Hideki Koike

Simple Camera Calibration From a Single Image Using Five Points on Two Orthogonal 1-D Objects

We propose a simple and practical calibration technique that effectively estimates camera parameters from just five points on two orthogonal 1-D objects, each which has three collinear points, one of which is shared. We derive the basic equations needed to realize camera calibration from just five points observed on a single image that captures the objects. We describe a new camera calibration algorithm that estimates the camera parameters based on the basic equations and optimizes them by the bundle adjustment technique. Our method is validated by both computer simulated data and real images. The results show that the camera parameters yielded by our method are close to those yielded by existing methods. The tests demonstrate that our method is both effective and practical.

Model based human motion tracking using probability evolutionary algorithm

A novel evolutionary algorithm called probability evolutionary algorithm (PEA), and a method based on PEA for visual tracking of human motion are presented. PEA is inspired by estimation of distribution algorithms and quantum-inspired evolutionary algorithm, and it has a good balance between exploration and exploitation with very fast computation speed. The individual in PEA is encoded by the probability vector, defined as the smallest unit of information, for the probabilistic representation. The observation step is used in PEA to obtain the observed states of the individual, and the update operator is used to evolve the individual. In the PEA based human tracking framework, tracking is considered to be a function optimization problem, so the aim is to optimize the matching function between the model and the image observation. Since the matching function is a very complex function in high-dimensional space, PEA is used to optimize it. Experiments on 2D and 3D human motion tracking demonstrate the effectiveness, significance and computation efficiency of the proposed human tracking method.

Task-Oriented Evaluation of Super-Resolution Techniques

The goal of super-resolution (SR) techniques is to enhance the resolution of low-resolution (LR) images. How to evaluate the performance of an SR algorithm seems to be forgotten when researchers keep producing algorithms. This paper presents a task-oriented method for evaluating SR techniques. Our method includes both objective and subjective measures and is designed from the viewpoint of how SR impacts many essential image processing and vision tasks. We evaluate some state-of-the-art SR algorithms and the results suggest that different SR algorithms should be utilized for different applications. In general, they reflect the consistency and conflict between objective and subjective measures as well as computer vision systems and human vision systems do.

Estimating Anomality of the Video Sequences for Surveillance Using 1-Class SVM

We have proposed a method to detect and quantitatively extract anomalies from surveillance videos. Using our method, anomalies are detected as patterns based on spatio-temporal features that are outliers in new feature space. Conventional anomaly detection methods use features such as tracks or local spatio-temporal features, both of which provide insufficient timing information. Using our method, the principal components of spatio-temporal features of change are extracted from the frames of video sequences of several seconds duration. This enables anomalies based on movement irregularity, both position and speed, to be determined and thus permits the automatic detection of anomal events in sequences of constant length without regard to their start and end. We used a 1-class SVM, which is a non-supervised outlier detection method. The output from the SVM indicates the distance between the outlier and the concentrated base pattern. We demonstrated that the anomalies extracted using our method subjectively matched perceived irregularities in the pattern of movements. Our method is useful in surveillance services because the captured images can be shown in the order of anomality, which significantly reduces the time needed.

Estimating the number of people in a video sequence via geometrical model

We propose a novel technique for estimating the number of people in a video sequence; it has the advantages of being stable even in crowded situations and needing no ground-truth data. By analyzing the geometrical relationships between image pixels and their intersection volumes in the real world quantitatively, a foreground image can be directly indicate the number of people. Because foreground detection can be done even in crowded situations, the proposed method can be applied to such situations. Also it can estimate the number of people in an a-priori manner, so it needs no ground-truth data which is necessary for existing feature-based estimating techniques. Experiments show the validity of the proposed method.

Contrasting shadow for occluder light suppression from one-shot image

Two main problems for front projection systems when a user appears between a screen and a projector are 1) shadows being cast on a screen and 2) the user being illuminated due to undesirable strong projection light. To solve these problems, it is necessary to know which projectors are occluded by a user and cast a shadow. We propose a method that suppresses occluder light from a single image based on the fact that the darkness of shadows on a screen is differentiated when projectors have different luminance. Contrary to previous approaches, our method utilizes different alpha masks for each projector to contrast shadows on a screen. This enables us to clarify the relationship between the shadow on a screen and the occluded projector. Additionally, by differentiating all projector luminances, our method makes it possible to discriminate shadows even when some shadow regions overlap on a screen. We implemented the proposed method into a three-projector system to demonstrate its effectiveness.

Monocular 3D tracking of multiple interacting targets

In this paper, we present a new approach based on Markov Chain Monte Carlo(MCMC) for the stable monocular tracking of variable interacting targets in 3D space. The crucial problem with monocular tracking multiple targets is that mutual occlusions on the 2D image cause target conflict (change ID, merge targetshellip). We focus on the fact that multiple targets cannot occupy the same position in 3D space and propose to track multiple interacting targets using relative position of targets in 3D space. Experiments show that our system can stably track multiple humans that are interacting with each other.

Online anomal movement detection based on unsupervised incremental learning

We propose an online anomal movement detection method using incremental unsupervised learning. As the feature for discrimination, we extract the principal component of the spatio-temporal feature by incremental PCA. We then detect anomal movements by an incremental 1-class SVM. In order to use principal component as the feature for discrimination while supporting incrementation of the subspace, we modify the SVM kernel function to take account of the difference in distance scale between the principal component feature vectors and that of the feature vectors after the subspace is incremented. This allows us to efficiently conduct the relearning process even though the dimension of the original input spatio-temporal feature is high. Experiments show that anomal scenes can be detected without the cost of preparing a lot of labeled data for preliminary learning.

Modeling timing structures in gait image sequences using bottom-up clustering

In this paper, we show that timing differences among body parts characterize human gait motions. To represent the timing differences, the timing structures are introduced and are derived using bottom-up clustering. In our method, multiple cameras capture image sequences to analyze human gait motions. A 3-D human articulated model is fitted to the reconstructed 3-D shape of a walking person. The 3-D articulated model is tracked by evaluating the space between the 3-D model and the reconstructed 3-D shape. Feature vectors of selected individual body parts are segmented by performing bottom-up clustering on linear dynamical systems. The timing structures of a walking person are described. Experimental results show that the proposed method is effective for timing difference analysis, and the timing structures can represent the individuality.