Design Of Data Transmission And Control Unit And Image Segmentation Algorithm For Terahertz Imaging

The Terahertz passive detection imaging system takes advantage of the high resolution and strong penetration of the terahertz imaging technology to detect hidden objects.Non-contact screening,early warning and detection imaging of dangerous objects such as prohibited items,hidden weapons,etc.at entrances and exits of important places have broad product development prospects and advantages that are difficult to be replaced.Among them,the data transmission and control unit in the detection system realizes data collection,processing,transmission and other functions.The servo control unit can complete the complete scan of the space.These two units are very important parts of the THz passive detection system.Before the detection and recognition of the detected image,it is usually necessary to perform image segmentation.How to achieve a better segmentation effect is always a research problem,and the quality of subsequent detection and recognition is closely related to image segmentation.This thesis relies on actual scientific research projects to carry out research: the design and implementation of the data transmission and control unit in the THz passive detection imaging system;and the image segmentation method based on traditional segmentation algorithms and deep learning.Verification of functions and indicators through experiments and simulations.The specific work content is summarized as follows:(1)Combined with the basic principles of terahertz passive detection imaging technology,according to the overall performance index requirements of the system,the parameters are analyzed and demonstrated,and the functional structure and hardware circuit design of the data acquisition and transmission and control unit are completed;Verilog HDL language is used for digital Logic design;realize the process of image data acquisition,image preprocessing,arrangement storage and transmission from terahertz and optics to the lower computer.Completed the joint debugging of the terahertz front end,data transmission and control unit,and receiving and storage modules in the display and control unit.(2)According to the overall indicators and functional requirements of the system,the technical indicators of the servo control unit were analyzed and demonstrated;the space scanning method of the servo was designed;the drive control of the servo control unit and the design plan of the mechanical structure were proposed,and the selection analysis of core components was completed.(3)For the problem of image segmentation,the advantages and disadvantages of traditional classic methods are studied and compared,and their application to THz passive imaging has certain limitations.Based on this,an improved idea of combining gray histogram with image spatial position and edge information is propose.Specifically using a combination of gray and gradient to segment the image.(4)Carried out research on terahertz image segmentation algorithm based on deep learning.Aiming at the requirements of small target detection and real-time performance,a multi-scale convolution window and deep separable convolution technology are used to construct a neural network model architecture for segmenting terahertz images.It can increase the operation speed,reduce the number of parameters,and reduce the network scale.Finally,it is proposed to use multiple indicators such as fuzzy entropy to comprehensively measure the quality of image segmentation.