A method of processing image information is disclosed, for detecting the movement of an object with a simplifed system structure, and high speed treatment. The method is comprised of the steps of entry of a stereoscopic image of an object, extraction of edges of the object, and dividing it into segments based on the local feature of the image. On each segment, tracing points (sampling points) are selected, the inter-frame correlation of selected points is obtained, the three-dimensional coordinates of the points are measured, and thus the position and attitude of the object are detected, wherein the inter-frame correlation is performed by evaluating the similarity between frames based on correlation between the tangential direction of tracing points and the luminance value of the surface.
In a moving object detection system, an edge image is generated from captured images taken by CCD cameras and a moving object distance image indicative of a distance to a moving object (such as a human being) is generated by extracting pixel corresponding thereto. Then, pixels in the moving object distance image are summed to generate a histogram and a profile extraction region is set in the moving object distance image with its center line focusing on a position whose histogram is greatest, while the edge image is superposed on the moving object distance image to correct the center line of the profile extraction region such that the moving object is detected by extracting its profile in the region. With this, it becomes possible to detect the moving objects respectively even when the two or more objects are present in neighborhood.
In an information extraction apparatus of the present invention, a distance image input section inputs a distance image including a plurality of objects to be respectively recognized. An area division section divides the distance image into a plurality of areas according to position information of a plurality of blocks to respectively execute an active event. An image processing section recognizes the plurality of objects in the areas corresponding to the blocks actually executed the active event, and supplies the recognition result to a system of the active event.
After stereo image is inputted (S1), the tracing point corresponding to the silhouette of a curved object observed by the stereo image is selected on the basis of the information on the present position of the object (S2), using a three-dimensional geometric model. And the corresponding points on the silhouette corresponding to respective tracing points are sampled from the stereo image (S3), and their three-dimensional coordinates are measured (S4). Then, from the group of the three-dimensional coordinates of tracing points and corresponding points, the position/posture of the object concerned and the measurement error are determined (S5), to discriminate the error (S6). In case the error is not small enough, processes starting from (S2) are repeated, regarding the detected position/posture as a renewed present position, while, if the error is small enough, processes are repeated starting from (S1), regarding the detected position/posture as a renewed present position, thereby providing the technique for high-speed tracing of three-dimensional rigid body motion of an arbitrarily-curved and smooth faced object utilizing an object model, applying a stereo camera system as sensors.
An image processor comprising an image pickup portion to take in an image of an object, a reference point extracting portion that extracts a reference point to be used in object extraction from images taken by the image pickup portion,a corresponding point detecting portion that detects a corresponding point in other images, a three dimensional coordinates value calculating portion that calculates a three dimensional coordinates value of an object by using the absolute position and posture parameters calculated by using image of a reference solid and coordinates values of reference point and corresponding point, a contour point extracting portion that extracts a contour point to show contour of an object on the basis of coordinate values of reference point and three dimensional coordinate values, and an object display control portion that extracts an image of the object from images photographed by the image pickup portion based on coordinates value of contour point and displays it on a display device.
A moving picture processing apparatus comprises an image capture unit for taking in time-series images (moving picture series) of a target object, a feature point extraction unit for extracting feature points constituting the surface of the object whose model is to be made, a three-dimensional position information extraction unit for finding three-dimensional position information on the feature points by correlating the feature points contained in the image at a point in time with those at another point in time in the time-series images and analyzing positional coordinate information on the correlated feature points, a three-dimensional patch group generating unit for generating three-dimensional patch groups constituting the surface of the target object on the basis of the feature points for which the three-dimensional position information has been found and supplying the three-dimensional patch groups as information on a model of the target object to an output section.
