JPH05135181A - Abnormality detection device - Google Patents

Abstract

PURPOSE:To increase the execution speed of an abnormality detection system which uses image processing technique or reduce the manufacture cost. CONSTITUTION:When the number of successive picture elements regarded as logic '1' among binary-coded picture elements counted by a scanning counter 73 in single-time scanning exceeds a constant value, the area value of the picture elements of logic '1' is increased by an area counter 51 and an abnormality decision means decides the abnormality of image data according to the incremental value; while the picture elements of logic '1' are counted, picture elements of logic '0' are counted and if the continuity of the picture elements of logic '1' is ceased halfway, the output of the scanning counter 73 is reset by a scanning counter resetting means 74 by considering that a noise is mixed with the image data.

Description

Detailed Description of the Invention

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to an abnormality detecting apparatus for detecting a normal portion and an abnormal portion of a subject appearing in image data from a difference in luminance of pixels forming an image.

[0002]

2. Description of the Related Art FIG. 4 shows a conventional abnormality detecting device shown in Chapter 12 of "Latest Trend of Image Processing Algorithms" (Sho 63.9.5), New Technology Communications, Chapter 12, for example. Is a camera for image input, 2
Is an image input device for A / D converting an image captured by the camera 1, and 3 is grayscale data converted by the image input device 2,
Binarization processing for binarizing with a certain threshold value, 4 is noise removal processing for removing noise of the data binarized by the binarization processing 3, and 5 is noise removal by the noise removal processing 4. Two
An area counter process that counts the number of pixels having the value “1” in the value data, and 6 is an abnormality determination process that determines whether or not there is an abnormality based on the number of data counted by the area counter process 5.

Next, the operation will be described. The image data input by the camera 1 is A / D converted by the image input device 2. Next, the A / D-converted grayscale image data is binarized by binarization processing 3. If there is an apparently large opening in the brightness of the normal part and the abnormal part in the input image, a threshold value is set to the intermediate brightness between the normal part and the abnormal part to binarize the normal part, for example. It is possible to divide the "0" and the abnormal portion into two like a logical "1".

Next, noise removal processing 4 removes noise generated during image input, noise generated during binarization, etc. to prevent erroneous operation of the area count 5 of an abnormal portion due to noise.

Then, the area count processing 5 counts the number of pixels of logic "1" which is an abnormal portion. The abnormality determination processing 6 compares the area count value counted by the area counter 5 with a preset abnormality determination reference value,
Output an error.

Next, a case where this system is realized by software will be described. In FIG. 4, 1 and 2 perform image input and A / D conversion by hardware. The binarization process 3 in FIG. 4 compares with a threshold value, and outputs a logic "1" when the threshold value is exceeded and a logic "0" when the threshold value is less than the threshold value.

The noise processing 4 performs, for example, expansion and contraction processing as a noise processing method for binary data. Expansion and contraction processing is logical "0" or logical "1" by binarization.
The image converted into is represented by the following function.

[0008]

[Equation 1]

Then, noise removal processing of binary data is generally realized by performing the following two logical operations several times. (A) Expansion processing result p ₁ (i, j) p ₁ (i, j) = Q (i-1, j-1) + Q (i, j-1) + Q (i + 1, j-1) + Q (i -1, j) + Q (i , j) + Q (i + 1, j) + Q (i-1, j + 1) + Q (i, j + 1) + Q (i + 1, j) (b) shrinking process result p ₂ (i, j) _{p 2 (i, j) =} Q (i-1, j-1) × Q (i, j-1) × Q (i + 1, j -1) × Q (i-1, j) × Q (i- 1, j + 1) × Q (i 1, j + 1) × Q (i + 1, j + 1) In noise removal, the expansion process is performed a plurality of times and then the contraction process is performed a plurality of times depending on the size of noise to be removed. It is necessary to perform such treatment, and by performing such treatment, the noise can be removed.

Next, in FIG. 4, the area count processing 5 counts the pixels having a pixel value of "1" of the binary data after noise removal, and the abnormality determination processing 6 determines the abnormality.

When this processing is to be realized by hardware, FIG. 5 shows "industrial image processing" (Sho 63.
5.31), Shokoido, is a circuit diagram for cutting out into a 3 × 3 local area shown in Chapter 3, and is a block diagram when the noise removal processing 4 of FIG. 4 is implemented by hardware.

Reference numeral 31 in FIG. 5 indicates the output of the binarization processing.
The image data has a pixel value of "0" or "1". Reference numeral 41 is a line buffer for holding the image data 31 input in time series by one horizontal line,
The horizontal synchronizing signal 33 is input. The image data 31 of 31 is cut out into three columns by the latch 42.

