JPS603273A - Image binarization threshold setting method - Google Patents

Abstract

(57) [Summary] This bulletin contains application data before electronic filing, so abstract data is not recorded.

Description

[Detailed Description of the Invention] [Field of Application of the Invention] The present invention provides a method for appropriately setting a threshold value according to illumination conditions when binarizing gray scale (multivalued) image data in an image processing system. This is a thresholding film value setting method for the image Z iii.

[Background of the invention]

Multivalued image data is usually treated as 6 to 8 bit digital data, and various image processing is performed on it.

In binary image processing, a threshold value is required to binarize the multivalued image data as described above. The threshold value varies greatly depending on lighting conditions, object surface conditions (color, etc.), optical system, sensor sensitivity, and the like.

When the above conditions are constant, the same threshold value can be used for different multivalued image data, and an online real-time image processing system can also be constructed. However, it is generally difficult to control lighting conditions to keep them constant.

Therefore, if illumination conditions cannot be controlled, it is necessary to reset the threshold value to an appropriate value each time image data is captured. This must be done automatically and quickly in an online real-time system.

In general, two-value image processing is suitable for performing binary image processing.
Regarding the distribution of the brightness (multivalued data) of each pixel of a dimensional multivalued image, when there is one peak in each of the bright and dark parts, it is possible to Correct binarization can be performed by setting a binarization threshold.

Such setting conditions can be determined through statistical analysis, and conventional methods have exclusively used them as a method for automatically determining the threshold value.

However, the statistical analysis method described above is
I Wl'' Since the amount of calculation is very large, in order to read out all the image data, for example, the image size must be 256 (pixels)
Even in the case of 256 (pixels), a huge number of memory accesses of 256×256 times were required, making online real-time image processing difficult.

[Purpose of the invention]

The object of the present invention is to eliminate the drawbacks of the prior art as mentioned above and
Image 2 that allows you to appropriately set the threshold value in a short time according to lighting conditions when converting multivalued image data into binarization
The object of the present invention is to provide a value threshold setting method.

[Summary of the invention]

The configuration of the image binarization threshold setting method according to the present invention is such that, in an image processing system that binarizes multivalued image data and performs the binary image processing, a mark fixed at a fixed position within the field of view is provided, and the above-mentioned The relationship between the brightness of the above-mentioned sign and the corresponding appropriate binarization threshold value is determined according to the illumination conditions within the field of view, and the brightness of the sign is determined based on the above-mentioned relationship each time when performing binary image processing. An appropriate binarization threshold value is set according to the above-mentioned field of view, and the multi-valued image data of the article to be imaged within the field of view is binarized.

Note that this will be supplemented and explained below.

The appropriate binarization threshold for binarizing multivalued image data changes depending on the lighting conditions, but in the conventional method, as mentioned above, it focuses on the brightness distribution of all pixels in the screen,
The binarization threshold was determined through extensive calculations.

In the present invention, the brightness of illumination related to changes in illumination conditions is measured, and an appropriate threshold value is determined based on this measurement. In other words, in order to measure the brightness of the illumination, an IT-J- sign is always placed at a certain position within the field of view that should be on the screen, and basic pixel data at several points on the sign is referred to. The brightness of the illumination at the time of inputting the multivalued image data is determined.

By determining the relationship between illumination brightness and appropriate binarization threshold in advance, it is possible to immediately determine an appropriate binary threshold corresponding to the lighting conditions.

Appropriate binarization thresholds have a certain range, and the range of acceptable values is also limited by the domain of pixel values (i.e., image information), but it is usually a relatively small value between 0 and 255. limited to a narrow range.

Therefore, the relationship between the above-mentioned spacing and the appropriate binarization threshold value can be easily expressed by a simple function.

[Embodiments of the invention]

Embodiments of the present invention will be described below with reference to the drawings.

FIG. 1 is an explanatory diagram of an example of an article to be subjected to image processing;
Figure 2 shows the brightness on a straight line on the surface and the 2 that should be set.
An explanatory diagram of the permissible variation range of the digitization threshold; FIG. 3 is a relationship diagram of an example of the relationship between brightness and the binarization threshold to be set according to the image binarization threshold setting method according to the present invention; Similarly, FIG. 4 is a configuration diagram of an embodiment of the image processing system.

Here, 1 is a conveyor that conveys a part 2 to be subjected to image processing, 3 is a sign, 4 is a ′]゛V camera, 5 is an image processing device, 6 is its multivalued image input unit, and 7 is the same. 8 is a threshold value setting section, 8 is a component image binarization section, and 9 is a component image processing section.

1 (a) is a front view, and (b) is a side view), the parts 2 as shown in FIG. It will be transported. It is assumed that this is picked up by a robot or the like and assembly work is performed.

In this case, the robot has the position of part 2.

Visual ability to detect posture, etc. is essential.

The image processing should be performed by binary image processing, since the surface of the conveyor 1 is horizontal and the shape of the part 2 means that changes in its posture are limited to within a two-dimensional plane. .

By the way, the screen size is 256 (pixels) x 256 (i
! ! ii) The number of brightness levels is about 64 or 256, and the part 2 has a front side with text display such as product name and a back side without such display, and it is necessary to identify the front and back sides.

Figure 2 OI) to (f) are straight lines AA' shown in Figure 1.
The brightness for each point (horizontal axis) above is plotted under the lighting conditions of each building.

