JPH01207679A - Extracting method for background information - Google Patents

Abstract

PURPOSE:To improve the accuracy of measurement by extracting the number of '1's on a scanning line which is generated most in a predetermined time as background information in an image. CONSTITUTION:A video signal inputted to an input part 10 is outputted to a shift register 11 and a difference calculation part 12. The calculation part 12 receives video signals from the register 11 and input part 10 and outputs the absolute value of the difference. The output of the calculation part 12 is sent to a feature value updating part 14 through a threshold comparison part 13. The updating part 14 reads the right and left values of an area out of an area storage part 15, adds a feature value, and the scanning line feature value to a stored instruction part 19. The instruction part 19 reads the right and left values of the area and increases the value of a corresponding area in a feature value storage part 20 by one. After the processing of the instruction part 19, a picture element number counting part 21 passes control to a period counter 22 when values of a row (i) and a row (j) reach the picture element number of the right lower end of the screen. The measured value of the counter part 22 and the value of a cycle quantity storage part 23 are compared 22 and an extraction part 24 extracts the maximum value of each scanning line feature value of the storage part 20 as background information on each scanning line when the values are equal to each other, thereby outputting the information to a background information storage part 25.

Description

DETAILED DESCRIPTION OF THE INVENTION [Field of Industrial Application] The present invention relates to a method for detecting a moving object such as a vehicle or an aircraft, and particularly to a background information extraction method for extracting background information other than a moving object.

[Conventional technology]

When detecting a moving object by processing a video signal obtained from a television camera or the like, it is very important to recognize parts that are not moving objects, so-called background information, in order to obtain good detection accuracy. Therefore, conventionally, patent application No. 55-18459
1r vehicle presence position method" has been proposed. This method takes the difference between neighboring pixel values in the scanning line direction in a two-dimensionally sampled image, and depending on the magnitude of the difference, the image is divided into "1" and "0".
”.

Then, the number of "1"s in the same scanning line direction within the measurement area is taken as a feature value of the scanning line, and this feature value is used to detect the presence of a running vehicle. In this case, to extract the background value, the maximum value of the background value expected in advance is set to the threshold value A.
Set as . When a feature value on a scanning line that is smaller than this threshold value A is obtained, it is recognized as a background value.

In addition, the method of automatically setting the feature values of the scanning line described above,
y, Child Communication Society Journal Vol. J68-D No. 3
It was proposed in March 1985 in "Traffic flow measurement system using image processing method". This method calculates the maximum value of
and find the minimum value y. After that, the background value is X+α, y-α
(α and β are constants) and are sequentially measured.

[Problem to be solved by the invention]

However, since the above-mentioned "vehicle presence position detection method" sets the background information threshold value A in advance, accurate position detection is impossible if the background information changes from moment to moment due to weather conditions or the like. Further, depending on the operation of the threshold value A, there is a possibility that feature values of scanning lines that occur when a vehicle is present may also be mistakenly treated as background information.

In addition, since the "traffic flow measurement system using image processing method" sequentially measures the background value as a value sandwiched between X + α and y + β, the background value of the nth cycle is x -)
-nα, y -nβ.

In other words, if measurements are continued sequentially, the upper limit of the background value (X+nα
) gradually increased, and the lower limit (y-nβ) approached O, making normal measurement impossible.

In this way, there is no precedent for a method of effectively extracting feature values corresponding to background information using a method of processing scanning lines, and of automatically doing so.

[Means to solve the problem]

The background information extraction method of the present invention calculates the difference between adjacent pixel values in the scanning line direction in a two-dimensionally sampled image, and divides the image information into binary values of "1" and "O" depending on the magnitude of the value. In a method that processes the number of "1"s on the same scanning line in an image as an I\fgi value, when processing a signal that periodically generates a certain number of images at a certain time, The number of "1"s that occur most frequently on the scanning line within the image is extracted as background information within the image.

In addition, "1" on the scanning line that occurred most frequently within a predetermined time, except when objects other than the background were present.
Extract the number of images as background information in the image.

In addition, the number of "1"s that occur most frequently on the scanning line within a predetermined time is extracted as background information in the image, except when objects other than the background are present between the set scanning lines. do.

[Effect]

Calculate the difference between neighboring pixel values in the scanning line direction within the image,
The image information is binarized into ``1'' and ``0'' depending on the magnitude of the value, and the ``1'' on the scanning line that occurs most frequently within a predetermined time is extracted as in-plane background information.

Further, the already extracted background information is compared with the number of "1"s on the scanning line, and the background information is extracted except when an object other than the background is present.

Also, after setting the range between the scanning lines on the screen and recognizing whether there are objects other than the background within this range,
Extract background information.

〔Example〕

Next, the present invention will be explained with reference to the drawings.

