We propose an efficient spectral clustering method for large-scale data. The main idea in our method consists of employing random Fourier features to explicitly represent data in kernel space. The complexity of spectral clustering thus is shown lower than existing Nyström approximations on large-scale data. With m training points from a total of n data points, Nyström method requires O(nmd+m³+nm²) operations, where d is the input dimension. In contrast, our proposed method requires O(nDd+D³+n'D²), where n' is the number of data points needed until convergence and D is the kernel mapped dimension. In large-scale datasets where n' << n hold true, our explicitly mapping method can significantly speed up eigenvector approximation and benefit prediction speed in spectral clustering. For instance, on MNIST (60,000 data points), the proposed method is similar in clustering accuracy to Nyström methods while its speed is twice as fast as Nyström.

Image registration is an important technique in many computer vision applications such as image fusion, image retrieval, object tracking, face recognition, change detection and so on. Local feature descriptors, i.e., how to detect features and how to describe them, play a fundamental and important role in image registration process, which directly influence the accuracy and robustness of image registration. This paper mainly focuses on the variety of local feature descriptors including some theoretical research, mathematical models, and methods or algorithms along with their applications in the context of image registration. The existing local feature descriptors are roughly classified into six categories to demonstrate and analyze comprehensively their own advantages. The current and future challenges of local feature descriptors are discussed. The major goal of the paper is to present a unique survey of the state-of-the-art image matching methods based on feature descriptor, from which future research may benefit.

A fundamental problem in Nyström-based kernel matrix approximation is the sampling method by which training set is built. In this paper, we suggest to use kernel k -means sampling, which is shown in our works to minimize the upper bound of a matrix approximation error. We first propose a unified kernel matrix approximation framework, which is able to describe most existing Nyström approximations under many popular kernels, including Gaussian kernel and polynomial kernel. We then show that, the matrix approximation error upper bound, in terms of the Frobenius norm, is equal to the k -means error of data points in kernel space plus a constant. Thus, the k -means centers of data in kernel space, or the kernel k -means centers, are the optimal representative points with respect to the Frobenius norm error upper bound. Experimental results, with both Gaussian kernel and polynomial kernel, on real-world data sets and image segmentation tasks show the superiority of the proposed method over the state-of-the-art methods.

Characterized by the ability to handle varying number of objects, tracking by detection framework becomes increasingly popular in multi-object tracking (MOT) problem. However, the tracking performance heavily depends on the object detector. Considering data association optimization and association affinity model are two key parts in MOT, an online multi-pedestrian tracking method is proposed to formulate a more effective association affinity model. It includes a two-step data association taking advantage of rank based dynamic motion affinity model. The rank based dynamic motion affinity model is used to estimate the object state and refine the trajectory for each of target to achieve the noiseless trajectory. Both these two strategies are beneficial to eliminate ambiguous detection responses during association. To fairly verify the proposed method, three public datasets are adopted. Both qualitative and quantitative experiment results demonstrate the superiorities of the proposed tracking algorithm in comparison with its counterparts.

Depth estimation is essential for many human-object interaction tasks. Despite its advantages, traditional depth sensors, including Kinect or depth camera, are always not wearable-friendly due to several critical drawbacks, such as over-size or over-weight. Monocular camera, on the other hand, provides a promising solution with limited burden to users and attracts more and more attentions in the literature. In this paper, we propose a depth estimation method with monocular camera. Our main idea lies in the weak-supervised learning model of monocular depth estimation based on left and right consistency. To learn an accurate depth estimation, on our training step, we employ LiDAR data, which are generated by laser radar with very high depth accuracy, to semi-supervise the learning scheme. We train our network on ResNet and propose a new penalty function, which takes into account the LiDAR depth loss in training. Compared with several state-of-the-art monocular camera depth estimators, our proposed method obtains the highest depth accuracy.

Point cloud is a discrete and unordered expression of 3D data. A lot of methods have been proposed to solve the problem in 3D object classification and scene recognition. To handle the huge amount of unordered point cloud, down-sampling before processing is needed. The shortage of existing sampling methods is the lack of geometry information consideration, which is essential for point cloud classification and segmentation tasks. Our method is mainly motivated by the observation that points with a high curvature variation can depict the outlines of objects. Thus, we propose a curvature variation based sampling method for point cloud classification and segmentation tasks. We aim to sample points with high curvature variations, which are considered to be more suitable for classification and segmentation tasks than the traditional sampling method. We combine the proposed sampling algorithm with the existing sampling method for multiple information fusion, and a higher accuracy and mean IoU can be achieved. The experimental results verify the advantage of considering curvature variation in classification and segmentation tasks.