3D Reconstruction Of Gastrointestinal Images From Capsule Endoscope Based On Six-dimensional Positioning And Orientation Technique

The invention of wireless capsule endoscope(WCE) improves the patient’s feeling when doing gastrointestinal endoscopy or colonoscopy. Wireless capsule can transmit high-definition GI images after patients swallowed capsule. The WCE make easier for doctors to diagnose diseased tissues, it is a potential GI examination. However, the current clinical capsule endoscopes still have some problems to be solved. The commercial capsule endoscope can only provide two-dimensional image, can’t provide the true 3D scene. The specific area of diseased tissues are usually obtained from the doctor’s experience, so there will be some error. Therefore, we need more accurate three-dimensional detection methods. In this paper, we do research in three-dimensional reconstruction of gastrointestinal based on sequence capsule endoscope image, the main contents are as follows:1. This paper do research on the camera model, including the linear model and the camera lens’ s internal and external parameters, and nonlinear model caused by optical lens distortion. Several camera calibration methods are designed, introduced a calibration method based on planar template, and design experiment calibration algorithm. This test use template calibration method to calibrate WCE camera and nonlinear distortion correction. And this paper do feature points matching from two images taken by different directions of the camera.2. This paper try to use the SIFT algorithm to do feature matching in capsule endoscope image. Because it is difficult to find the feature points in the smooth image, this paper presents a matching algorithm based on template and variable scales. The size of template and search area are optimized, and the matching result is accurate. Meanwhile, this algorithm has good anti-noise ability,the number of feature points calculated by matching algorithm is enough to make 3D reconstruction.3. This paper do research on 3D reconstruction based on feature points. We use a combination of single camera and magnetic location data. Eight point algorithm can calculate the position and direction changes(including the translation vector and rotation matrix) of the capsule. Calculate the three-dimensional coordinates of feature points in the scene through the camera vision mechanism. We use Delaunay triangulation algorithm to constructed three-dimensional grid. Experiments gastrointestinal scene are designed to calibrate and test the algorithm. We do experiment based on C++ and Open GL, experimental results show that the three-dimensional reconstruction satisfied, and have high accuracy.