Research On Segmentation And Measurement For Moving Object

With the development of computer vision,it still been a hot field to process the object.There are a lot of excellent methods which be used in object segmentation,object recognition,object tracking,image retrieval,object measuring,as well we can get the-start-of-art result.While when we aim to process the moving object,the image degradation will seriously endanger the effectiveness of those methods.Particularly object segmentation,be the foundation of higher-class vision research,whether having a good segmentation results directly decide the subsequent quality of object process.This paper start the research with the moving object,design a non-uniform motion degradation images based moving object segmentation method.Moreover,on the basis of object segmentation,designing a measuring rotational angular velocity method of moving object based on monocular feature-based SLAM.The main work and achievements are as follows:1.A moving object segmentation method for non-uniform motion degradation image is proposed.Aiming at the non-uniform motion degradation image caused by multiple relative motion,a super-pixel level composite feature of localized autocorrelation consistency and fuzzy connection degree is constructed on the basis of over-division preprocessing.And an ultra-pixel background search algorithm is designed to reconstruct the moving object from the background using the composite feature.2.Take the problem that the SLAM algorithm is slow and difficult to initialize into consideration,we come up with a monocular SLAM fast initialization method.Based on the image vanishing point detection,we generate the initial map by constructing the depth by normalizing the total distance from each feature point to vanishing point,or normalizing the normal random number.Meanwhile,the optimization of targeted algorithms lays the foundation for real-time rotation moving measurements.3.To solve the problem of visual motion object measurement,we propose a measurement method of uniform rotational motion.under the foundation of the pre-segmentation of the Gaussian mixture model,the optimized monocular feature-based SLAM is used to calculate the angle of rotation of the object between frames,remove the inversion error cause by the SLAM error estimate,and calculate the rotation angle of the object and angular velocity finally.Quantitative and qualitative experiments show that the method of moving object segmentation for non-uniform motion degradation images,and measuring rotational angular velocity of moving object based on monocular feature SLAM proposed in this paper have good results,and have great following research and application value.