JPS623585A - Stereo visual device - Google Patents

Abstract

PURPOSE:To easily identify the color of an object without lowering the accuracy of a stereo visual process by detecting the corresponding point out of plural pictures regarding the same object image-picked up and inputted respectively through each TV camera on which filters having different color characteristics with each other are loaded. CONSTITUTION:On each of TV cameras 1a and 1b, for example, a filter 5a of a thin blue or a filter 5b of a thin red is respectively loaded and at right and left picture memories 4a and 4b, a picture image of thick blue component or that of thick red component in an object 2 is stored respectively. An edge detection circuit 6 detects respectively the edge component of the image of the object 2 in each picture. A stereo visual process circuit 10 performs the line segment resolution of the edge component of each picture detected at the edge detection circuit 6 by every straight line unit. A three-dimentional coordinate calculating part 8, according to the relation of corresponding line segment, calculates respectively a three-dimentional coordinate position at the two end points of each line segment by a trigonometrical survay method and a color indentification part 9 identifies the color information of the object 2 with respectively calculating the average density of both picture parts that represent the same portion of the object.

Description

DETAILED DESCRIPTION OF THE INVENTION [Technical Field of the Invention] The present invention relates to a stereo visual apparatus that can easily identify color information of a subject in stereo visual processing.

[Technical background of the invention and its problems]

Recently, stereo vision processing has been widely introduced as an object recognition technology. In this technology, the same subject is viewed from multiple different positions, for example, two TV cameras lined up on the left and right, and two me large inputs are received by each TV camera.
The system detects the same corresponding points from the two images, and recognizes the position (stereoscopic) of the subject in three-dimensional space by triangulation based on the positional relationship between the two TV cameras.

By the way, when identifying the color information of a subject in this type of stereo vision processing, conventionally a color TV camera is used,
We are getting the color information. When such a color TV camera is used, color information of a subject can be obtained with high precision. However, for image processing, for example, a large amount of image memory is required to process each color image of R, G, and B obtained by color separation of two color captured images, and the amount of image processing becomes extremely large. Then, there was a problem.

Moreover, when a color TV camera is generally used, its resolution is degraded compared to a monochrome TV camera, resulting in a problem that stereo viewing processing accuracy (three-dimensional coordinate measurement performance) is degraded.

[Purpose of the invention]

The present invention has been made in consideration of these circumstances, and its purpose is to provide a stereo visual device that can easily identify the color of a subject without causing a decrease in stereo visual processing accuracy. It's about doing.

[Summary of the invention]

In the present invention, filters with mutually different color characteristics are attached to a plurality of TV cameras that are provided at predetermined intervals and view the same subject, and the same image is input via each of these TV cameras. Corresponding points of the subject are detected from a plurality of images related to the subject, for example, by decomposing the edge components in each image into line segments, and detecting corresponding line segments from the connection relationship of each line segment in the plurality of images. , the distance to the point is determined by triangulation using the information of this corresponding point, and the relationship of the brightness of the same part of the subject in each of these images found by the corresponding point detection, that is, the relationship between the brightness of each image. The color information of the object is identified from the relationship between the brightness information of the object and the color characteristics of the filter used to obtain the image.

〔Effect of the invention〕

Thus, according to the present invention, it is possible to obtain a high-resolution image without color separation, perform stereoscopic processing on this image, and detect corresponding points with high precision. Based on the relationship between the obtained color (luminance) information of the image, it becomes possible to easily identify the color of the subject.

In other words, the color information of the subject in each input image (11
1 degree information) changes at a predetermined ratio depending on the color of the subject, the color of the subject can be easily identified from the relationship of the color information of the subject in each image,
A great practical effect can be achieved in object recognition processing.

[Embodiments of the invention]

Hereinafter, one embodiment of the present invention will be described with reference to the drawings.

FIG. 1 is a schematic configuration diagram of an embodiment device. The two TV cameras 1a and 1b are arranged in parallel with a predetermined spacing between them, and are provided so as to view the same direction and simultaneously take Wi images of the subject 2. Each of these TV cameras 1a and 1b captures and inputs the subject 2 as a high-resolution monochrome image (multivalued), and the input image is sent to the left side via A/D converters 3a and 3b, respectively. Image memory 4
a and the right image memory 4b.

The TV cameras 1a and 1b are respectively equipped with filters 5a and 5b having different color characteristics, for example, a light blue filter 5a and a light red filter 5b, and the left image memory 4a is equipped with filters 5a and 5b having mutually different color characteristics, for example, a light blue filter 5a and a light red filter 5b. An image of the subject 2 with a strong blue component is stored in the right image memory 4b, and an image of the subject 2 with a strong red component is stored in the right image memory 4b. The edge detection circuit 6 receives the information of the left and right images respectively stored in the image memories 4a and 4b in this way, and detects the edge components of the image of the subject 2 in each image. This edge detection is performed, for example, by detecting a pixel portion in an image where the luminance changes sharply as a connection between the pixels.

