JPH0546732A - Detector for number of persons - Google Patents

Abstract

PURPOSE:To highly accurately detect the number of persons without being influenced by a setting place or a setting environment of a television (TV) camera by reducing the influence of a low contrast between an objective person and a floor to be a background or a change in brightness due to an illuminating state or the like in the case of processing a video signal obtained from the TV camera and detecting the number of persons. CONSTITUTION:A device for picking up the image of a person to be detected by the TV camera is provided with an arithmetic means 6 for automatically updating a background image including no person, arithmetic means 9, 10 for detecting the number of persons to be detected, a means 7 for changing a threshold level for binarizing an image signal at the time of detecting the number of persons based on the average value of brightness density gradation in an unmanned visual field photographed by the TV camera, a means 8 for compensating the shape of a person image based on the fixing height of the TV camera up to a ceiling and the focal distance length of a camera lens, and a means 12 for compensating the number of persons to be detected.

Description

Detailed Description of the Invention

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a device for detecting the number of people and the number of passers-by in a specific place such as in a department store, elevator hall, underground mall, etc. by processing a video signal of a television camera. is there.

[0002]

2. Description of the Related Art A method of converting a video signal of a television camera into a digital signal and arithmetically processing an image to be detected from the obtained digital image signal is widely used. For example, FIG. 5 is a block diagram of a conventional people detection device. Figure 5
In FIG. 1, 1 is an image input unit configured by a television camera or the like for capturing an image of a detection target person and outputting an image signal, and 2 is an image signal of the image input unit 1 for removing a noise component and performing binarization or the like. A preprocessing unit for processing and outputting a binarized image signal, 3
Is an object extraction unit for inputting the image signal of the preprocessing unit 2 to extract a detection target person image, and 4 is a target person image of the target extraction unit 3 and a reference area image which is input in advance and stored therein. An area comparing section 5 for comparison is a counting section for counting images larger than the reference area image.

Next, the operation will be described. Image input section 1
When an image of a certain place such as a department store or an elevator hall is captured and an image signal is output, the preprocessing unit 2 takes in the image signal, removes noise components by masking, and then performs binarization processing. And outputs a binarized image signal. The target extraction unit 3 extracts a detection target person image from the binarized image signal. A reference area image input in advance is stored in the area comparison unit 4, and the reference area image and the detection target person image of the target extraction unit 3 are compared to obtain a person image of a portion larger than the reference area image. The counting unit 5 counts these images and outputs a person number signal.

By thus digitally processing the image signal of the image input section, the number of persons to be detected can be detected.

[0005]

In the conventional number-of-people detecting apparatus as described above, it is necessary to confirm that there is no person in the background image of the reference area, and manually input the background image within the field of view of the camera. Every time the installation environment changed, such as when you put something, you had to input a new standard background image each time. Since area determination is performed based on a simpler binary image signal, when the contrast between the person and the background is low,
For example, when the hair color and the background floor are similar colors, the image area (number of pixels) of the detection target person may be small, or the brightness of the installation location of the image input unit may change.
It was difficult to set the threshold level when performing the binarization process. Also, when an image of a person to be detected is taken with a TV camera installed on the ceiling, the area image of the person to be detected becomes smaller in the central part of the TV camera and larger in the peripheral part due to the mounting height of the TV camera and the focal length of the camera lens. The detection accuracy has decreased. After all, the detection accuracy of the conventional people detection device is greatly affected by the installation location and installation environment of the TV camera.

The present invention has been made in order to solve the above-mentioned problems, and can automatically update the background image in a state where there is no human being, and can cope with the change in brightness due to the color of the floor or the lighting condition. An object of the present invention is to obtain a person number detecting device capable of detecting the number of persons with high accuracy without being affected by the installation place or the installation environment by reducing the influence.

[0007]

[Means for Solving the Problems] A person number detecting device for solving the above-mentioned problems includes an arithmetic means for automatically updating a background image in the absence of a person, and an arithmetic means for detecting the number of target persons. A means for changing the threshold level for binarizing the image signal when detecting the number of people based on the average value of the density gradation of the brightness in the unmanned field of view imaged by the TV camera, and the installation to the ceiling of the TV camera. A means for correcting the shape of the human image by the height and the length of the focal length of the camera lens,
And means for correcting the number of persons to be detected.