FIG. 1 is a block diagram showing a first embodiment according to the present invention. In the figure, 10 is an input section for inputting video signals from a television camera, 11 is a shift register, and 12 is an input section 1.
1; a difference calculation unit that calculates an absolute value by subtracting 0 and the signal of the shift register 11; a threshold comparison unit that compares the calculation result of the 13-company difference calculation unit 12 with a predetermined threshold;
Reference numeral 4 denotes a feature value updating unit that updates the feature values on the scanning line based on the results of the threshold comparison unit 13; 15, an area storage unit that stores measurement areas; and 19, a feature value storage unit that updates the feature values on the scanning line based on the results of the threshold comparison unit 13; 20 is a feature value storage section that stores the frequency of occurrence of the scanning line feature value in each scanning line; 21 is a pixel number counting section that measures the pixel number of the screen; 22 is the period of the screen. 23 is a cycle storage unit that stores the number of screen cycles; 24 is a background information extraction unit that extracts background information from the feature value storage unit 20; 25
is a background information storage unit that stores the extraction results of the background information extraction unit.

Next, the operation will be explained. The video signals ai,j, which can generally be assumed to be arranged in a two-dimensional array of i rows and j columns, are input to the input section 10 after being quantized. The input section 10 sequentially receives the values corresponding to each pixel and inputs them to the shift register 11 and the difference calculation section 12, respectively. The difference calculating section 12 receives and receives video signals from the shift register 11 and the input section 10, and calculates the absolute value of the difference. The delay time from input to output in the shift register 11 means the interval between pixels to be compared on the same scanning line, and if the delay time is k, the output of the shift register will be alJ-. Therefore, the difference calculation section 12 outputs -ai j l to the threshold comparison section 13 for fat j-. The threshold comparison unit 13 compares the predetermined threshold TH with the result of the difference calculation unit 12, and determines if Iat j − to −as j1≧TH, rlJ
r Iat, 3-k ``I, J If l≦TH, then ``
0'' as the feature value Pij- of the pixel and the feature value update unit 14
send to When the feature value updating unit 14 counts the sum of “1” in the measurement area, it reads the left and right values (31yJt) of the area in the i row from the area storage unit 15 that stores the measurement area, and calculates that j,≦3− If k<J, the above-mentioned feature value Plj
-k is added to V1+Ptj-, and the scanning line feature value v
Update 1. When the feature value updating unit 14 completes the processing, it outputs the scanning line feature value Vl to the storage command unit 19. As mentioned above, the storage instruction unit 19 stores the left and right values (31+Jt
) is read and if J -k = J t, the value in the corresponding area of the feature value storage section 20 that stores the frequency of occurrence of the scanning line feature value Vi of each scanning line is incremented by 1.

The feature value storage unit 20 is a two-dimensional memory that stores the frequency of occurrence of the scanning line feature value vi in the i-th row of scanning lines as an element in the i-th row and the Vith column. When the processing of the storage instruction section 19 is completed, control is transferred to the pixel number counting section 21. In addition,
In the above explanation, unless j-=j, control passes through the storage instruction section 19 and is transferred to the pixel number counting section 21. The pixel number counting unit 21 measures the pixel number of the screen and calculates the pixel number.
When the column value reaches the pixel number at the right edge of the screen, the scanning line number in the i row is incremented by 1, and the value in the j column is adjusted to the value 1 at the left edge of the screen. Then, when the value in the i row and the j column reaches the pixel number corresponding to the lower right edge of the screen, control is transferred to the cycle counting section 22. If the pixel number corresponding to the lower right corner of the screen has not been reached, control is returned to the input unit 10 and the above-described process is repeated.

The period counting section 22 measures the period of the screen, and increases the count by 1 when control is transferred. Further, the value of the cycle number storage unit 23, which stores the cycle for extracting the background in advance in terms of the number of screens, is compared with the value measured by the cycle count unit 22. If the values match, the contents of the cycle count section 22 are reset to O and control is transferred to the background information extraction section 24, and if they do not match, control is returned to the input section 10.

In order to extract background information, the background information extraction section 24 extracts the maximum value among the scanning line feature values of each row of the feature value storage section 20 as background information of each scanning line, and outputs it to the background information storage section 25. Then, the value in row i and column j is reset to O and control is returned to input unit 10. The background information storage unit 25 can store background information in a one-dimensional array.

FIG. 2 is a block diagram showing a second embodiment according to the present invention. In the figure, the same parts as in FIG. 1 are given the same reference numerals. Reference numeral 16 denotes a region determining section 0. In this embodiment, the operations from the input section 10 to the feature value updating section 14 are the same as in the first embodiment. Now, the area determination unit 16 receives the scan a% characteristic value vi from the feature value update unit 14.
When it is received, the area storage unit 15 also reads the left and right values (jz-Jt) of the area of one line, and if j-=jt, control is transferred to the storage instruction unit 19. Furthermore, if this condition is not satisfied, control is transferred to the pixel number counting section 21. The storage instruction unit 19 refers to the content 1ho1 of the background information storage unit 25 that stores background information, and uses the preset values α and β to calculate +1lel−α (vI≦+)
If lol+β, control is transferred to the pixel number counting section 21. Also, if this condition is not satisfied, the feature value updating unit 1
The operations after 0 in which the value in the corresponding area of the feature value storage unit 20 storing the frequency of occurrence of the scanning line feature value determined in step 4 is incremented by 1 are the same as in the first embodiment. Therefore, in the second embodiment, the area determination unit 16 limits the measurement area, and the storage instruction unit compares it with the data in the background information storage unit 25 that has already been stored, thereby detecting the presence of objects other than the background. It is possible to recognize and extract the number of scanning line feature values that occur most frequently as background information in the image.

