Hideo Shimazu

Tasting robot with an optical tongue: real time examining and advice giving on food and drink

This paper describes the tasting robot with an artificial sense of taste. Its optical tongue examines the taste of food and drink within several seconds and gives the name as well as its ingredients without crashing them and without opening paper or plastic packages. It is the integration of infrared (IR) spectroscopic technologies with pattern recognition technologies. The robot is the worlds first tasting robot which gives advices on the food and on health issues based on the information gathered. It sees if foods are best to eat or if they are good for health from the point of the content of sugar and fat. The robot was first exhibited in EXPO 2005 AICHI, Japan.

Multi-resolution Analysis based Salient Contour Extraction

Detecting salient contours in complex backgrounds is important in image analysis and scene understanding. The local context of an edge or line segment feature is commonly used to measure its saliency degree as a part of the object boundary. However, traditionally the context information is captured by studying several features in the predefined neighborhood. In this paper, a novel salient contour extraction algorithm based on the multi-resolution analysis is proposed and a new saliency measure is defined to characterize the significance of feature. Relation of features that are corresponding to the same part of the object boundary across resolutions is utilized to estimate the context information and feature significance value. Experimental results show that the proposed method can extract salient contours more efficiently than center-surround interaction based methods and still provide robust results

Geometric Primitives Detection in Aerial Image

Geometric primitives are important features for aerial image interpretation, especially for understanding of manmade objects. With the increasing resolution of aerial image, growing size and complexity of image make it more difficult to efficiently extract dependable geometric features such as lines and corners. In this paper, we propose a novel linear feature extraction approach called trichotomy line extraction. According to the knowledge of geometric properties of interested objects in aerial image, i.e. manmade objects, a rule is designed to remove line segments meaningless for boundaries of interested objects. Then line updating is carried out based on spatial and geometric relation between lines, to improve connectivity of boundary lines and also to extract corners on object boundary. Experiment results show that proposed line extraction method can perform efficiently with accurate linear features of objects in large aerial image and meaningless line segments removing process is effective to improve the geometric features description of object and to reduce computing burden of following step

RealScape: Metropolitan Fixed Assets Change Judgment by Pixel-by-pixel Stereo Processing of Aerial Photographs

The Tokyo Metropolitan Government, the largest municipality in Japan, routinely conduct s building-change identification work. Recently, Tokyo terminated its traditional visual identification work, which had been used for 20 years, and shifted to a new automated system. This paper introduces the Fixed Assets Change Judgment (FACJ) system and its core tool, RealScape. RealScape automatically detects changes in the height and color of buildings based on three-dimensional (3D) analysis of aerial photographs. It employs a unique pixel-by-pixel stereo processing method and enables a foot-level precision for each building. RealScape detects building changes more accurately than visual judgment operations by humans and reduce s the labor costs to one third of the traditional approach a nd the required judgment duration to about two weeks per 100 km 2 .

A New Multiscale Line Detection Approach for Aerial Image with Complex Scene

Straight lines are important geometric features for aerial image understanding tasks like man-made object detection. As image scene becomes more complex, traditional method like Hough transform may produce false detections and cannot work efficiently. In this paper, the authors propose a new multi-scale line detection approach that can efficiently detect semantic lines in aerial image with complex scene. Firstly, a method called trichotomy line extraction detects reliable line segments locally. Then multi-scale image system is constructed by wavelet decomposition, from which global information is obtained to detect semantic lines. Experimental results show that proposed method can extract accurate linear features on complex scene aerial images in a robust and efficient way

Personal robot with a sense of taste

This paper describes an artificial sense of taste and the world’s first “Tasting Robot”, which was developed under support by the New Energy and Industrial Technology Development Organization (NEDO) for the Robot Project: Prototype Robot Exhibition at EXPO 2005 Aichi, Japan. The robot has a sense of taste based on an optical tongue concept. The optical tongue is the integration of infrared (IR) spectroscopy with pattern recognition technologies.