Image analysis – Applications – Target tracking or detecting
Reexamination Certificate
1999-01-26
2002-09-24
Boudreau, Leo (Department: 2721)
Image analysis
Applications
Target tracking or detecting
C382S190000, C348S169000
Reexamination Certificate
active
06456728
ABSTRACT:
BACKGROUND OF THE INVENTION
The present invention relates to an object detection apparatus incorporated into a robot, a toy, and a stuffed animal or doll, which runs in accordance with a person's motion and repeats its predetermined motion.
Conventionally, infrared rays, supersonic waves and sound have been used in order to move a toy or a stuffed animal or doll in response to a person's motion. The infrared rays are the same as those for operating an automatic door.
In the object detection apparatus employing the infrared rays, one approaches the apparatus and blocks one of the infrared rays emitted therefrom. In this apparatus, even when a person approaches a toy or a stuffed animal or doll without any intention of moving it, he or she blocks an infrared ray to cause the toy or the stuffed animal or doll to respond in vain. Since, furthermore, a predetermined distance is required between a source for emitting the infrared rays and that for receiving them, the apparatus cannot be miniaturized.
In the object detection apparatus using the supersonic waves, when a person approaches a toy or a stuffed animal or doll, the supersonic waves are generated from the toy or the stuffed animal or doll and reflected by the person to thereby detect the person who is approaching. Since, however, this apparatus also detects only whether the supersonic waves are blocked or not as in the case of using the infrared rays, even when the person approaches a toy without any intention of moving it, the toy responds in vain.
According to the object detection apparatus using sound, sound is sensed by a small-sized microphone attached to a toy and converted to an electrical signal to move the toy. As in the case of using infrared rays and supersonic waves, this apparatus detects only whether the sound is blocked or not. If, therefore, one makes a sound without any intention of moving it, the toy responds in vain.
On the other hand, in a self-running apparatus incorporated into a robot such as a cleaning robot which senses an obstacle and runs avoiding it, one or both of a sensing method using supersonic waves and a sensing method using image processing is employed.
According to the former method, the apparatus generates ultrasonic waves from two portions to measure a distance to an object or an obstacle therefrom, like the triangular surveying, with the result that the obstacle can be sensed with high precision, though this method is more expensive than a toy. In this method, however, the presence of the obstacle can be known but the size or shape thereof cannot be obtained. It is thus difficult to plan how to avoid the obstacle. Therefore, a map of the area where the robot moves is stored in advance, and the position of a sensed object on the map is assumed from the location and direction of the robot itself, thus making a plan for the robot to move. Consequently, the sensing method using the supersonic waves cannot be practiced so well in the area whose map is not stored in advance.
Moreover, there is a method of recognizing an object, existing in which direction a robot is headed, by analyzing an image captured by an image capture apparatus such as a CCD camera. In this method, first, an image of the object in the direction is captured by the image capture apparatus, and then pre-processing of removing an extra portion such as a background from the image and to cut out only the object to be recognized as an obstacle. After that, the shape of the object is recognized using the processed image.
The above pre-processing of cutting out the object to be recognized will now be described. In order to cut only the image of a target object out of the image captured by the camera, any difference between an image region of the object and that of the others is employed as a clue. There are two image cutting methods utilizing a variation in hue and utilizing a differential image. In the former method, a portion of an image having a wide difference in hue is extracted and subjected to thinning to extract an edge. In the case of a room, the edge corresponds to the boundary between a wall and a pillar and that between a door, a floor and a wall. These boundaries are inclined in the depth direction toward a disappearing point, but basically they are recognized as vertical and horizontal lines. When these lines are disturbed, it is determined that there is an obstacle other than the structures such as the door, floor and wall. It is however difficult to clearly distinguish the floor, wall and door from one another if their hues are close or they have conspicuous stains. Since a tone of color varies with lighting, it is difficult to cut out an image in a steady state. Since, furthermore, no lighting makes a captured image dark completely, it is difficult even for persons to recognize an object from its photograph taken in the dark.
There is another method of cutting an object to be recognized out of a captured image, in which a motion vector between frames of the image is calculated to analyze an object in motion. When the number of objects in motion is small, no problems occur; however, if a large number of objects are moving, the number of vectors is increased suddenly and so is a load for calculating the motion vectors between the frames, with the result that the objects cannot be detected in real time.
To compensate for the above method using only the image processing, it is employed together with a distance calculation method using supersonic waves. In this method, however, points are observed and thus it is difficult to acquire information for all images captured as a plane.
As described above, in the conventional method for analyzing an image captured by image capture means, the flow of the analysis and the information to be analyzed are fixed; therefore, when the image is varied with external conditions every moment, a load is applied to a specific processing section, and the information analysis cannot follow the variation.
One method for resolving the above problem is to use a high-performance computer and a high-speed transmission system so as to allow real-time object detection processing (e.g., 30 images per second) in spite of an increase in load. It is however difficult to incorporate such a high-performance computer into a small-sized apparatus such as a toy, let alone a robot.
In the above-described conventional object detection apparatus for detecting an object itself and its motion using image processing, an inexpensive and robust image analysis is difficult to conduct under the external conditions varying every moment. In order to do so, a high-performance computer and a high-speed transmission system are used and thus costs therefor are too high, and they can be used neither in a simple automatic apparatus having intensive functions of a robot, a stuffed animal or doll, and a toy which runs avoiding an obstacle or acts according to a person's motion nor in a wearable object detection apparatus for, when the blind walk, detecting an obstacle in place of their eyes.
When the foregoing object detection apparatus is incorporated into playing equipment such as a boxing game, a user is so absorbed therein that he or she is likely to hit and break the apparatus by mistake. Moreover, the user's hand or body partly falls within a limited image capture range in which an image can be captured and thus a normal image cannot be obtained.
BRIEF SUMMARY OF THE INVENTION
An object of the present invention is to provide an object detection apparatus which is capable of robustly detecting an object at high speed and which can easily be miniaturized, and a motion control apparatus and a pattern recognition apparatus both using the object detection apparatus.
In order to attain the above object, an object detection apparatus according to a first aspect of the present invention comprises image capture means for capturing a distance image indicating distance information to a subject and representing intensity distribution of light reflected by the subject, and detection means for detecting one of a
Doi Miwako
Morishita Akira
Numazaki Shunichi
Umeki Naoko
Boudreau Leo
Choobin Barry
LandOfFree
Object detection apparatus, motion control apparatus and... does not yet have a rating. At this time, there are no reviews or comments for this patent.
If you have personal experience with Object detection apparatus, motion control apparatus and..., we encourage you to share that experience with our LandOfFree.com community. Your opinion is very important and Object detection apparatus, motion control apparatus and... will most certainly appreciate the feedback.
Profile ID: LFUS-PAI-O-2858903