3d Measurement Of Pulse Image Feature Points Based On Binocular Vision

The pulse signal is one of the important characterization parameters of human body physiology, and contains the rich pathological information. Traditional Chinese medicine pulse-taking lack of objectivity and the pulse signal sensors cannot express human physiological depth information, it has limitations for cardiovascular disease prevention and family care. In order to solve this problem, because the image can represent more abundant information, combined the image processing technology and stereo vision technology, the binocular vision pulse image sensor is designed based on the idea of traditional Chinese medicine pulse-taking, this image sensor collects image information of pulse signal, eventually gets the depth of the visual pulse information through the binocular vision measurement principle.Firstly, The pulse image sensor of the binocular vision is introduced, it from the perspective of bionics, combined the best pulse-taking pressure of the traditional Chinese medicine pulse-taking and the principle of thin film deformation of stress space displacement measurement to design the image sensor, the introduction mainly includes the hardware system, the working principle of the sensor, camera calibration,and the camera calibration method choose Zhang Zhengyou calibration way. Secondly,it introduces the image processing, feature extraction, stereo matching principle of the binocular vision, using binocular vision pulse image sensor collects synchronously pulse images with round logo, through the image filter processing can reduce noise and light difference, due to the circular logo becomes ellipse after the camera imaging,using gray moment subpixel edge detection and ellipse fitting method to extract the center of the circular logo as the desired feature points. Finally, on the basis of the feature extraction need to match these points one by one to find the corresponding point of the same name, in this paper uses SIFT feature descriptor combined with parallax constraints to match feature points, through experimental analysis, this method has good matching results, according to the imaging model of camera,combined feature point matching results and the camera calibration results, so we can get the spatial distribution form of the feature points and complete three-dimensional detection of feature points, it lays a good foundation for three-dimensional reconstruction of the pulse shape.The pulse images acquisition provides a new method and fresh idea for the research of the pulse signal, at the same time it makes a more in-depth research forrealizing pulse-taking objectification and gaining multipoint pulse dynamic information.