JPH1115982A - Method for detecting moving object - Google Patents

Abstract

PROBLEM TO BE SOLVED: To generate a background image that follows up the variation of environments without receiving effect of a detection object body and to detect a stable object by making an updation unit of a background from a pixel unit to an area unit. SOLUTION: An inter-frame difference processing part 11 detects the positions of variation pixels from two pieces that are an input image 2 photographed at (t-1) and an input image 2 photographed at (t). Next, an area split processing part 12 performs area split processing that splits into each object from a processing result of detected inter-frame difference processing and the image 2 photographed at (t). An image variation area detection processing part 13 performs image variation area detection processing according to the processing result. Next, the part 13 decides the updation area of a background image, and a synthesization processing part 14 performs synthetic processing of an input image and a past background image based on the decision result and generates a new background image. With this, it is possible to detect an object due to background difference even at an unstable photographic site.

Description

DETAILED DESCRIPTION OF THE INVENTION

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a moving object detection method, and more particularly to a moving object detection method using digital image processing.

[0002]

2. Description of the Related Art A conventional moving object detecting method of this kind will be described with reference to the drawings.

FIG. 5 is a flowchart showing an algorithm showing a conventional moving object detection method.

Conventionally, as a method for setting uniform integration parameters for a motion extraction processing area in an image, there is a method described in Japanese Patent Application Laid-Open No. Sho 62-114064. In this method, an input image f (n) and an old background image are g (n−
Assuming that 1), the background image g (n) to be updated is g (n) = A × f (n) + (1−A) × g (n−1) (1) It is. Here, A is a constant and is common to all pixels.

That is, the moving object is detected by the simple addition method disclosed in Japanese Patent Application Laid-Open No. Sho 62-114064.
Was detected by the method shown in the algorithm.

However, according to this method, depending on the moving speed of the object to be extracted, the reflectivity of the surface, and the threshold value,
There is a problem that a hole or a tail occurs in the region to be extracted and the shape of the extraction region is deformed.

As a method for solving this problem, there is the following method.

FIG. 6 is a flow diagram showing a background integration process in a conventional moving object detection method.

In FIG. 6, the background integration processing of this conventional moving object detection method is disclosed in Japanese Patent Application Laid-Open No. Hei 6-223188. When performing the difference processing 105 using the threshold value spatially set by the threshold control processing 106 with the background image that has been set, the threshold value is set to half of the difference value image, thereby achieving good motion extraction. Becomes possible. The background integration parameter is an estimated value of the time at which the object to be extracted at the moment obtained by the situation determination process 116 is stopped moving, and the threshold value of the binarization spatially set by the threshold value control process 106. And difference processing 10
5 is determined from the ratio of the difference values obtained.

In this conventional example, a difference value image between an input image and a background image is obtained, and a threshold value is set to half of the difference image at each pixel.
A threshold value for binarization and a background integration parameter are determined by using a difference value image between an input image and a background image and an estimated value of a time at which the extraction target object stops moving. That is, the accuracy of the extraction area of the extracted moving object when the background is updated using the background integration method is improved.

[0011]

The conventional method for detecting a moving object is disclosed in Japanese Unexamined Patent Publication No. Sho 62-1140 by a simple addition method.
In the case of Japanese Patent No. 64, the addition time (time axis) is shortened in order to quickly generate a background image. That is, when the value of the background integration parameter A in the equation (1) is increased, the detection target object is also affected by being added to the background image,
There is a problem that the detection target object appears in the background image as a shadow.

Based on this simple addition method, Japanese Patent Laid-Open No.
Japanese Patent Application Laid-Open No. 23188 discloses a method of estimating and calculating the value of the background integration parameter A based on the time at which the object stops moving, and determining the value based on the calculated time. There is a problem.

That is, in the background subtraction method under the condition that a moving object is detected in the conventional moving object detection method,
There is a problem that the background image cannot be generated promptly without being affected by the detection target object.

