阵列调制激光三维成像雷达编解码及信号处理技术研究与实现

发布时间：2017-12-31 17:26

本文关键词：阵列调制激光三维成像雷达编解码及信号处理技术研究与实现　出处：《南京大学》2016年博士论文　论文类型：学位论文

【摘要】：阵列成像LIDAR具有高灵敏、高隐蔽、高精度、高帧频、无扫描、结构轻小等优势,因而成为成像LIDAR领域的一个重要的发展对象。近年来,随着集成电路制造技术的飞速发展,传统的小面阵探测器已经朝着大面阵或超大面阵规模发展,使得阵列成像LIDAR的成像分辨率、成像精度都有了巨大的提高。然而,阵列成像LIDAR的研究过于集中在探测器件的生产制造上,硬件成本和体积大幅度增加的同时,成像性能却提高有限,传统的阵列LIDAR系统面临着极大的局限性。在此情况下,从根本上改进阵列LIDAR的探测机制,摆脱硬件设计和工艺的束缚,将会打开一条全新的发展方向,而目前却尚未出现在这一方面的创新性研究。因此,本文基于阵列成像LIDAR开展深入的研究和探索。目前的阵列LIDAR探测技术表明,探测机理建立的是探测器与目标像素的单一性对应关系,因而造成分辨率受限于探测器规模的特点。以此作为突破口,引入编码技术,建立目标像素与编码位数的对应关系;再通过复用技术,实现单一探测器对多目标像素的复合接收；最后通过解码技术实现对混合复用信号的多像素识别。基于这一全新的思路,本文开展了如下的研究工作：介绍阵列调制LIDAR的系统结构和工作原理,以此为依据开展后续工作。与传统阵列LIDAR相比,阵列调制LIDAR的光源需要经过整形和编码,以实现对目标像素的编码。接收端对编码像素后向散射的信号进行分组耦合和复用接收。接收端通过有效的数字解调算法对混合编码波形进行处理,分离出所有像素信息。如此设计的效果是,M元探测器阵列的每一单元复用N位经过编码的目标像素,并实现对所有像素的高精度识别。也就是说,M元探测器阵列借助N位编码最终获得M×N像素阵列的信息,从而极大地减少了传统方法对探测器及相应的读取电路的要求。针对全新的阵列调制编码LIDAR系统,首先开展了编解码技术的研究。本文建立了包括阵列编码、像素复用、编码全波形解码的完整的编解码技术理论体系。将实际需要解决的物理问题转化为数学模型,经过推导和演算,证明了激光编解码实现多像素探测的理论可行性。同时,设计合适的硬件模块实现方法完成对每一步数学操作的功能,进而证明了激光编解码技术的可实现性。阵列调制编码LIDAR在结合了编解码技术后,获得的观测数据波形与传统全波形有所区别,因而需要进一步探索这种特殊编码全波形的数据处理方法。首要的任务是采用何种信号增强方法来提高信噪比,以提高探测距离和成像精度。本文提出了一种基于数字解调积累、小波优化阈值、Vondrak平滑的信号增强处理方法,在完成像素信息解码的同时实现了周期噪声积累、小波阈值多尺度去噪和局部毛刺消解,有效地提高了各解调像素特征波形的信噪比。经过数字解调后的各像素特征波形因目标形状的差异往往包含多个特征参数,需要借助高效的拟合算法实现特征参数的分解。本文提出了优化LM波形分解方法,实现了更高速、更高精度的波形数据分解。首先改进初始参数估计方法,减少了运算的时间。接着提出LM算法模型的最优调谐模型,大幅度地提高了迭代算法的收敛速度。其次,通过改进传统高斯分解的基波,使之匹配实际激光波形,可有效地提高了波形拟合的精度,从而提高最终像素的分解精度。该算法不仅适用于本系统,也适用于大部分全波形LIDAR系统,可实现大幅宽、高精度的目标探测。最后,阵列调制LIDAR的大视场线阵探测方式要求对波形分解所得到的阵列像素的飞行距离信息进行正射投影修正,本文推导了相应的正射投影距离转换的算法。在理论研究的基础上,本文最后展示了阵列调制LIDAR系统的原理样机和数据处理系统,并进行了现场实验。在进行试验之前,需要对系统各模块、编码驱动器、系统同步控制参数进行检查,以确保整个系统正常运行。基于虚拟仪器技术设计数据处理系统,不仅实现了软件系统与硬件系统的高度集成,完成实时数据获取及处理,而且可以方便系统移植和系统升级。虚拟仿真系统采用三维控件设计仿真测试目标和虚拟环境,结合数据处理模块完成仿真实验,对数据处理系统的功能性验证起到了关键性作用。最后,对软硬件系统联调,完成外场实验。一代样机包括4元APD与8位编码,即单帧32像素。将其应用到室内动态测距实验,以初步验证阵列调制LIDAR测距的有效性。二代样机包括4元APD与64位编码,即单帧256像素。将其应用到室外动态三维成像实验,证明了阵列调制LIDAR在更复杂的环境中可实现小规模探测器阵列完成高分辨率、高精度的三维成像。最后,对阵列调制LIDAR的横向分辨率、扫描分辨率、距离分辨率等指标进行了综合评价,以指导未来的新一代样机的研究。
[Abstract]:LIDAR array imaging with high sensitivity, high concealment, high precision, high frame rate, without scanning, light structure and other advantages, and thus become an important development object field of LIDAR imaging. In recent years, with the rapid development of integrated circuit manufacturing technology, the traditional facet array detector has a large array or large surface the scale of the development of the array, the resolution of the imaging array LIDAR imaging, the imaging accuracy has been improved dramatically. However, the research of LIDAR array imaging is too focused on the production of the detection device, the hardware cost and the volume increases greatly, but improve the imaging performance, array of traditional LIDAR system faces great limitations. In this case, the detection mechanism fundamentally improved LIDAR array, get rid of the shackles of hardware design and process, will open a new direction of development, and now has not yet appeared in the party The study of innovation. Therefore, this paper carried out an in-depth study and exploration of LIDAR array imaging based on LIDAR array detection technology. That at present, the detection mechanism is established for the single correspondence between the detector and the target pixel, resulting in the characteristics of the detector resolution is limited scale. As a breakthrough, the introduction of encoding technology, and established the corresponding relationship between the target the pixel and the encoding bits; through multiplexing technology, realize receiving single detector of composite target pixel; finally through decoding technology to achieve multi pixel identification of mixed multiplexing signals. This new method based on the research works in this dissertation are as follows: this paper introduces the system structure and working principle of the LIDAR modulation array, in order to carry out on the basis of the follow-up work. Compared with the traditional light source array array LIDAR, LIDAR modulation by shaping and encoding, in order to achieve the target Pixel encoding. The receiver encoding pixel backscattering signal packet coupling and multiplexing receiving. The receiving end to mixed encoding process the waveforms through digital demodulation algorithm, separated all the pixels information. So the design effect is that each unit of multiplexing N M yuan detector array after the target pixel encoding, and to achieve high accuracy recognition for all pixels. That is to say, N encoding for M * N pixel array information through the M element detector array, thereby