JPS6167187A - Image processor - Google Patents

Abstract

PURPOSE:To facilitate the estimation of relative position to an object to be chacked from the wide field of view by providing a reference mark having an optional characteristic. CONSTITUTION:Picked-up data from ITV camera are A/D converted through an image converted unit 2 and stored in a memory 3. An extractor 4 extracts features of a cross mark based on the image data received the wide field of view including the preset reference cross mark through the conveter 2. Each reference pattern for detecting 4 corners of the cross mark is stored preliminarily and the whole image is compared by a sequential pattern matching unit and the 4 matching points 9 are extracted. A discriminator 5 cuts the reference mark 7 from these points 9 to calculate the coordinates of center of the mark and a displacement information computer 6 estimates the relative position of the object and the camera from the coordinates of center of mark.

Description

[Detailed Description of the Invention] [Field of Application of the Invention] The present invention relates to an image processing device, and in particular to an image processing device suitable for obtaining positional information for cutting out a target object from a wide field of view. It relates to a processing device.

[Background of the invention]

In conventional image processing apparatuses, an object is cut out using a fixed image input device, as shown in, for example, Japanese Patent Laid-Open Nos. 58-87668 and 57-10883. A method of cutting out a target object from an area and a figure cutting method based on a mark, which is relatively used for figure processing, are known. However, in recent years, the scope of application of image processing has rapidly developed, and it is now being expanded to the field of monitoring objects, conditions, and the like. For this reason, in order to extract an object from a fixed imaging area using a conventional image input device that is firmly fixed, it is necessary to easily extract the desired object from a wide field of view. .

[Purpose of the invention]

SUMMARY OF THE INVENTION An object of the present invention is to provide an image processing device that extracts positional information in order to extract only the minimum necessary target object from a wide field of view using a highly efficient search method.

[Summary of the invention]

The gist of the present invention is to set a reference mark with a unique characteristic point in a specific positional relationship to an object, use the mark as a reference point for a wide field of view, and then Detect the aforementioned mark from the image obtained by inputting the image, extract the coordinate position of the mark in the captured image, recognize the current position coordinate of the image input device with respect to the reference point, and calculate the distance to the target object. The objective is to estimate the position, calculate the amount of movement of the image input device to capture the object at the center of the screen, and output position information for cutting out the desired object.

[Embodiments of the invention]

The present invention will be explained in detail below.

FIG. 3 shows the relationship between the object 11.12 and the fiducial mark 7. FIG. The reference mark 7 is a reference point in a wide field of view to be monitored, and is set at a certain fixed position relative to the target object 11, 12. Also, mark 7
Due to the wide field of view and background noise, shapes without similar feature points and standard patterns can be minimized, and can be easily captured without depending on the distance between the object 11 or 12 and the camera 10 at the time of imaging. It is desirable to have a mark 7 that can be pattern matched as a feature point, and further, in order to make it possible to easily extract the mark 7 described above, the structure is designed such that the border that differentiates the contrast between the mark and the background is devised.

With reference to FIG. 1, the configuration and operation of a preferred embodiment of the present invention will be explained. The image input unit 1 is an industrial television camera using an MO8 type image sensor as an image sensor, and converts image data of an object to be recognized into an electrical signal. next,
The image changing section 2Fi performs A/D conversion on the electric signal and stores it in the image memory 3 as a digital image. The image memory 3 can be implemented with several memories as required, and has two independent memory: - @light memory with one bit per pixel, and - binary image memory with one bit per pixel. can be expanded to a maximum of four sides, 512 x 512 x 8 bits as a gray level memory, and 512 x 512 x 1 bit as a binary memory. The reference mark extraction unit 4 extracts feature points of a cross based on the image data inputted by the image conversion unit 2 from a wide field of view including reference marks set in advance.

This feature point extraction method is based on the existence of an object with certain characteristics.
This is a technology that recognizes the position coordinates of objects in images. This time, we applied the pattern matching method as the most basic method to clearly capture the feature components in an image. This method consists of 8
A binary standard pattern of x12 pixels is compared with the input image, and the result is 0''.

This is a method of finding the point with the largest number of matching points in the pattern '1. Because the size of the ° mark changes depending on distortion, noise, and the distance between the object and the camera, standard patterns vary, and to satisfy this requirement, a huge database would be required and the processing algorithm would be complicated. applied this pattern matching method.

We attempted to put it into practical use by using the local mouth point pattern matching method. Figure 2 shows each matching location. This pattern matching method stores each standard pattern internally in advance to detect four elements, and sequentially
Performs pattern matching and compares the entire screen,
Extract four matching points 9.

Next, the reference mark detection and determination section 5 and the reference mark extraction section 4
For the four matching points extracted by Vc, X! The rationality of the coordinate positions of the four points is checked by considering the allowable deviation amount in the Y coordinate direction. This results in
A reference mark 7 is cut out from each matching point, and the center coordinates of the mark are determined.

On the other hand (if an abnormality is detected in the rationality check, the threshold conditions are changed to some extent, adjustments are made to capture the mark more clearly, and the reference mark 7 is extracted again. In order to estimate the relative position between the object and the camera from the center coordinates of the reference mark 7 extracted by the mark extraction unit 4, the size of the reference mark 7 captured on the screen is measured and compared with the existing value. From the ratio, find the actual size (pixel size) indicated by one pixel.On the other hand, the camera posture that captures the field of view including the reference mark 7 is determined by capturing the preset reference mark 7 at the center of the screen. A certain positional relationship is established, but depending on the positioning accuracy of the moving mechanism on which the camera is mounted, it is not always possible to obtain the specified angle (the point at which the reference mark is captured at the center of the screen).
Therefore, by determining at which coordinate position the center of the mark exists with respect to the center point of the image from an image that captures the field of view including the reference mark 7, the position at the time of imaging can be determined. The deviation width is measured, and the amount of error during movement is determined from the determined pixel size and positional deviation width. This allows the camera to accurately capture the target object within the screen (capture it at the center of the screen) based on the amount of error and the existing positional relationship, without being affected by stopping errors, etc. Calculate the moving scene.

According to the invention, a fiducial mark with certain characteristics can be created.

【Effect of the invention〕

According to the present invention, it becomes easier to cut out a target object from a wide field of view, and the amount of information handled by the image processing device and
It minimizes database capacity and does not require complex image pattern recognition processing.

[Brief explanation of drawings]

FIG. 1 is a system block diagram, FIG. 2 is an explanatory diagram of matching points in turn matching, and FIG. 3 is a conceptual diagram of the positional relationship between a reference mark and an object. DESCRIPTION OF SYMBOLS 1... Image input part, 2... Image changing part, 3... Image memory, 4... Reference mark extraction part, 5... Reference mark detection determination part, 6... Movement information calculation part .

Claims (1)

[Claims] 1. An image input device that captures a wide field of view to be monitored as a video signal, the video signal from this image input device is converted into a digital image signal, the digital image is stored in an image memory, and various feature points are extracted. It consists of an image processor for extraction, and a control device that estimates the position of the object from the captured image data, and sets a unique reference mark with a specific positional relationship in a wide field of view to be monitored. An image processing device comprising: a calculation mechanism that extracts the reference mark from the image data including the reference mark and estimates the amount of movement of the image input device to the monitoring target.