Vision based road recognition and tracking are crucial tasks in a field of autonomous driving. Road recognition methods based on shape analysis of road region have the potential to overcome the limitations of traditional boundary based approaches, but a robust method for road region segmentation is the challenging issue. In our work, we treat the problem of road region segmentation as a classification task, where road pixels are classified by statistical decision rule based on the probability density function (pdf) of road features. This paper presents a new algorithm for the estimation of the pdf, based on sequential Monte-Carlo (SMC) method. The proposed algorithm is evaluated on data sets of three different types of images, and the results of evaluation show the effectiveness of the proposed method.

View full abstract

This paper is aimed at employing mirrors to estimate relative posture and position of camera, ie, extrinsic parameters, against a 3D reference object that is not directly visible from the camera. The key contribution of this paper is to propose a novel formulation of extrinsic camera calibration based on orthogonality constraint which should be satisfied by all families of mirror-reflections of a single reference object. This allows us to obtain a larger number of equations which contribute to make the calibration more robust. We demonstrate the advantages of the proposed method in comparison with a state-of-the-art by qualitative and quantitative evaluations using synthesized and real data.

View full abstract

In this paper, we propose a novel algorithm to extrinsically calibrate a camera to a 3D reference object that is not directly visible from the camera. We use the spherical human cornea as a mirror and calibrate the extrinsic parameters from its reflection of the reference points. The key contribution of this paper is to present a cornea-reflection-based calibration algorithm with minimal configuration; there are three reference points and one mirror pose. The proposed algorithm introduces two constraints. First constraint is that the cornea is virtually a sphere, which enables us to estimate the center of the cornea sphere from its projection. Second is the equidistance constraint, which enables us to estimate the 3D position of the reference point by assuming that the center of the camera and reference point are located the same distance from the center of the cornea sphere. We demonstrate the advantages of the proposed method with qualitative and quantitative evaluations using synthesized and real data.

View full abstract

This paper presents a novel shape descriptor for topology-based segmentation of 3D video sequences. 3D video is a series of 3D meshes without temporal correspondences which benefit for applications including compression, motion analysis, and kinematic editing. In 3D video, both 3D mesh connectivities and the global surface topology can change frame by frame. This characteristic prevents from making accurate temporal correspondences through the entire 3D mesh series. To overcome this difficulty, we propose a two-step strategy which decomposes the entire sequence into a series of topologically coherent segments using our new shape descriptor, and then estimates temporal correspondences on a per-segment basis. As the result of acquiring temporal correspondences, we could extract rigid parts from the preprocessed 3D video segments to establish partial kinematic structures, and could integrate them into a single unified kinematic model which describes the entire kinematic motion in the 3D video sequence. We demonstrate the robustness and accuracy of the shape descriptor on real data which consist of large non-rigid motion and reconstruction errors.

View full abstract