Research On LiDAR Image Reconstruction Technology And Multi-core DSP Implementation

GM-APD(Geiger-Mode of Avalanche Photo diodes)array lidar has become a research hotspot in the field of military target detection due to its advantages of low energy consumption and long detection distance.Aiming at the problem of high computational complexity and time-consuming reconstruction of laser radar target detection images,this paper studies the laser radar image reconstruction technology,selects multi-core DSP as the image reconstruction processing hardware platform,and applies the reconstruction algorithm proposed in this paper to the multi-core DSP Parallel design and implementation are carried out on the platform,and the code and memory performance are optimized,and the processing speed of image reconstruction is improved.The specific research content is as follows:Firstly,the working mechanism of the GM-APD array lidar is analyzed,and then starting from the target echo signal model,the GM-APD trigger probability model is constructed to explore the mapping relationship between the echo signal intensity and the GM-APD detection probability density.Based on the cumulative detection system,the 3D image reconstruction method of GM-APD echo data based on half-maximum width fitting is studied to achieve high-quality reconstruction of the target image.The performance of the proposed algorithm is verified based on the PC platform.The results show that the proposed reconstruction method has a 114% increase in computing speed compared with the 3D image reconstruction method of GM-APD echo data under low SNR.Aiming at the problems of large size and high power consumption of the PC platform,the multi-core DSP TMS320C6678 is selected as the image reconstruction hardware platform,and a parallel processing scheme based on the combination of the Fork-Join model and data flow is designed,and the synchronous communication between cores based on the shared memory mode is constructed method,and tested the image reconstruction algorithm based on multi-core DSP in the bare-metal development environment.Experimental results show that the running speed of the algorithm based on multi-core DSP is shortened from 7.37 s to 4.75 s compared with single-core,and the operation speed is increased by 36%.In order to improve the running speed of the 3D image reconstruction algorithm on the multi-core DSP,the clock cycle required by each function in the program is obtained by using the performance detection tool,and the execution time and call times of each function are analyzed to clarify the performance bottleneck of the program.Keyword optimization,loop optimization,and compiler optimization are performed on the code,memory bandwidth-based optimization and Cache-based optimization are performed on the memory,which effectively utilizes the storage space,reduces redundant calculation of data,and improves the efficiency of code execution.The optimization test was carried out on the multi-core DSP platform.The experimental results show that the running time after optimization is shortened to 3.62 s compared with that before optimization,and the operation speed is increased by 24%.