This paper presents a new hardware and software calibration technology to improve the process of automatic multiple projector calibration using photo detectors to support Spatial Augmented Reality. This research builds on existing calibration methods that employ point-based photo detectors by advancing the hardware with a larger surface area photo detector to locate sub-pixel position with structured light. This data is used to expand the Gray-coding algorithm with an additional step that improves the measurement. The novel hardware allows a sub-pixel position to be calculated and leveraged to improve alignment of multiple-projector environments. The results show the new approach improves position measurement from pixel accuracy using a point photo detector with Gray-code by an order of magnitude providing sub-pixel accuracy by leveraging the planar photo detector and additional algorithm steps.
We have been developing FoodLog, a multimedia system that enables users to maintain a record of their food intake simply by taking photos of their meals. In this paper, we describe the functions and various applications of FoodLog, rather than its technical details. FoodLog is beneficial not only for monitoring food intake but also for several other purposes. During the research and development of FoodLog and in collaboration with a nonprofit organization, we employed it to generate donations for children in African countries. We also outline some extensions to FoodLog that are currently under development.
An Active grid-based method for estimating pass regions from broadcast soccer videos is presented in this paper. It is assumed that the pass region has a high probability of the pass succeeding. In soccer matches, players discover pass regions based on previous and current player positions. In conventional methods, pass regions are estimated by applying Active Net to only a single frame of a soccer video. In the proposed method, Active grid is applied to three-dimensional data by which frames of the soccer video are connected with the temporal dimension. The proposed method then realizes robust estimation of pass regions based on multiple frames of player positions. The proposed method was applied to actual TV programs to verify its effectiveness.
In this paper, we propose a method for robust tracking of a moving finger in an image sequence. The method is suitable for application to our input interface system, which recognizes a moving finger in the air. The proposed method extracts edges from input images, and then estimates the position and rotation of a finger in the input images by matching points in a template to edges. The most remarkable feature of our method is that it also takes into account the presence or absence of edges in regions in the input images corresponding to the inside of the finger in the template for estimating. This makes it possible to estimate the position and rotation of a finger exactly in images with complex backgrounds. Our method successfully tracked a finger in several situations with an average processing time of 6.32 [ms/frame], and the finger was tracked with good accuracy.
This paper presents a new method to improve performance of SVM-based classification, which contains a target object detection scheme. The proposed method tries to detect target objects from training images and improve the performance of the image classification by calculating the hyperplane from the detection results. Specifically, the proposed method calculates a Support Vector Machine (SVM) hyperplane, and detects rectangular areas surrounding the target objects based on the distances between their feature vectors and the separating hyperplane in the feature space. Then modification of feature vectors becomes feasible by removing features that exist only in background areas. Furthermore, a new hyperplane is calculated by using the modified feature vectors. Since the removed features are not part of the target object, they are not relevant to the learning process. Therefore, their removal can improve the performance of the image classification. Experimental results obtained by applying the proposed methods to several existing SVM-based classification method show its effectiveness.
We investigated the individual differences in the use of binocular disparity and proposed a method for improving stereopsis in observers who do not perceive depth from disparity in 3D-graphic environments. In Experiment 1, non-stereoanomalous observers, aged 19-25 years, were asked to roughly evaluate the depth of 3D-graphcial stimuli containing binocular disparity and shading. The results of Experiment 1 showed 30% of the observers were pseudo-stereoanomaly who perceived depth only from shading. In Experiment 2, 60% of the pseudo-stereoanomalous observers were able to use disparity when they had to assess the depth concretely. In Experiment 3, all of the observers who participated in Experiment 2 learned to use the disparity information when retasked with roughly estimating again. These results suggest that quite a few people have difficulty in experiencing rich depth perception in current 3D-graphic environments. However, appropriate training procedures would improve their use of disparity information over the long term.