Further, the data stored in the line buffer 41 under the control of the horizontal synchronizing signal 33 is divided into three rows. As a result, the data cut out in the 3 × 3 local area enters the address selection line of the memory 43.

By connecting each latch and the memory address line in this manner, only one of the 512 memory cells of the memory 43 can be selected depending on the state of "0" or "1" of the pixel of interest and the neighboring pixel. The memory cell of can be selected. Then, the processing result corresponding to the expansion and contraction processing is written in the corresponding memory cell in advance, and the image data to be processed is input, so that the expansion (or contraction processing) can be performed once.

In normal noise removal, noise removal is configured by vertically connecting the devices shown in FIG.
The data thus noise-removed is subjected to area counting in the area counting processing 5 of FIG. 4 and judged in the abnormality judging processing 6.

[0016]

Since the conventional anomaly detection apparatus is constructed as described above, in order to carry out such noise removal processing by software, expansion processing is performed according to the magnitude of noise to be removed. After performing a plurality of times, it is necessary to perform the contraction processing a plurality of times, and there is a problem in that it takes a lot of time for the noise processing.

Further, in order to carry out the removal processing by such noise with hardware, a memory for writing a pattern for noise removal and a line buffer for cutting out a local area are required, so that it is expensive to implement. There was a problem such as this.

The present invention has been made in order to solve the above problems, and can realize high-speed processing in software implementation, and also in hardware implementation, an expensive chip such as a memory or line buffer. The purpose is to detect abnormalities without using.

[0019]

An abnormality detecting device according to the present invention comprises a binarizing means for converting each pixel of image data converted into a digital signal into a binary signal of logic 1 or 0 in accordance with shading. A scanning counter that scans binarized image data and outputs a count value when counting a predetermined number or more of logical 1s at the time of counting each pixel, and a continuous logical 1 when a logical 0 count is determined while the continuous logical 1 is being counted by the scanning counter. Scanning counter resetting means for clearing the count value, comparing means for comparing the count value output from the scanning counter with an abnormality judging reference value for judging an abnormal portion in the image, and the comparing means for judging the abnormal portion. An area counter that counts the number of pixels of logic 1 and an abnormality determination unit that compares the area count value with a preset abnormality determination reference value and outputs an abnormality are provided.

[0020]

According to the present invention, when there are a certain number or more of the pixels regarded as logic 1 continuously among the binarized pixels counted by the scan counter in one scan, the area value of the logic 1 pixel is set. The area counter is incremented, and the abnormality determination means determines the abnormality of the image data based on the increment value. Further, while counting the pixels of logic 1, the pixels of logic 0 are counted, and the pixels of logic 1 continue. When the image quality is cut off in the middle, noise is mixed into the image data, and the scanning counter output is reset by the scanning counter setting means, so that noise is removed and the area is counted at one time by one scanning. Data can be accessed once for each pixel, and the processing speed can be increased.

[0021]

【Example】

Example 1. An embodiment of the present invention will be described below with reference to the drawings. In FIG. 1, binary data of abnormal values and normal values of 0 or 1 is input in the first to third areas as in the conventional case.
Next, 7 is an area count controller which controls the operation of the area count 5 so as to count the area of the abnormal portion while removing the noise of the binarized image.

In these processes, first, the input binary image data noise for one screen is removed, and then the area is counted. In these processes, the pixels of "0" or "1" are consecutive in the input image, The area counter is controlled so as to increment when a certain number or more continues.

On the other hand, when pixels in the input image do not continue for more than a certain run length, it is determined that the image is noise, and the area counting operation is not performed, so that the area can be counted while removing the influence of noise. Also, these processes
Binary data for one screen can be processed only by accessing and determining once.

In the abnormality determination processing 6, the abnormality determination value set in advance in the abnormal area value setting register 52 is compared with the area count value in the same manner as in the conventional case.
Abnormality is judged.

Next, the operation will be described. For example, FIG. 2 shows an embodiment in which the control 7 of the area count value in FIG. 1 is realized by hardware. In the figure, the scan counter 73 to which the image data 31 is input in synchronization with the dot clock 32 is incremented when the pixel value "1" is input, and the scan counter reset control 74 when the pixel value "0" is input. Is cleared to 0 by. As a result, the output of the scanning counter is continuously "1" in the scanning direction.
Outputs the number of consecutive pixels only when pixels are input,
Otherwise, "0" is output.

