数的运动方向。并根据红外面阵探测器的积分时间、瞬时视场和图像帧频等信息推导出运动长度。利用该方法推导的退化函数优点在于运动方向是任意的,而不局限于特定方向下的运动。3)、提出了将推导得到的线性运动退化函数作为EM算法的起始值,经过E步和M步的有限次迭代,得到了更加精细的退化函数。该方法提高了红外模糊图像去模糊复原的清晰度,并在实验中得到了验证。4)、针对图像去模糊的复原算法存在振铃效应的现象,提出了循环边界法抑制边界截断引起的边界振铃纹波;并利用改进的受限制自适应复原算法复原图像以及抑制棱边振铃效应引起的纹波。实验表明,该方法在有效抑制振铃效应的同时,突出了图像细节,图像复原效果更加清晰。5)、围绕图像数据的快速传输和图像复杂处理算法的高速处理两个方面的核心问题,研制了一块以TMS320C6678八核DSP为主要处理芯片的高性能信号处理板,并将红外图像的消旋复原算法和去模糊复原算法移植到具有高性能计算能力的八核DSP上。该移植工作充分考虑了并行技术的计算模型、各个内核之间的负载均衡以及核间的通信与同步等,提高了并行性和计算效率。依据图像复原算法流程和算法的计算量,设计了以数据流模型和主从模型的混合模型并行方式。实验测试验证了并行模型设计方案的可行性,基本达到各个内核负载的均衡化,为进一步的程序优化和工程实践奠定了良好的基础。
[Abstract]:Shipborne infrared tracking system causes rotation and blurring of infrared array images because the ship is disturbed by sea wind and waves, and detects weak and small targets in the operation of the command center as well as in the subsequent processing of the image. Recognition and tracking have brought a certain impact. In addition, in view of the trend of larger and larger image data and more complex image processing algorithms, the performance of the image processing platform is also put forward higher requirements. Based on the above three problems, the research of infrared image deconvolution, infrared image deblurring restoration, development of high performance signal processing board and parallel implementation of restoration algorithm are carried out in this paper. The main work of this paper is as follows: 1) the rotation angle of infrared image of shipborne moving platform is deduced. Based on the analysis of shipborne infrared tracking system and the reason of image rotation, the rotation angle of infrared image is defined from the angle of space geometry. According to the information of bow angle, pitch angle, roll angle, azimuth angle and pitch angle, the rotation angle of infrared image is deduced by using quaternion method combined with space analytic geometry knowledge on the physics knowledge of rigid body rotation around fixed point, and according to the five angles information of bow angle, pitch angle, roll angle, azimuth angle and pitch angle. Finally, the image is racemized on the basis of rotation angle, the stable image is restored, the degree of sloshing of infrared image is alleviated, and the electronic image stabilization is realized. 2) the motion length and direction of linear motion degradation function are deduced. On the basis of quaternion and spatial analytic geometry, the angle difference of the field of view and the direction of motion of the degradation function of infrared image are obtained on the basis of quaternion and spatial analytic geometry, and the spatial geometric relationship of infrared array image sequence under shipborne motion platform is analyzed and defined. According to the integration time, instantaneous field of view and frame rate of infrared array detector, the motion length is deduced. The advantage of the degradation function derived by this method is that the direction of motion is arbitrary and not limited to the motion in a specific direction. 3) the linear motion degradation function derived from this method is proposed as the initial value of the EM algorithm. After the finite iterations of E-step and M-step, a more precise degenerate function is obtained. This method improves the clarity of infrared image deblurring restoration and is verified in the experiment. 4) aiming at the phenomenon of ringing effect in the restoration algorithm of image de-blurring. The cyclic boundary method is proposed to suppress the boundary ringing ripple caused by boundary truncation. The improved limited adaptive restoration algorithm is used to restore the image and suppress the ripple caused by edge ringing effect. The experimental results show that this method can effectively suppress the ringing effect while highlighting the details of the image, and the image restoration effect is clearer (5). Focusing on the core problems of fast transmission of image data and high speed processing of complex image processing algorithms, a high performance signal processing board based on TMS320C6678 eight-core DSP is developed. The algorithm of deconvolution and de-blurring of infrared image is transplanted to the eight-core DSP with high computing capability. The model of parallel technology, load balancing among cores and communication and synchronization among cores are fully considered in this work, which improves the parallelism and computing efficiency. According to the flow of image restoration algorithm and the computational complexity of the algorithm, a hybrid model parallel mode based on data flow model and master-slave model is designed. Experimental tests verify the feasibility of the parallel model design scheme, basically achieve the load balance of each kernel, and lay a good foundation for further program optimization and engineering practice.
【学位授予单位】：中国科学院大学(中国科学院上海技术物理研究所)
【学位级别】：博士
【学位授予年份】：2017
【分类号】：TP391.41

【相似文献】