Kenji Terabayashi

A simple indoor self-localization system using infrared LEDs

In this paper, a simple and robust method for indoor self-localization is proposed. Infrared LEDs are used as landmarks. They are set at known positions and observed by a CCD camera on a mobile robot. As a consequence, self-localization is carried out. The acquisition of the position and orientation of the robot with two or more LEDs is formulated. The nonlinear least squares method is applied. Experiments with a constructed experimental system evaluate the proposed method. Error analysis is performed by simulations. The proposed method realizes a very simple indoor GPS using only several infrared LEDs and a CCD camera.

Multi-object Segmentation in a Projection Plane Using Subtraction Stereo

We propose a method for multi-object segmentation in a projection plane. Our algorithm requires a stereo camera system called Subtraction Stereo, which extracts foreground information with a fixed stereo camera. The main contribution of this paper is how the image sequences that include partial occlusion of the foreground objects can be accurately segmented using mean shift clustering in real-time processing. The proposed method is suitable for inside a medium-sized environment, such as a room. Finally, we try to segment the sequences that include occlusion and show the accuracy of the proposed method.

Construction of a compact range image sensor using multi-slit laser projector and obstacle detection of a humanoid with the sensor

Detection of obstacles on a plane is important for a mobile robot that moves in a living space, especially for a humanoid that falls down even with a small obstacle. In this paper, a range image sensor for detecting small obstacles on a plane is constructed using a multi-slit laser projector. The sensor consists of a commercially available laser projector and a CCD camera. It measures a relative disparity map (RDMap) whose measurement errors are not affected by the distance. From the obtained RDMap, a plane is estimated using RANSAC and regions out of the plane are detected as obstacles. Experiments show that planes can be obtained with small errors in RDMaps with the constructed sensor, and that a humanoid with the constructed sensor can detect small obstacles such as a moving ping-pong ball and a LAN cable on a plane by the proposed methods while walking.

Fast and stable human detection using multiple classifiers based on subtraction stereo with HOG features

In this paper, we propose a fast and stable human detection based on “subtraction stereo” which can measure distance information of foreground regions. Scanning an input image by detection windows is controlled in their window sizes and number using the distance information obtained from subtraction stereo. This control can skip a large number of detection windows and leads to reduce the computational time and false detection for fast and stable human detection. Additionally, we propose two-step boosting as a new training way of classifier with whole and upper human body models. Experimental results show that the proposal is faster and less false detection than the method described in the reference [1].

Measurement of Three-Dimensional Environment with a Fish-Eye Camera Based on Structure from Motion - Error Analysis

This paper proposes a method for measuring 3dimensional (3D) environment and estimating camera movement with two fish-eye images. This method deals with large distortion of images from a fish-eye camera to calibrate internal and external camera parameters precisely by simultaneous estimation. In this paper, we analyze 3D measurement accuracy based on a theoretical model and evaluate it in practical analysis in experimental and real environments. These analyses show that the theoretical measurement error model works over a wide range of fish-eye views.

Registration of 3D geometric model and color images using SIFT and range intensity images

In this paper, we propose a new method for 3D-2D registration based on SIFT and a range intensity image, which is a kind of intensity image simultaneously acquired with a range image using an active range sensor. A linear equation for the registration parameters is formulated, which is combined with displacement estimations for extrinsic and intrinsic parameters and the distortion of a cameras lens. This equation is solved to match a range intensity image and a color image using SIFT. The range intensity and color images differ, and the pairs of matched feature points usually contain a number of false matches. To reduce false matches, a range intensity image is combined with the background image of a color image. Then, a range intensity image is corrected for extracting good candidates. Moreover, to remove false matches while keeping correct matches, soft matching, in which false matches are weakly removed, is used. First, false matches are removed by using scale information from SIFT. Secondly, matching reliability is defined from the Bhattacharyya distance of the pair of matched feature points. Then RANSAC is applied. In this stage, its threshold is kept high. In our approach, the accuracy of registration is advanced. The effectiveness of the proposed method is illustrated by experiments with real-world objects.

Gesture recognition based on the detection of periodic motion

In this paper, we propose a method to recognize periodic gestures from images. The proposed method uses a amplitude spectrum and a phase spectrum that are obtained by applying Fast Fourier Transform (FFT) to a time series of intensity images. FFT is applied to each pixel of low-resolution images. The method consists of 2 steps. First, the method detects p eriodic motion regions from the amplitude spectrum. Secondly, the method uses the phase spectrum in the detected periodic motion region to classify the gestures. The proposed method is robust to lighting conditions and individual differences in skin color because it does not rely on color information. Several experiments are performed to demonstrate the effectiveness of the proposed method.

Construction of a compact range image sensor using a multi-slit laser projector suitable for a robot hand

In this paper, a compact range image sensor used for short-range measurements is constructed with a multi-slit laser projector. Three-dimensional (3D) information obtained with a sensor is important for a robot that grasps an object. Sensors attached to robots for measurement are often hindered by occlusion by the hand immediately before an object is grasped. The constructed sensor is compact enough to be attached to a robots hand, and occlusion can thus be avoided Compactness of the sensor is achieved by using a small CMOS camera and a laser projector and setting the baseline length between them as short as possible. The short baseline length also enables measurement in a short distance. Some experiments verify that the constructed sensor obtains accurate 3D information of objects in a short distance.

Detection of Moving Objects with Removal of Cast Shadows and Periodic Changes Using Stereo Vision

In this paper we present a method for the detection of moving objects for unknown and generic environments under cast shadow and periodic movements of non relevant objects (like waving leaves), using a combination of non-parametric thresholding algorithms and local cast shadow analysis with stereo camera information. Good detection rates were achieved in several environments under different lighting conditions, and objects could be detected independently of scene illumination, shadow, and periodic changes.

Improvement of Human Tracking in Stereoscopic Environment Using Subtraction Stereo with Shadow Detection

The combination of subtraction stereo with shadow detection we propose improves people tracking in stereoscopic environments. Subtraction stereo is a stereo matching method which is fast and robust for the correspondence problem – one of the most serious issues in computer vision – restricting the search range of matching to foreground regions. Shadow detection gives adequate foreground regions of tracked people by removing cast shadows. This leads to accurate three-dimensional measurement of positions in stereoscopicenvironmenttracking. Byfocusing ondisparity images obtained by subtraction stereo, we can detect people easily based on standard labeling. Objects can also be measured directly in size by subtraction stereo without geometric information about environments for tracking. This is important for installing the tracking system easily. To track multiple passing people, we use the extended Kalman filter to address the occlusion problem usually encountered in crowded environments. The proposed method is verified by experiments using unknown stereoscopic environments.