On the other hand, the scan setting register 72 holds run-length data for determining whether the pixel is an abnormal one when "1" pixels are continuously input in the input image, and the output of the scan counter 73. And the scan setting register value are compared by the comparator 71, and when the scan counter output is larger than the scan setting register value, the area counter 51 outputs a control signal to enable counting.

The content of the scan counter is cleared every time one scan is completed by the scan counter reset control 74 for inputting the horizontal synchronizing signal 33, and if "1" pixels are not continuously input for a certain set value or more, Since the operation does not count the area, even if the input image contains noise, it is possible to measure the area of the "1" portion while removing the influence. The area value thus obtained is used to determine whether or not there is an abnormal object as in the conventional case.
Further, the area counter reset control 75 receiving the vertical synchronization signal clears the area counter 51 to zero when the data for one screen is completed.

Example 2. In the above embodiment, the case where it is realized by hardware has been described, but the processing flow when it is realized by software is shown in FIG. First, data is input at the binary image input 7a, and the area counter and scan counter are reset at 76. It is determined at 7c whether the pixel value is "1". If "1", the scan counter is incremented at 7d.

At 7e, it is judged whether "1" is continuous for a set value or more, and if it is continuous, 7f is incremented, and it is judged as an abnormal object, and the area counter is incremented. In 7d to 7f, abnormal values continue,
In addition, those larger than the judgment value are regarded as abnormal, and those which are discontinuous are regarded as noise, and are cleared at 7 g.

It is 7 g that it is limited to repeating one scanning line. In other words, one scan line break is not considered to be continuous. The amount of data for one screen is determined in 7h. In 5a, the abnormal area count value is read, and in 6a, the abnormality is determined.

As described above, the software only accesses the image data once per pixel at 7c. After that, noise can be removed and area can be counted by judging the counter value. Therefore, the processing time for accessing the image data to the memory is reduced, and fast processing is possible.

Example 3. Next, another embodiment will be shown as a system. This is a fire abnormality detection system using an infrared camera. Images of a fire scene taken with an infrared camera show a large difference in brightness due to the temperature difference between the fire-occurring part and the normal part. Binarization is performed at an intermediate brightness level so that a difference in brightness appears, and a fire part is "1" and a normal part is "0".

In addition, when the fire is viewed as an infrared image, it is a lump of a high temperature part to some extent, and therefore it is continuous as pixels. Abnormality is defined as continuous for a certain amount or more, and normal is defined as less than that. If the abnormality exceeds a certain level, it is judged that a fire has occurred and an alarm is issued.

Example 4. It is also used to detect defects on the printed circuit board. For this, the base image is binarized and the number of pixels having a value of "1" or "0" of a normal pattern is held as a representative value. Next, it is possible to determine the presence or absence of a defect by binarizing the target substrate under the same imaging conditions as in the normal state, scanning the pixels, and comparing with a representative value.

[0035]

As described above, according to the present invention, since the line buffer and the memory for temporarily holding the image data by hardware are not used, the system can be constructed at low cost. In addition, when this device is constructed with software, the processing speed is improved because noise removal, area counting, and image data can be accessed only once.

[Brief description of drawings]

FIG. 1 is a configuration diagram of an image abnormality detection system according to the present invention.

FIG. 2 is a block diagram when the present invention is implemented by hardware.

FIG. 3 is a flowchart when the present invention is implemented by software.

FIG. 4 is a system configuration diagram of a conventional abnormality detection system.

FIG. 5 is a block diagram when a conventional abnormality detection system is realized by hardware.

[Explanation of symbols]

 2 image signal A / D converter 3 image data binarization processing 5 abnormal part area count processing 6 abnormality determination processing 7 control processing for counting areas 31 binary image data 51 abnormal area counter 52 abnormal area value setting register 71 Comparator 73 Scan counter 74 Scan counter reset control

Claims (1)

[Claims] 1. A binarizing means for converting each pixel of image data converted into a digital signal into a binary signal of logic 1 or 0 according to shading, and scanning the binarized image data to count each pixel. A scan counter for outputting a count value when counting a predetermined number of logic 1s or more, and a scan counter resetting means for clearing the count value of the continuous logic 1 when a logic 0 count is judged while the continuous logic 1 is being counted by the scan counter; Comparing means for comparing a count value output from the scanning counter with an abnormality judging reference value for judging an abnormal portion in an image; an area counter for counting the number of pixels of logic 1 judged to be an abnormal portion by the comparing means; An abnormality detecting device comprising: an abnormality determining unit that compares a count value with a preset abnormality determination reference value and outputs an abnormality.