[0008]

[Operation] Since the number-of-people detection device according to the present invention automatically updates the background image that is the reference when detecting the number of people, even if the background changes, such as when an object is placed on the floor. There is no need to change the setting artificially. Also, even if there is a change in brightness due to the lighting condition, etc., the image signal at the time of detecting the number of people is binarized based on the average value of the density gradation of the background brightness in the absence of people. The threshold level is changed, and the shape of the portrait image is corrected by adjusting the height of the TV camera and the focal length of the camera lens, so it is affected by the installation location and environment of the TV camera. The number of persons to be detected can be detected with high accuracy.

[0009]

DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS An embodiment of the present invention will be described below with reference to FIGS. FIG. 1 is a block diagram showing the configuration of the number-of-people detecting device of the present invention.

Reference numeral 1 denotes an image input unit composed of the same TV camera as shown in FIG. 5, and reference numeral 6 denotes a background image calculation unit for calculating a background image in an unmanned state which serves as a reference when detecting the number of people from the image signal of the image input unit 1. , When there is no person, the background image is updated periodically. Reference numeral 7 denotes a density gradation calculation unit, which calculates the average value of the density gradation of the background image signal in an unmanned state, and changes the binarization level of the image signal when detecting the number of people based on the magnitude of this value to change the lighting state, etc. The influence of the resulting brightness is reduced. For example, when the illuminance is high and bright, a person's shadow is likely to appear, so the binarization level is increased. Reference numeral 8 is a shape correction unit. In the peripheral portion of the detected image, the human image appears longer than in the central portion. For this reason, the image of the peripheral portion is compressed and the shape of the image is corrected so that the size of the image for one person becomes substantially equal. Here, if the value of the shape correction level is reduced, the degree of shape correction is increased and the peripheral portion is further reduced. On the contrary, if this level value is increased, the peripheral portion is not reduced so much. The shape correction level is changed according to the mounting height of the TV camera on the ceiling and the focal length of the camera lens used.

Reference numerals 9 and 10 are the number-of-people detector. The number-of-people detection unit A of 9 captures only one screen of the image for which the number of people is desired to be detected, and performs a difference from the image (background image) of the state in which there is no person stored in the image storage unit of 11 (this is called a fixed difference Make a picture of only people. This fixed difference image
The binarization level determined based on the average value of the grayscale levels obtained by the grayscale level calculation unit 7 is binarized for each pixel to form a binary image. In order to detect the number of people from this binary image, the number of pixels of the image is first reduced in order to speed up the calculation process, and the pattern of the person of a predetermined size is overlapped on the reduced image and overlapped. A kind of pattern matching processing is performed in which the image area of a portion is calculated, and this area is used as the correlation value of the pixel at the center position of the overlapped patterns, and this is obtained over the entire image. Then, since the correlation value peak is the position where the pattern of the person best fits the image area, there is a person at that position. Therefore,
Each detected correlation value peak is counted as one person.

Next, the ten-person detection section B of 10 captures two images continuously at a predetermined time interval (for example, 0.1 seconds), subtracts these images (this is called a movement difference), and this movement difference After that, as with the processing of the number of people detection unit of 9,
After binarization and reduction of the number of pixels, the peak of the correlation value is detected by the pattern matching method and the number is counted. The feature of the movement difference process in the number-of-people detection unit B is that
Even in an environment where the contrast between the person to be detected and the floor as the background is low, the influence is small and highly accurate detection is possible. However, since this moving difference image is obtained in principle by the movement of a person, a completely stationary person cannot be detected. Reference numeral 12 is a number-of-people correction section, which compares the calculation results of the number-of-people detection section A of 9 and the number-of-people detection section B of 10, and selects the larger of these numerical values and outputs it as the number of people. In addition, for example, in an installation environment in which people move, the calculation result of the number-of-people detector B is selected.

