| Owing to the incomparable superiorities of high sensitivity, good foliage penetration, high precision, high frame rate, non-scanning, light and compact structure and others, the array LIDAR is becoming the main focus and hot topic in the field of imaging LIDAR. In recent years, with the rapid development of integrated circuit manufacturing technology, conventional APDs of small scales have been greatly upgraded to those of large scales which lead to a huge improvement in both imaging resolution and precision. However, the imaging performance is less improved for the array imaging LIDAR when greater cost is spent on the production-manufacturing of larger scale of detectors. Thus, traditional array LIDAR is encountering a bottleneck. In order to overcome the limitations and open up a new prospect, we have to change the traditional mechanisms for the array LIDAR fundamentally and get rid of constraints of the hardware design and manufacturing. However, few innovative researches appear to achieve this goal at present. Therefore, the author has made in-depth researches and explorations into the array imaging LIDAR.The current array LIDAR technology indicates that most detection meachnisms lie in the one-to-one correspondence between the detectors and target pixels, thus resulting in the resolution limitation by the detector scales. Based on the mechnism, we introduce the encoding technique into the array LIDAR system and build a new correspondence between the encoding bits and target pixels; and then employ the multiplexing technique to achieve the multi-pixel acquisition by every single detector; we finally achieve muli-pixel distinction and recognition by decoding technique. Based on the brand-new mechanism, various researches are made as follows.The structure and principle of array-modulated LIDAR system are introduced. Compared with conventional array LIDAR system, the laser pulse of the novel system has to be shaped and encoded first in order to achieve multi-pixel encoding. The back-scattered signals of all pixels are classified into different complete encoding groups, and then multiplexed and respectively coupled to different APD detectors. The digital demodulating algorithm at the terminal is employed to respectively process different full encoded waveforms. The effect of the design is that we achieve M×N pixels detection with M-unit detector array by employing N-bit encoding, thus greatly reducing the requirements of large-scale detectors and readout circuits.Based on the novel idea of array-modulated LIDAR, the encoding and decoding technologies are studied. In this paper, a complete theoretical system is proposed involving laser encoding, pixel multiplexing and full encoded waveform decoding. The reasonable mathematical models are established to solve the physical problem. After the mathematical deduction, the idea of multi-pixel detection by employing the laser encoding and decoding technologies is proved to be feasible in theory. Meanwhile, hardware modules are designed and used to support the theoretical deduction, which provides a verification for the physical feasibility of the novel technology.After the encoding and decoding technology has been applied to LIDAR system, appropriate data processing approach has to be explored for the full encoded waveforms, which are quite different from traditional full waveforms. Thus, a demodulation-based wavelet approach composed of demodulation accumulation, optimized wavelet thresholding denoising and Vondrak smoothing is proposed for global and multi-scale signal enhancement. By using this signal enhancement method, all pixel charaterastic waveforms can be demodulated and effectively enhancedConsidering different pixel characteristic waveforms usually include many characteristic parameters due to the difference of target shape, high-efficient fitting algorithm is necessary to achieve the decomposition of characteristic parameter. In this paper, an optimized LM algorithm is proposed to improve the convergence and precision of waveform decomposition method. In addition, the optimized algorithm is well suited to other full waveform LIDAR systems. Considering that the times-of-flight of all pixels cannot be directly applied to the 3D reconstruction of large FOV array-modulated LIDAR system, the orthographic projection conversion algorithm is proposed based on the space architecture of all pixels.Based on the theoretical research, the array-modulated LIDAR prototype is developed and put into on-site experiments. Before experiments, the hadware modules including all optical elements, encoder driver and synchronism parameters have to be first examined to ensure the regular operation of entire system. The data processing system is designed based on virtual instrumentation technique, which can achieve seamless integration between hardware system and software system and can also facilitate system transplant and system upgrade. By combing the data processing system with 3D control elements, the simulation platform is built to test the practicability of the software system. Then, the 1st generation prototype consisting of 4-unit APD array and 8-bit encoder, namely 32 pixels-per-frame, is applied to the range detection dynamic experiment indoors, in order to verify the physical feasibility of the whole system. The 2nd generation prototype consisting of 4-unit APD array and 64-bit encoder, namely 256 pixels-per-frame, is then applied to dynamic 3D imaging experiment outdoors. The experimental results demonstrate that the array-modulated LIDAR system can achieve 3D imaging of higher resolution and accuracy with small scale of detectors in a more complex environment. Finally, the comprehensive evaluation involving the lateral resolution, scanning resolution and range resolution is described based on the two experimental results, thus providing a reference for the research and development of the next-generation prototype. |
PDF Full Text Request