基于PICO384的图像处理与跟踪算法研究

发布时间：2018-06-25 15:20

本文选题：PICO384 + 红外图像增强　；参考：《南京理工大学》2017年硕士论文

【摘要】：红外成像导引技术以其全天时、全天候、精度高、隐蔽性好和抗干扰能力强等优点被广泛应用于精确制导和导航等领域,其中红外目标跟踪是红外导引头的关键技术之一。本文围绕PIC0384非制冷型红外焦平面实现了基于FPGA的红外图像实时预处理算法,以便于在后续跟踪系统中进一步提取目标特征,并在此基础上实现单一红外目标的实时跟踪。本文首先介绍了非制冷红外成像技术和红外目标跟踪技术的发展概况和应用前景,阐述了非制冷红外探测器原理,红外图像特点,图像处理算法以及红外目标跟踪技术,在上述理论基础上搭建基于PIC0384红外焦平面的红外图像处理和目标跟踪硬件平台。本文在已有的384×288红外成像系统电子组件的硬件基础上,实现了盲元的综合检测补偿和连续均值量化变换增强等实时红外视频处理算法。其中均值量化变换增强算法与直方图均衡化算法相比能在提高图像对比度的同时保留更多的图像细节,有效防止过增强和"斑块"等传统增强算法中容易出现的问题。在获得清晰实时红外图像后添加基于FPGA的单目标跟踪系统,通过LVDS(低压差分信号)传输技术将红外视频数据从成像系统传输到目标跟踪系统。在对基于Shi-Tomasi特征点的TLD跟踪算法和基于SIFT特征点的匹配跟踪算法进行理论仿真和验证的基础上,结合FPGA的特性,提出一种基于改进SIFT特征点的光流跟踪算法并硬件实现。采用DoG(高斯差分)算子检测特征点,运用局部非极大值抑制方法去除SIFT特征点的边缘响应,利用简化的SIFT描述子,通过光流算法获得目标的位移与尺度,采用中值算法叠加跟踪窗口,实现实时精确跟踪,并给出了硬件实现步骤和相关处理效果。
[Abstract]:Infrared imaging guidance technology is widely used in the fields of precision guidance and navigation, such as all weather, all-weather, high precision, good concealment and strong anti-jamming ability, in which infrared target tracking is one of the key technologies of infrared seeker. In this paper, a FPGA based infrared image is realized around PIC0384 uncooled infrared focal plane. In order to facilitate the further extraction of target features in follow-up tracking system, the real-time tracking of single infrared target is realized on this basis. First, the development and application prospect of uncooled infrared imaging technology and infrared target tracking technology are introduced, and the principle of uncooled infrared detector and infrared image are described. Point, image processing algorithm and infrared target tracking technology, based on the above theory, the infrared image processing and target tracking hardware platform based on PIC0384 infrared focal plane is built. Based on the hardware of the 384 x 288 infrared imaging system electronic components, this paper realizes the comprehensive detection compensation and the continuous mean quantization transformation of the blind element. Enhancement and other real-time infrared video processing algorithms, in which the mean quantization transform enhancement algorithm can retain more image details while improving image contrast, and effectively prevent problems which are easy to appear in traditional enhancement algorithms, such as over enhancement and "patch". The single target tracking system of FPGA is transmitted from imaging system to target tracking system by LVDS (low voltage differential signal) transmission technology. Based on the theoretical simulation and verification of the TLD tracking algorithm based on Shi-Tomasi feature points and the matching tracking algorithm based on the SIFT feature points, a kind of FPGA is proposed. Based on the improved SIFT feature point optical flow tracking algorithm and hardware implementation. Using the DoG (Gauss difference) operator to detect the feature points, using the local non maximum suppression method to remove the edge response of the SIFT feature points, using the simplified SIFT descriptor to obtain the position shift and scale of the target through the optical flow algorithm, and use the median algorithm to superpose the tracking window. Real time and accurate tracking, and hardware implementation steps and related processing effects are given.
【学位授予单位】：南京理工大学
【学位级别】：硕士
【学位授予年份】：2017
【分类号】：TP391.41

【相似文献】