[0014]

A moving object detecting method according to the present invention is a moving object detecting method for detecting a moving object based on a difference between an input image and a background image. The input image and the second image taken at (t)
Based on an inter-frame difference process for detecting the position of a fluctuating pixel from two of the input images, a region division process for dividing each second object from the second input image captured at the time (t) is performed.
An update area of the background image is determined from the area division processing result by the image fluctuation area detection processing, and based on the determination result, the first input image and the past background image are combined,
An area determination synthesis unit that generates a new background image is provided.

The moving object detecting method of the present invention is a moving object detecting method for detecting a moving object based on a difference between an input image and a background image. And a region determining / combining unit for performing a combining process with (t-1)
The elements of the first input image at the time and the second input image at the time (t) are compared between frames, and the first pixels having different grayscale values are used to determine the first moving area by a predetermined first threshold value. An inter-frame difference processing means for judging, wherein the second input image at the time (t) is adjacent to the previously determined moving area and has a gray level value of the moving area within a predetermined second threshold. Area division for detecting a second moving area by inferring a second element having a gray value as being caused by the same moving object
A fluctuating area detecting means, and masking the input image at time t from the second moving area to update a background image at time t-1;
image synthesizing means for generating a background image at time t.

[0016]

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

FIG. 1 is a flowchart showing an algorithm according to an embodiment of the present invention, FIG. 2 is a diagram showing an example of a time change of an input image and a background image for explaining the algorithm of the embodiment, and FIG. FIG. 4 is a view showing the concept of the area division processing and the image fluctuation area detection processing based on FIG. 2 for explaining the algorithm of the present embodiment. FIG. 4 is a view for explaining the algorithm of the present embodiment. FIG. 4 is a diagram showing a concept of a combining process based on FIG.

The algorithm of the moving object detection method according to the present embodiment considers a problem that occurs because the average image of the simple addition method is regarded as a background image, and detects a moving object. This is an algorithm devised from the condition of generating a background image for updating a background image with respect to an input image by taking only a still (stop) object region as a background.

That is, an algorithm is not considered as a background image, but is focused on a moving object of moving image data, thereby determining an image obtained by removing the moving object from an input image as a background image.

In FIG. 1, the algorithm of the moving object detection method of the present embodiment is as follows. The input image 2 shot at (t-1) and the input image 2 shot at (t)
An inter-frame difference processing unit 11 for detecting the position of a fluctuating pixel from a sheet, an area division processing unit 12 for performing an area division process for dividing each object from the processing result and the input image 2 captured at (t) An image change area detection processing unit 13 that performs image change area detection processing based on the above processing results, and an update area of the background image is determined by the image change area detection processing unit 13. Based on this determination result, the input image and the An area determination synthesizing unit 1 includes a synthesizing unit 14 that performs a synthesizing process with a past background image and generates a background image. Other processes (the background image 3, the background difference unit 4, and the object detection unit 5) are the same as those shown in FIG.

First, as shown in FIG. 2, the inter-frame difference processing section 11 outputs an image I ₋₁ (x, y) of I (t−1) of the input image 2 at time t−1 and an input image at time t. Pixel I of I (t)
(X, y) is compared between the frames, and the images having different grayscale values are determined as the moving area M (x, y) by the threshold value T1. That is, when this is expressed as an equation, the equation shown in (2) is obtained.

[0022]

(Equation 1)

Next, as shown in FIG. 3, the area division processing section 12 and the image fluctuation area detection processing section 13 determine the moving area M (x1, y1) determined earlier in the input image 2 at time t.
= 1 and the gray value I of the moving area
Pixel I having (x1, y1) and a gray value within threshold value T2
It is assumed that (x2, y2) is caused by the same moving object, and is determined as the moving area M (x2, y2). This determination is repeated until the gray value of an adjacent pixel becomes a value other than the threshold value T2, and a moving area of the object is obtained. That is, when this is expressed by an equation, the following equation (3) is obtained.

[0024]

(Equation 2)

As shown in FIG. 4, the input image I (x, y) at the time t is masked by the moving area M (x, y) thus obtained, as shown in FIG. ) Time background image B _-1 (x, y) is updated, and the background image B at time t is updated.
Generate (x, y).