FIG. 3 is a block diagram showing a third embodiment according to the present invention. In the figure, the same parts as in FIGS. 1 and 2 are designated by the same reference numerals. 17 is a background validity determining section, and 18 is a flag storage section.

In this embodiment, the operation from the input section 10 to the area determination section 16 is the same as in the second embodiment. Now, when the background validity determining unit 17 receives control from the area determining unit 16, it refers to the content +fiol of the background information storage unit 25, uses preset values α and β, and +hoI
If -α≦Vi≦+h61+β, the value of the corresponding scanning line i row in the flag storage unit 18 that stores the scanning line characteristic value of each scanning line in the same image is written as the scanning line characteristic value vt. If this condition is not satisfied, a flag of %% is written and control is transferred to the storage instruction section 19. As shown in the array diagram of scanning line feature values shown in FIG.
When the scanning line feature value Vi of the scanning line i row is obtained, the scanning line i
It is determined based on the value in the flag storage unit 18 whether the feature values from −n −m to scanning line i are valid as a background.

That is, if the gutter flag is written in the flag storage unit 18, it is determined that there is something different from the background from scanning line i-n-m to scanning line i, and the scanning line feature value of scanning line i-n is determined. The feature value storage unit 2 does not recognize Vi-0 as background information.
Do not manipulate 0 values. On the other hand, if the masu-hakama flag is not written in the flag storage unit 18, the scanning line i −n −m
It is determined that the area from to scanning line i is the background, and scanning line i
Using the scanning line feature value V of -n and -a as background information, the value in the i-n row V and -0 column of the feature value storage unit 20 is incremented by 1. Note that m and n indicate the number of scanning lines. When the storage unit 1911 completes the above operation, the pixel number counting unit 2
Transfer control to 1. The subsequent operations are the same as in the first embodiment. Therefore, in the third embodiment, after setting the range between scanning lines by the storage instruction unit 19 and recognizing whether or not there is an object other than the background within this range, the most frequently occurring scanning line feature value The number of images can be extracted as background information within the image.

[Effects of the present invention]

As explained above, according to the background information extraction method of the present invention, the difference between neighboring pixel values in the scanning line direction in a two-dimensionally sampled image is taken, and υ image information is obtained based on the magnitude of the value. By converting "1" into rOJK binarization and extracting the number of "1"s that occur most often on the scanning line within a predetermined time as background information in the image, the background can be changed from moment to moment depending on the Even if the information changes, background information that satisfies the conditions can be extracted. Thereby, when detecting a moving object such as a conventional vehicle or aircraft, measurement accuracy can be improved.

In addition, background information can be automatically extracted, so
Processing speed can be increased.

Furthermore, even if there is an object other than the background in the same image information, it is possible to extract the background information by excluding only that object, which eliminates confusion between moving objects, etc. and the back information. It has remarkable effects such as

[Brief explanation of the drawing]

FIG. 1 is a block diagram showing a first embodiment according to the present invention,
FIG. 2 is a block diagram showing the second embodiment, and FIG. 3 is a block diagram showing the third embodiment.
FIG. 4 is a block diagram showing an embodiment of the present invention, and FIG. 4 is an array diagram of scanning line feature values. 10...Input section, 11...Shift register, 1
2...Difference calculation unit, 13...Threshold comparison unit,
14... Feature value update unit, 15... Area storage unit,
16... Area determination unit, 17... Background validity determination unit, 18... Flag storage unit, 19... Storage instruction unit, 20... Feature value storage unit, 21... ... Pixel number counting section, 22... Period counting section, 23... Q
- Period number storage section, 24... Background information extraction section.

Claims (3)

[Claims] (1) Calculate the difference between neighboring pixel values in the scanning line direction within a two-dimensionally sampled image, and binarize the image information into "1" and "0" depending on the magnitude of that value. In a method that processes the number of "1"s on a scanning line as a feature value, when processing a signal that periodically generates a fixed number of images, the number of "1"s on the scanning line that occur most often within a predetermined time is
'' as background information in an image. (2) Calculate the difference between neighboring pixel values in the scanning line direction within a two-dimensionally sampled image, and binarize the image information into "1" and "0" depending on the magnitude of that value. In a method that processes the number of "1"s on a scanning line as a feature value, when processing a signal that periodically generates a fixed number of images, a predetermined A background information extraction method characterized by extracting the number of "1"s that occur most frequently on a scanning line within a time as background information in an image. (3) Calculate the difference between adjacent pixel values in the scanning line direction within a two-dimensionally sampled image, and binarize the image information into "1" and "0" depending on the magnitude of that value. In a method that processes the number of "1"s on a scanning line as a feature value, when processing a signal that periodically generates a fixed number of images, there was an object other than the background between the set scanning lines. A background information extraction method characterized in that the number of "1"s that occur most frequently on a scanning line within a predetermined time period, excluding times, is extracted as background information in an image.