reducing the traditional method of detector and the corresponding read circuit requirements greatly. Aiming at the array modulation encoding the new LIDAR system, first launched a research series decoding technology. In this paper, including the establishment of an array of pixel multiplexing, encoding, decoding and encoding full waveform complete coding theory. The actual need to solve the physical problem into a number of The model, after deduction and calculation, proved the feasibility of laser decoding theory of multi pixel detection. At the same time, the design of appropriate hardware module method to complete each step of the mathematical operation function, and prove that the realization of laser encoding and decoding technology. In combination with LIDAR modulation array encoding decoding technology. The difference was the waveform of the observational data and the traditional full waveform, and needs further exploration of this special encoding full waveform data processing method. The primary task is to enhance the method to improve the signal-to-noise ratio of the signal to improve the detection distance and imaging precision. This paper presents a digital demodulation based on wavelet threshold optimization, accumulation the signal processing method to enhance Vondrak accumulation of smooth, while finishing the pixel information decoding to achieve periodic noise, wavelet threshold denoising and multi-scale local burr elimination solution, a Effectively improve the demodulation pixel characteristic waveform signal to noise ratio. After each pixel feature waveform after digital demodulation due to differences in the shape of the object often contain multiple parameters, need to decompose characteristic parameters by using the fitting algorithm efficient. This paper proposed a decomposition optimization method of LM waveform, to achieve a more rapid, more waveform data high precision decomposition. First improve the initial parameter estimation method, reduces the computation time. Then the optimal tuning model LM algorithm, greatly improve the convergence speed of the iterative algorithm. Secondly, through the improvement of the traditional fundamental Gauss decomposition, to match the actual laser waveform, which can effectively improve the waveform fitting the precision, so as to improve the accuracy of the final pixel decomposition. This algorithm is not only suitable for this system is also suitable for most of the full waveform of the LIDAR system, can achieve a wide, high precision of target. Test. Finally, large field line array LIDAR modulation array detection for flight distance information of the pixel array waveform decomposition of orthographic projection correction, derived orthographic projection distance transformation algorithm. On the basis of theoretical research, this paper shows the principle prototype array and data processing system modulation of the LIDAR system, and carried out field experiments. Before the test, the need for each module, encoding system drive, synchronous control system parameters for inspection, to ensure the normal operation of the whole system. The technology design of data processing system based on virtual instrument, not only to achieve a highly integrated software system and hardware system, real-time data acquisition and processing, and can facilitate the system transplantation and system upgrade. The virtual simulation system using three-dimensional control design and simulation test target and virtual environment, combined with the number of According to the processing module to complete the simulation experiment, the key role of the functional verification of the data processing system. Finally, the software and hardware of the system debugging, complete experiment. A prototype including 4 yuan APD and 8 bit encoding, namely single frame 32 pixels. Its application to the ranging experiment indoor dynamic, in order to effectively preliminary validation of array LIDAR modulation range. The two generation prototype including 4 yuan APD and 64 bit encoding, namely single frame 256 pixels. Its application to 3D imaging experiments proved that the outdoor dynamic array, LIDAR modulation can achieve small scale array detector with high resolution in a more complex environment, 3D imaging with high precision. Finally, the transverse resolution of array LIDAR modulation scanning resolution, range resolution and other indicators of the comprehensive evaluation and research to guide the future of the new generation prototype.

【学位授予单位】：南京大学
【学位级别】：博士
【学位授予年份】：2016
【分类号】：TN957.51

【相似文献】