By the way, the background image calculation unit 6 needs to correctly capture the background image in the absence of any person. The background image update procedure at this time will be described with reference to the flowchart of FIG. .. First, two images are continuously captured (procedure a). The commercial frequency is 50Hz for this acquisition time interval.
In consideration of the fact that it is 60 Hz, and in order to eliminate the illuminance change of the illumination due to the difference of the power supply frequency,
It is necessary to capture at intervals of (integers above) (especially for fluorescent lighting). Therefore, for example, capture is performed at intervals of 0.1 seconds. Next, the absolute value of the difference between the two captured images, that is, the movement difference is obtained (procedure b), and the difference image is binarized by a threshold value capable of removing noise (procedure c), and the temporal difference between the images is determined. Cut out only the changed part, in this case the part where the person has moved. Then, in order to shorten the processing time, a plurality of pixels are combined into one by reducing the number of pixels (procedure d),
After that, the correlation value is calculated (procedure e). First, an appropriate pattern that is close to the size of a person is prepared, and this pattern is scanned on a pixel-by-pixel basis to obtain the correlation value between the pattern and the image. That is, each value in the pattern (0 or 1)
And the pixel value (0 or 1) of the image at the position overlapping with that, and adding the value over the entire pattern is taken as the correlation value of the pixel at the center of the pattern.
Then, since the correlation value becomes high at the position where the pattern and the image shape of the person are well matched, the number of peaks of this correlation value is counted. When a plurality of peaks are detected at a short distance, they are counted as one. Since the number of peaks corresponds to the number of people, it is determined in step f whether or not there are people, and if there are 0 people, this is repeated and the number of detected people is 0 in m consecutive times. (Procedure g), assuming that there is no person, one of the original two images is saved as a background image (procedure h). Here, the procedure g is performed as follows.
It is difficult to detect a person's image by the movement difference when the lighting condition is bad or when there is almost no person's movement. Therefore, by increasing the number of detections, the image of a person is not saved as a background image. This is because.

In the background image updating procedure of FIG. 2, it is necessary to store, as an initial value, an image of a person without a person as the initial background image with the television camera attached.

Next, the calculation method of the shape correction unit 8 in FIG. 1 will be described with reference to FIG. As the shape correction, the radius r is converted around the position directly below the camera to obtain the radius r '. Consider the conversion formula as follows.

[0016]

[Equation 1]

Here, k is a correction coefficient. In considering the correction coefficient k, the length in the peripheral portion of the detected image is set to be equal to the length in the central portion. That is,
In FIG. 3, H is the ceiling mounting height (m) of the camera, H ₀
Is the height from the floor to the head of the person (m), (for example 1.5m),
F is the focal length (mm) of the camera lens, T is the lateral dimension (mm) of the camera element (for example, 8.8 mm for a 2/3 inch camera), M ₀ and M ₁ are the center of the camera field of view and A person standing at the edge of the field of view of the camera.

As is apparent from FIG. 3, the effective visual field L _{1 of the} camera is

[0019]

[Equation 2]

Here, L may be represented by the number of pixels even if the unit is (m). Now, assuming that the size m of the person's head is one side of the person's pattern, this value depends on the size of L. Where L
_{If 1} = A≡L ₁₀ , then m = a≡m ₀ (where A, a
Is a known number), and at any L ₁ , m is as follows.

[0021]

[Equation 3]

Therefore, r ₀ can be expressed as follows.

[0023]

[Equation 4]

In addition,

[0025]

[Equation 5]

[0026]

[Equation 6]

Therefore, r ₀ ′, r ₁ ′ and r ₂ after conversion are obtained.
′ Can be described as follows based on the above equation 1.

[0028]

[Equation 7]

[0029]

[Equation 8]

[0030]

[Equation 9]

Further, it is necessary to satisfy the following equation after the shape correction.

[0032]

[Equation 10]

Therefore, k may be obtained from the equations 7-10. That is, the correction coefficient k is the ceiling mounting height H of the camera,
It can be determined by the length of the focal length of the camera lens used.

Next, the number detection units A and 1 shown in FIG.
The processing procedure of the number-of-people detection unit B, which is 0, is shown in FIG.
A description will be given based on the flowcharts of (b) and (b).

