Research On Laser 3D Imaging Of Compressed Sensing Based On Range Encoding

Compressed sensing technology is a new type of signal processing methodology developed in recent years.The advantage of compressed sensing is that the projection data of the signal is much smaller than the amount of data obtained by the traditional sampling method based on Shannon's theorem,so it is widely used in the related fields of signal processing.This thesis analyze the shortcomings of the 3D imaging technology,the processing data is numerous and the processing speed can not meet the requirements.The accuracy of the ranging can not be guaranteed and the image SNR needs to be improved.This thesis presents a 3D imaging method of compressed-sensing laser based on distance coding,which reduces the calculation times of compressed sensing technology based on slice method from N times to 2 times.We constructed a simulation platform of 3D laser imaging system based on distance coding.The module of modulation and transmission system,receiving and collecting system module,control and processing system module are prepared.The simulation results show that the SNR of H objects are 2.90 and 2.97,respectively,and the SNR of T objects are 3.92 and 4.87 in the 200-sparse of OMP algorithm based on DCT and FFT transformation.The SNR of H and T object images with GPSR algorithm based on DCT transformation are 2.98 and 4.85,and the SNR of TV algorithm are 13.19 and 19.78.It is verified that the TV algorithm is the most suitable algorithm.And then apply the TV algorithm for 3D imaging simul ation,the results proved the feasibility of 3D imaging in distance coding.The results of the semi-physical simulation show that the SNR of H and T are 5.25 and 5.05,respectively,under the OMP algorithm with the sparse degree of 200 based on the DCT transformation,which are 5.08 and 4.63 based on FFT transformation,respectively.And the corresponding SNR of GPSR based on DCT transformations are 3.34 and 4.69.Comparing to the Bernoulli random sampling matrix,the Gaussian random sampling matrix improved picture quality on the whole image,especially in the H image.GPRS algorithm also increases the SNR of H objects.The results show that the Gaussian random sampling matrix is better than the Bernoulli random sampling matrix in measurement result.The laser imaging system is composed of a laser detector,a DMD micromirror and a convex lens.The receiving and collecting system is composed of a PIN detector,a transimpedance amplifier and a high-speed sampler.The control and processing system contains a PIN detector for triggering a signal and a computer used to control the system.The experimental target is a shape of the H,I,T or other letters of the object,the experimental environment is the dark closed environment.The SNR of the system is 35.97,and the regular objects of different reflectivity can be reconstructed to a clear image.In the end,the influence of the sampling rate and the average number of measurements on the imaging is explored.It is shown that the TV algorithm can restruct a regular image at the sampling rate of 5%.With the increase of the average cumulative number of measure times,the SNR of the image is also improved further.Then three different coding methods are adopted,which proves that the three coding methods are able to maintain the accuracy of distance measurement well.The accuracy of linear coding on the H is the smallest in the accuracy of distance measurement,reaching 0.38 m,and the accuracy of exponential coding on the T is the smallest,reaching 0.71 m.For the distance deviation,the ranging deviation on H of the root number is the smallest,reaching 0.38 m,and the ranging deviation on T of the exponential coding is the smallest,reaching 1.34 m.Finally,we use the method of spatial filtering.According to the results of the three sets of measurements,the minimum accuracy of distance measurement on the H is the median filter method in root coding,which is 0.23 m and the minimum accuracy of distance measurement on the T is the ordfilt method in the exponential coding,which is 0.34 m.In terms of ranging accuracy,H's best distance deviation is linear adaptive Wiener filtering method,0.38 m,and the best distance deviation of T is the exponential coding under the ordfilt method,1.31 m,all accuracy of distance measurement was improved.In the spatial filtering of the image,the median filter has the highest SNR,which is 1.26 times as much as that of the original image.