A calibration method for a structured light system by observing a planar object from unknown viewpoints is proposed. A structured light system captures a 3D shape by a camera that observes a light stripe on an object illuminated by a projector. The 3D shape, obtained from the system defined by a pinhole model for the projection of a light stripe, is solved using the equation of a plane model for the projector. The coefficients of each light stripe¿s equation are estimated using the 4×3 image-to-camera transformation matrix that is expressed by camera parameters. Experimental results demonstrate a high degree of accuracy when following the proposed approach.

A method of constructing a 3D model of human face by integrating multiple range images that were measured at different view directions is presented. The images were obtained by a high-speed rangefinder system which is capable of measuring the frontal face and capturing a texture image in approximately one second. A complete face model construction is thus performed by measuring the face in different orientations and integrating corresponding range maps using a least squares estimator. Range image integration is carried out in two ways: 1) by direct application of the least squares estimator to range values, and 2) by applying it to corresponding texture index values. A Sobel filter is applied to the intensity images to extract feature boundaries for the estimator. Both methods proved to have high accuracy returns and the average error of integration was estimated to be below 1.3 min.

Describes a high speed face measurement system and an algorithm for pose estimation of a human face using both color intensity and range images. The measurement system is designed so that we can obtain both color intensity and range images of a face without occlusion at high speed. Both color intensity and range images can be acquired within 37/60 seconds. The obtained face data are expressed as 3D wire-frame models with texture. Pose estimation is executed accurately by utilizing a property that the shape of face is almost symmetrical. Locations of the eyes and eyebrows are also detected through the pose estimation. The generated 3D face model and pose estimation results are shown. Both this system and technique are effective for 3D face modelling. Generated models are available for many applications such as virtual reality, man-machine interfaces and teleconferencing.

The extended-range time-shared code tracking loop for frequency hopped signals has been presented and analyzed under the assumption of white Gaussian noise at the input to the loop. The code tracking jitter performance of the loop has been derived and compared to that of the traditional time-shared loop. The results show that the presented loop can have a 5/2 times wider tracking range without the serious jitter performance degradation.

In this paper, a high accuracy, high density and high speed range finding method is described. The space encoding rangefinder has error on space sampling in principle. To improve this error, we introduce pattern shifting into laser scanning rangefinder. With this method, we can detect the projected pattern edge accurately, by taking advantage of the charge storing characteristics of a CCD device. Using this method we can obtain range images more accurate than normal space encoding with little increase in the measurement time. We can control the quantity of error with changing the angle of pattern shift. The realization of this method using the laser rangefinder and the simulated results in the computer are also described.

A compact and high-speed rangefinding system which is applicable to active robot vision is described. The principle of depth measurement is based on the space encoding method. Spatial coded patterns of light are generated by a semiconductor laser. A slit-ray of the laser is scanned by a rotating polygonal mirror and switched according to temporal switching patterns. Each pattern of light is generated, and its reflected image is taken by a CCD camera, in one image frame with no lag time. A 512/spl times/256 range map is obtained continuously every 0.2 seconds.

A new type of rangefinder is proposed, and a method for 3-D object shape reconstruction is described. The proposed rangefinder is compact and light and consists of a charge-coupled device camera, a galvano mirror, and a semiconductor laser. A slit-ray from the laser is scanned by the mirror and patterns of light are made by temporal switching of the laser. Range maps are obtained by the space encoded technique. A 512/spl times/256 range map is captured within 0.3 s. The rangefinder is installed at a tip of a robot manipulator, and takes range maps from various directions. Experimental results are described for range map integration and 3-D shape reconstruction.>

The authors present an adaptive on-line recognition system for recursive writing, which is referred to as Cursivewriter. The system is user-dependent and adapts to writers by updating their character templates. The recognition strategy is based on dynamic programming matching. The system was implemented on a personal computer and recognizes alphabets, numerals, Hiragana and Katakana as well as some symbols. The system performs real-time recognition and the recognition speed is 0.2 seconds per pattern.>

A vision system is proposed that has two differentiated visual fields, i.e. peripheral and central, and the fixation point to the scene is actuated actively. The proposed vision system is aiming at an active vision system that recognizes environment actively by gazing at each object of interest. To build this system, the authors developed a vision system with two CCD cameras of which fixation points are controlled by a computer. A fundamental vision model is proposed that consists of five independent neural network modules. Each module is educated individually and processes visual information independently. To signify the feasibility of the proposed vision model, some fundamental experiments were carried out.>

A system based on slit-ray projection method is presented that uses a nonscanning type image plane. It consists of a 2-D array of discrete photodetectors, attached to an individual memory element. The whole system is designed to latch positional information of a slit-ray into the memory when the corresponding photodetector is triggered by slitlike image passing across the image plane. Basic data for range computation are acquired during only one scanning of slit-ray. Trial implementation of this system has verified the method to be effective, showing data acquisition time within a few milliseconds.>

A method for the description and identification of curved objects is described in this paper. The shape of a curved object is considered as a surface which is an accumulation of horizontal section boundaries. This is described by the coefficients of a Fourier series expansion of the proposed representative function. Each complicated feature of the shape of the object is decomposed into five simple features: elongatedness, horizontal strain, section shape, torsion, and displacement. Differences of objects as patterns are defined as pseudodistances of corresponding features. Some essential properties such as completeness of the pseudodistance family are explained theoretically. Finally, experimental results of identification of wooden block models and a doll are displayed.

A method for the shape measurement of curved objects has been developed using multiple slit-ray projections and a turntable. The object is placed on a computer-controlled turntable and irradiated by two directed slit-ray projectors. An ITV camera takes a line image of the reflection from the object, and a computer calculates the space coordinates of the object surface. By using multiple projections, the total shape measurement of the object surface is attained accurately.