Research On 3D Measurement System Based On Binocular Endoscope

Electronic medical devices are now widely used in clinical practice,and many advanced electronic medical devices help doctors to diagnose and manage conditions more scientifically and effectively.In the use of endoscopy,doctors are able to visualise the tissue pattern of the patient’s internal organs,which helps them to diagnose the condition.However,one of the limitations of endoscopy or endoscopy-assisted surgery is the lack of binocular or stereoscopic vision.Monocular endoscopes and monitors produce twodimensional images,which can impair depth perception and the ability to estimate dimensions.With the development of stereoscopic 3D measurement technology,stereoscopic 3D measurement is widely used in the medical field and the addition of stereoscopic 3D measurement algorithms to endoscopic systems enables the surgeon to obtain accurate and intuitive 3D information in the surgical scene.This paper focuses on a binocular endoscopy system to achieve 3D measurement of real endoscopic scenes.The main work includes the design of the overall system scheme,the implementation of highprecision 3D measurement algorithms,the evaluation and analysis of the accuracy of 3D measurement algorithms,and the design and development of the software for the binocular endoscopic 3D measurement system.According to the research task of this project,a 5.5mm diameter binocular endoscope is designed for a small space and the overall solution of the system is designed and developed.Before solving the homography matrix,the LSD straight line detection algorithm is used to detect the straight lines of the checkerboard target and each pixel point in the image belongs to at most one straight line according to the LSD straight line detection algorithm.The characteristics of the straight line are extracted from the corner points of the tessellation grid,and the extracted initial corner point coordinates do not meet the requirements of high precision 3D measurement requiring sub-pixel optimisation of the initial corner point coordinates using the grey scale gradient method.The extracted corner point coordinates of the tessellation grid based on LSD line detection combined with the results of the camera marker can solve the homography matrix and use the homography matrix to calculate the 3D coordinates of the target.For the evaluation and analysis of the accuracy of the 3D measurement algorithm,the binocular endoscope is first calibrated using a high-precision checkerboard calibration plate,and the intrinsic matrix and aberration coefficients obtained from the calibration were used to correct the aberrations and polar lines of the binocular images taken by the endoscope.The Pan-Tilt is then fixed with a 0.001 mm precision Pan-Tilt and panned to the right by 2.5mm.The LSD-based corner point extraction is performed on the board before and after the movement and the homography matrix is solved.Finally,the influence of the corner point extraction accuracy and the object distance between the binocular endoscope and the target on the 3D measurement accuracy are analysed.The software for the binocular endoscope based 3D measurement system is developed on the Qt platform using the C++ programming language.The main function of the system software is to track the target object in the video stream to obtain 3D information and visualise its trajectory.The system software consists of three main parts,namely the camera calibration module,the initialisation module and the visualisation module.