As shown in FIG. 2, for example, the stereo vision processing circuit 10, which includes a corresponding point detection section 7, a three-dimensional coordinate calculation section 8, a color identification section 9, etc., is first composed of the edge detection circuit and the like. In step a), the edge components of each detected image are decomposed into line segments, and corresponding line segment candidates between the images are detected by referring to the coordinates of both ends of each line segment (
Step b).

Specifically, two TV cameras 1a arranged horizontally in parallel.
, When a rectangular parallelepiped object 2 is imaged and input in 1b,
Images obtained by decomposing the edge components in the leftward @ and right images into line segments are obtained, for example, as shown in FIGS. 3(a) and 3(b). Here, line segments that may correspond to line segment a1 in the left image are line segments bl, b2. b
3 is calculated, but since the coordinate position of line segment b2 is significantly different in the vertical direction, the corresponding line segment candidate is line segment bi
, b3 is obtained.

Similarly, line segment candidates corresponding to line segment a2 in the left image are line segments 1), 4, and 4 in the same direction in the right image. b5. b
The line segment b4 in 6 is determined as the corresponding line segment. The corresponding point detection unit 8 mutually compares and determines the relationship in which each line segment corresponds without contradiction according to the relationship between the corresponding line segment candidates and the connection relationship between a plurality of line segments in each image, and determines the relationship between the subjects in the left and right images. The corresponding line segments are detected and determined (step C). For example, in the illustrated example, line segment a1 has line segment b.
1 corresponds to other line segments a1. a2. bl,
It is determined from the connection relationship with b2.

The three-dimensional coordinate calculation section 9 calculates the three-dimensional coordinate positions of both end points of each line segment according to the triangulation method according to the relationship of the corresponding line segments, thereby recognizing the object shape of the subject.

The color identification unit 10 also includes an area surrounded by each of the corresponding line segments,
In other words, the average density of both image parts representing the same part of the subject is calculated in step e), the relationship between the average densities is compared with each other, and, for example, the average density pair of the left and right images and the corresponding color information are stored. The color information of the subject 2 is identified using a ROM that has been prepared (step f).

That is, as mentioned above, stereo y! The above-mentioned two images subjected to processing are images captured and input through a light blue filter 5a and a light red filter 5b, respectively. For this reason, for example, when the subject 2 is red, the left image 811111 input through the light blue filter 5a is obtained as an image with low brightness. On the other hand, the right image input through the light red filter 5b is obtained as an image with high brightness. In other words, as a monochrome image, an image with low brightness on one side and a high brightness on the other side is obtained. Such a relationship is determined by the color of the subject 2 and the color characteristics of the filters 5a and 5b. Therefore, from the relationship between the brightness and brightness of the corresponding area and the color characteristics of the filters 5a and 5b, it is possible to determine whether the subject 2
color information can be identified to some extent. In other words, although strictly speaking color identification is required according to the principle of three primary colors, it is possible to identify colors such as blue, red, green, and yellow using only the above-described method.

Thus, according to the present device, it is possible to perform high-precision stereoscopic processing from a high-resolution image input as a monochrome image, perform stereoscopic recognition, and easily and effectively identify the color of the subject 2. . Moreover, unlike when a color TV camera is used, there is no need to prepare image memories corresponding to each color component, and the amount of information to be processed does not increase. Therefore, it is possible to identify color information without complicating the system configuration, and it is possible to perform high-speed processing, which provides great practical effects.

Note that the present invention is not limited to the embodiments described above. For example, after detecting corresponding points from an image with the color filter removed, the image with the color filter attached may be input to identify color information.

Furthermore, when detecting corresponding points, not only edge information but also image density information may be used.

Of course, it is also possible to configure the apparatus by using one camera as a monochrome camera and the other as a color camera. Furthermore, the imaging characteristics of the plurality of cameras may be made to be different, so that the Ia image latitude with respect to the subject may be varied. In addition, the present invention can be implemented with various modifications without departing from the gist thereof.

[Brief explanation of drawings]

FIG. 1 is a schematic configuration diagram of an apparatus according to an embodiment of the present invention, FIG. 2 is a diagram showing the flow of processing in the embodiment apparatus, and FIG. 3 is a diagram illustrating an example of an input image. 1a, 1b...TV camera, 4a, 4b...image memory, 5a, , 5b...filter, 6...edge detection circuit, 7...corresponding point detection section, 8...3-dimensional Coordinate calculation section, 9...color identification section.

Claims (2)

[Claims] (1) A plurality of TV cameras equipped with filters with mutually different color characteristics are installed at predetermined intervals to view the same subject, and multiple images of the same subject are input through each of these TV cameras. The corresponding points of the subject are detected from each image, and the average density of the area surrounded by the corresponding line segment in each image is calculated according to the corresponding point detection result, and from the relationship of these average densities, it is possible to calculate the corresponding points of the subject. A stereo visual device characterized by identifying color information. (2) Detection of corresponding points of a subject in multiple images is performed by decomposing each edge component in each image into line segments and detecting corresponding line segments from the connection relationship of each line segment in the multiple images. A stereo visual device according to claim 1.