In FIG. 4 (a), first, an image for which the number of people is desired to be detected is captured in one screen (procedure i), and the difference between this image and the background image in the state where no person is stored is determined (procedure j). The feature of this fixed difference is that, unlike the case of the moving difference, an image of a person can be obtained if there is a contrast with the background even when there is no movement of the person. Next, the difference image is binarized using a constant density level (this is called a binarization level) as a threshold value (procedure 1). The binarization level at this time is equal to or higher than a level at which noise emitted from a television camera or the like can be cut, and is determined based on the average value of the density gradation obtained by the density gradation calculation unit 7 in FIG. In step p, the number of pixels of the image is reduced in order to speed up the calculation of the number of people, and a pattern of a person of a predetermined size is superimposed on the reduced image to obtain a correlation value (step q). At this time, if the person's pattern is about the same size as the person's image and is as small as possible (for example, a pattern of 3 × 3 pixels), the operation speed can be increased and the accuracy can be improved. Next, the peak of the correlation value is detected, and the number of peaks is counted to be the number of people (procedure s). At this time, the peaks of the correlation values are only those which are separated from each other by a predetermined distance or more in consideration of the size of the person, and erroneous counting is avoided.

FIG. 4 (b) is a flowchart showing the processing procedure of the number-of-persons detecting section B shown in FIG. 1, in which two images are first consecutively captured (procedure t), and two consecutive images are captured. Determine the absolute value of the difference between the two images (procedure u). Next, this difference image is binarized with a threshold that can remove background noise (procedure v), and only the temporally changing portion of the image, in this case, the portion where the person has moved is cut out. After that, the same processing as steps p, q, and s in FIG. 4A is performed, that is, pixel number reduction (step w), correlation value calculation (step x), and correlation value peak counting (step y).

[0037]

As described above, according to the present invention, even if the contrast between the person to be detected and the floor as the background is low, or even if there is a change in the brightness due to the lighting condition or the like, the influence is small, and the television camera This has the effect that the number of people can be detected with high accuracy and in a short time without being affected by the installation location or installation environment.

[Brief description of drawings]

FIG. 1 is a block diagram showing a configuration of a people detection device according to an embodiment of the present invention.

FIG. 2 is a flowchart showing a background image update process.

FIG. 3 is an explanatory diagram of shape correction.

FIG. 4 is a flowchart for explaining the processing operation for detecting the number of people.

FIG. 5 is a block diagram showing a configuration of a conventional people detection device.

[Explanation of symbols]

 DESCRIPTION OF SYMBOLS 1 Image input unit 2 Pre-processing unit 3 Target extraction unit 4 Area comparison unit 5 Counting unit 6 Background image calculation unit 7 Density gradation calculation unit 8 Shape correction unit 9 Number of persons detection unit A 10 Number of persons detection unit B 11 Image storage unit 12 Number of persons Correction unit

Claims (6)

[Claims] 1. A person number detecting apparatus for detecting a number of persons by converting a video signal of a television camera into a digital signal, calculating a person image with a binarization level of the obtained digital image signal as a threshold value, A calculation means for updating the background image in the absence of people and a calculation means for detecting the number of people are provided, and the average value of the density gradation of the brightness in the unattended field of view photographed by the television camera is used to detect the number of people. The means for changing the binarization level of the image signal, the means for correcting the shape of the person image by the mounting height of the television camera and the length of the focal length of the lens of the television camera, and the number of persons to be detected. A person number detecting device, comprising: a correcting means. 2. The number-of-people detection device according to claim 1, wherein the calculation means for updating the background image in the absence of a person is a moving difference for obtaining a difference between two consecutive images. 3. An apparatus for detecting the number of people according to claim 1, wherein the arithmetic means for detecting the number of people is a fixed difference that takes a difference between an image to be detected and an image in which no person is stored in advance. .. 4. An apparatus for detecting the number of people according to claim 1, wherein the arithmetic means for detecting the number of people is a moving difference that takes a difference between two consecutive images. 5. The number-of-people detector according to claim 1, wherein the means for correcting the number of people compares the calculation result by the fixed difference with the calculation result by the movement difference and selects the larger calculation result. .. 6. The means for correcting the number of persons can select either the calculation means based on the fixed difference or the calculation means based on the movement difference depending on the installation location or installation environment of the television camera. The number-of-people detection device according to claim 1.