Binocular Stereo Vision System For Surgical Navigation: Design And Its Key Technique Research

With the rapid development of computer science and technology, stereo vision especially binocular vision is widely used in medical examination, industrial inspection, video surveillance, machine vision systems and other fields. A boom driven by machine vision is rising, which liberate the human eyes, and also promote the tremendous progress to scientific visualization. With the extensive application of machine vision, it is more and more important to realize the target tracking fast and effectively by video processing. And real-time tracking to dynamic targets is widely used in the military target detection, intelligent surveillance systems, intelligent transport systems and surgical instruments tracking based on the surgical navigation systems which is involves in this paper.Robot-assisted surgery is a hot issue in the field of robotics, many systems of it have been used for many years in the clinical environment, such as the Da Vinci robot-assisted surgery system, which have been widely used in gynecological surgery, heart surgery and pediatric surgery and other fields. Tracking surgical instruments or robots in real-time is the basis of secondary surgery. If a visual locator is used in three-dimensional positioning system, the video of operation process can also be provided while targets are located, which can improve the system efficiency, and ensure safe operation.Binocular stereo vision system for surgical navigation is studied, such as system design, camera calibration, image distortion correction, feature extraction, indicator design and surgical instrument calibration, which is intend to provide a rapid targets locating programs for surgical navigation system, and develop a real-time surgical instrument tracking system.It is one difficulty of visual locator to realize online and high-precision detection to targets. On the basis of evaluation and comparison the detection ability to X-corner about the commonly used corner detection algorithms, this paper improved and adopted a segment test based X corner detection and sub-pixel localization method. This method can detect the X corners effectively, exclude other types of corners, and process image in real-time with high positioning accuracy, which can meet the requirements of assisted surgery. A no similar distortion correction model is adopted, which realize local correction with high speed and accuracy. A specialized graphic pattern (named indicator) attached to the surgical instrument is designed, and multiple objectives can be identified and tracked online by distinguishing different geometric parameters of the indicator. The influence of indicator structure to surgical instruments calibration accuracy is analysied.Surgical instruments calibration is the basis of surgical instruments navigation, which is intend to find the structural parameters describing the spatial relations between surgical instruments and indicator. This paper give up the application of measurement calibration method which has low precision and complicated process, and proposed a visual calibration method using binocular vision camera. In this method, the calibration parameters are obtained by least squares, and the tracks that instruments rotate around the given point is recorded. On this basis, theoretical analysis about the errors of calibration and location is made. The results show that the calibration accuracy depends not only on the positioning accuracy of the visual locator, but also on the indicator structure and the relative position of indicator and surgical instruments; The surgical instrument calibration error will not greater than visual localization error, when the indicator structure is reasonable,.Experiments results show that four more targets can be tracked online at the same time by the system, the mean average positioning accuracy can up to 0.23mm,and the positioning refresh rate can up to 12Hz.