Here, (t) at the start of background generation
The background image B _-1 (x, y) at _-1 ) is the input image I _-1 (x, y) at (t-1).

Background image B (x, y) generated at time t
Becomes the equation shown in (4).

[0028]

(Equation 3)

As described above, in the present embodiment, since the inter-frame difference is used for the determination of the moving area, it can be determined as the moving area if the object moves even a little.
Therefore, the present invention can be applied to various photographing environments (such as an image photographed from the side of the road body in the vehicle traveling direction or an image photographed from above the road). In addition, since the background image is updated for each frame, the response to a change in the imaging environment such as illumination is prompt.

[0030]

As described above, according to the present invention, in a moving object detection method for detecting a moving object based on a difference between an input image and a background image, a first input image photographed at (t-1); (T) 2 of the second input image captured at the time
Based on the inter-frame difference processing for detecting the position of the fluctuating pixel from the image, a region division process for dividing each object from the second input image photographed at the time (t) is performed, and an image variation is performed based on the region division processing result. By determining the update area of the background image by the area detection processing, performing the synthesis processing of the first input image and the past background image based on the determination result, and providing an area determination synthesis unit that generates a new background image, Change the image generation method from the simple addition method to the area judgment synthesis method,
That is, since the update unit of the background image is changed from the pixel unit to the area unit, each pixel of the input image can have a meaning such as whether or not it is a background area. Since only the area can be updated as a background image, it is possible to generate a background image that is not affected by the object to be detected and that follows environmental changes, and thus is considered to be impossible in the past, such as outdoors. Even in an unstable shooting environment, it is possible to detect an object based on the background difference, so that there is an effect that stable object detection can be realized.

[Brief description of the drawings]

FIG. 1 is a flowchart showing an algorithm according to an embodiment of the present invention.

FIG. 2 is a diagram illustrating an example of a temporal change of an input image and a background image for explaining an algorithm of the present embodiment.

FIG. 3 is a diagram illustrating the concept of a region division process and an image variation region detection process based on FIG. 2 for explaining the algorithm of the present embodiment.

FIG. 4 is a diagram illustrating the concept of a combining process based on FIGS. 2 and 3 for explaining the algorithm of the present embodiment.

FIG. 5 is a flowchart showing an algorithm showing a conventional moving object detection method.

FIG. 6 is a flow diagram showing background integration processing in a conventional moving object detection method.

[Explanation of symbols]

 REFERENCE SIGNS LIST 1 area determination synthesis unit 2 input image 3 background image 4 background subtraction unit 5 object detection unit 11 inter-frame difference processing unit 12 region division processing unit 13 image fluctuation region detection processing unit 14 synthesis processing unit

Claims (2)

[Claims] 1. A moving object detection method for detecting a moving object based on a difference between an input image and a background image, comprising: a first input image photographed at (t-1), and a second input image photographed at (t). Based on the inter-frame difference processing for detecting the position of a fluctuating pixel from two of the input images of (a) and (b), a region division process of dividing each object from the second input image photographed at the time (t) is performed. An area to be updated of a background image is determined from the result of the division processing by an image change area detection processing, and based on the determination result, the first input image is combined with a past background image to generate a new background image. A moving object detection method, comprising: 2. A moving object detecting method for detecting a moving object based on a difference between an input image and a background image, wherein a background image in which the moving object is removed from the input image is determined, and an area for performing a combining process with a past background image is determined. A determination synthesizing unit, wherein the area determination synthesizing unit compares the elements of the first input image at the time (t-1) with the second input image at the time (t) between frames, and compares the elements of the first input image at the time (t). An inter-frame difference processing means for determining a first moving area by a pixel based on a predetermined first threshold; and adjacent to the moving area previously determined in the second input image at the time (t); Area division / variation area for detecting a second moving area by inferring a second element having a shading value of the moving area and a shading value within a second threshold value determined in advance as being caused by the same moving object Detection means, the second movement area The mask the second input image when (t) (t-1) updates the background image when, (t)
A moving object detecting method, comprising: an image synthesizing processing unit for generating a background image at the time.