The levels P1 to P6 of each town indicate the brightness of an arbitrary area that is assumed to have uniform brightness under each illumination condition, but there may be some variation in pixel values within that area. If so, take the average value for several points. In addition, in the case of Figure 2, levels 1) 1 to P6 of each town are as follows:
This is the value of the lightest (for example, white) part of the surface of the component 20.

For each level waveform diagram as shown in FIG. 2, in order to identify a character (ie, the front side of part 2), the threshold value must be within each tolerance IIi range ΔP1 to ΔP6.

For each of these allowable ranges ΔP1 to ΔP2, when binarizing a certain brightness 1〕 or more as 1'' and a certain brightness Pn or less as °'0'', the brightness P (the above P1 to P6 is It is desirable that the representative general values) be set in the following relationship.

P R (p < P A (]) This is essentially a problem of the S/N ratio of binarization.

Curve B in FIG. 3 corresponds to the relational expression % (2) between the brightness P and the appropriate binary threshold Q, which was determined experimentally in consideration of the above conditions. Using this relationship, an appropriate binarization threshold can be easily obtained.

In this case, when the brightness P<50, the noise level overlaps the permissible variation range ΔP of the binarization threshold Q, so it is impossible to set the binarization threshold Q appropriately. this is,
This is because the S/N ratio deteriorates extremely if the illumination is too dark, and the above equation (2) cannot be used under such conditions.

Here, the unit of brightness in the above equation (2) will be explained.

The value of each pixel on the screen has a monotonically increasing relationship with the intensity and brightness of light, but the relationship therebetween is not necessarily constant depending on the characteristics of the image sensor and the like.

Generally, in the field of digital image processing, the A/D conversion output (ti/tal value) of electricity (g (for example, voltage) after imaging is treated as "brightness", and in particular, the original measurement unit of brightness is It doesn't matter how it relates to "looks" etc. Therefore, lighting, lens system.

Since the image sensor etc. are collectively considered as an optical system and treated separately from digital value processing, pixel values are processed in relation to others, and calibration that forces their absolute values is not possible. There's no need to do it.

In such a situation, the above equation (2) includes a constant determined from a certain specific experimental equipment.

Next, this method will be explained based on FIG.

'1゛For the sign 3 fixedly provided within the field of view F of the V camera 4, set the appropriate binarization threshold Q and brightness in advance as shown in the curve B in Fig. 3 or the above equation (2)) P
(hereinafter referred to as the optimal threshold function),
This is stored in the threshold value setting section 7 of the image processing device 5.

The serial image signals sequentially output from the TV left camera are:
The multi-value image input unit 6 converts the image into a two-dimensional multi-value image array.

Since the relative position between the TV left camera and the sign 3 is fixed, the corresponding pixel fixed at a fixed position on the two-dimensional multivalued image array is extracted for the sign 3, and its brightness is obtained.

Since this sign brightness is given to the threshold value setting section 7, the appropriate 2
A valorization threshold is sent.

The component image binarization section 8 converts the image of the target component 2 into a binary image using the binarization threshold value, and the binary image processing section 9 converts the position of the component 2, based on the binary image. Performs posture recognition processing and outputs the parts recognition results. This allows a robot or the like to perform desired assembly work.

In this way, according to this embodiment, an appropriate binarization threshold can be set in a time that is negligible from the image processing process lead, so when high-speed processing is required in binary image processing. In addition, it is possible to automatically respond to changes in environmental conditions without special consideration of stabilizing lighting conditions.

As a result, the cost of controlling the illumination conditions can be reduced, and the illumination conditions can be adjusted to
Since the obstacles to the introduction of value image processing technology are overcome, it is possible to utilize and improve the functionality of two-frame image processing technology in fields such as robot vision technology, for example.

〔Effect of the invention〕

As described in detail above, according to the present invention, it is possible to properly set the binarization threshold value in multilevel image processing in a short time, thereby improving the economy and performance of this type of image processing system. In addition to improving reliability and reliability, remarkable effects can be obtained in expanding the scope of its introduction.

[Brief explanation of drawings]

FIG. 1 is an explanatory diagram of an example of an article to be subjected to image processing;
Figure 2 shows the brightness on a straight line on the surface and the 2 that should be set.
An explanatory diagram of each permissible variation range of the digitization threshold, FIG. 3 is a relationship diagram of an example of the relationship between the brightness and the binarization threshold to be set according to the image binarization threshold setting method according to the present invention, Figure 4 shows
FIG. 2 is a configuration diagram of an embodiment of the image processing system. 1... Conveyor, 2... Parts, 3... Sign, 4...
...i' V camera, 5... image processing device, 6...
Multivalued image input section, 7... Threshold value setting section, 8... Component image binarization section, 9... Binary image processing section. Agent Patent attorney Kosaku Fukuda (and 1 other person) Kaya 1 Illustration No. 2 Convict Kaya 3 Convict No. 4 Continuation of page 1 0 Inventor Yoshie Matsuzaki 292 Yoshida-cho, Totsuka-ku, Yokohama City Hitachi, Ltd. Production Technology Research Co., Ltd. Inside the office

Claims (1)

[Claims] 1. In an image processing system that binarizes multivalued image data and performs the binary image processing, a mark is fixed at a fixed position within the field of view, and the brightness of the mark is adjusted depending on the illumination conditions within the field of view. , set the appropriate binarization threshold value according to the brightness of the sign based on the above relationship, and set the appropriate binarization threshold value according to the brightness of the sign each time. An image binarization threshold setting method characterized in that multivalued image data of an article to be image processed within a field of view is binarized.