Research And Implementation Of Color Single-Pixel Imaging Quality Optimization Based On Random-Circulant Matrix

The rapid development of information technology has caused people's increasing demand for the amount of information,and has brought great difficulties for the traditional signal processing technology to meet the practical demands.The theory of compressive sensing(CS)has given scientists opportunities to break the compression limit of Shannon-Nyquist sampling theorem,and it has also led to a revolution in the field of information technology.CS theory consists of three parts:sparse transformation,measurement matrix and recovery algorithm.The performance of the measurement matrices are related to the signal acquisition and recovery.The most impressive achievement of CS theory is the single-pixel camera developed by RICE University.The system has both verified the practical significance of CS theory,and put forward a new set of imaging system.The single pixel camera adopts single photon sensor.It compresses the captured image during sampling,and restores it through CS algorithm.Hence,single pixel camera acquires the image information by "sampling while compressing",which has saved sampling resources and reduced cost of sensor.The main contributions of this thesis are as follows:(1)Based on the deep research on the theory of compressive sensing,firstly,the typical binarized measurement matrices were simulated and compared.Based on the performance analysis of the improved Logistic chaotic system,the binarized deterministic measurement matrices based on Logistic chaotic sequence was proposed.Then,on the basis of circulant matrix and Toeplitz matrix,this thesis made full use of the exterior determinacy and inherent randomness of chaotic sequences,and designed the random-circulant measurement matrix.Compared with other deterministic measurement matrix,the restoration effect was greatly improved,and it had advantages in hardware and storage.(2)Based on the research on the single pixel camera system,this thesis improved the light path and signal processing module on the basis of original platform.Including the following aspects:using the lens of digital single lens reflex with better focusing capability instead of a single convex lens;selecting a silicon photodiode with a larger imaging area as a single pixel sensor;using more accurate and efficient amplification filter circuit and A/D converter,ect.The imaging quality and practical application ability of the system were greatly improved.(3)This thesis proposed two groups of implementation schemes of color single-pixel camera and could accurately obtain color information.Firstly,the principle and workflow of the two schemes were introduced.Then the advantages and disadvantages of the two schemes were analyzed and contrasted.Finally,the second design scheme with better real-time performance was adopted.The scheme could get the structure and RGB information of target images.Through a variety of experimental data to prove that the proposed improved measurement matrix combined with the hardware system could accurately reconstruct the RGB image and has good robustness.