JPH07249127A - Traveling object detection device - Google Patents

Abstract

PURPOSE:To securely detect an object which travels so that it crosses a traveling direction under a situation that an image input-side travels by detecting the traveling object from the polarity of a speed vector component in the horizontal direction of optical flow with an infinite far point direction as a reference. CONSTITUTION:An image input part 1 continuously image-picks up an area in the traveling direction of a vehicle by a video camera and obtains an animation image. An optical flow operation part 2 samples the respective images in an image element unit, obtains data on density and calculates optical flow showing the distribution state of a speed vector on the image, which follows the spatial density gradent and the timewise density gradent of the moving image. A smoothing processing part 3 spatially smooths the vector of optical flow. An infinite far point direction extraction part 4 obtains the minimum value of the size of the speed vector component in the horizontal direction from optical flow and detects the infinite far point direction. A detection part 5 detects the traveling object from the porality of the speed vector component in the horizontal direction of optical flow with the infinite far point direction as the reference.

Description

Detailed Description of the Invention

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a moving object detecting device for detecting a moving object existing in a moving image.

[0002]

2. Description of the Related Art Conventionally, as a method for detecting a moving object in a moving image, a difference in density between frames of a moving image is calculated, and the difference between the frames is detected.
A density change area is extracted using a preset density threshold value, and the area corresponding to the extracted area is detected as a moving object.

However, according to such a method, in order to detect a moving object in an image based on a difference in density between frames, the side capturing the moving object needs to be stationary, and the image capturing side moves. In the state where the background is moving, the density of the background also changes and the background and the moving object cannot be separated.

Further, conventionally, from a moving image obtained by imaging the outside world with a camera mounted on a moving body, an optical showing a distribution state of velocity vectors using a spatial concentration gradient and a temporal concentration gradient is used. A method of calculating a flow and recognizing a three-dimensional movement of an image from the obtained optical flow has been proposed (USP).
attend No. 4,980,762).

However, this method detects a parallel movement or a rotation movement of the entire image due to the movement of the moving body which is the input side of the moving image, and cannot detect a moving object existing in the image. Has become.

Further, conventionally, as a method for detecting a moving object in a moving image, a frame image is divided into small areas,
The velocity vector is calculated for each small area to create a velocity vector distribution map, the components of the velocity vector distribution pattern are removed from the distribution map, and the remaining velocity vector components in the frame image are removed. Is extracted as a moving object area (Japanese Patent Laid-Open No. 4-3).
28689).

However, according to such a method,
It is necessary to register the velocity vector distribution pattern in advance. Further, even if the velocity vector is calculated for each small area in this way, it is still difficult to detect only the moving body with high accuracy due to the inclusion of noise. For example, when such a device is mounted on a vehicle, even a device that is not moving may be detected as a moving body due to the influence of noise, resulting in poor reliability.

[0008]

The problem to be solved is that when a moving object is detected in a moving image, it moves in the direction opposite to the moving direction when the image input side is moving. That is, the method of detecting the moving object or the object moving transversely with respect to the moving direction is not surely established.

[0009]

According to the present invention, a moving image is input so that an object moving transversely to the moving direction can be detected under the condition that the image input side is moving. Means, an optical flow calculating means for calculating an optical flow indicating the distribution state of the velocity vector on the image from the input moving image, and a minimum of the magnitude of the velocity vector component in the horizontal direction from the calculated optical flow. An infinity point direction extracting means for obtaining a value and finding the infinity point direction, and a moving object is detected from the polarity of the velocity vector component in the horizontal direction of the optical flow with reference to the extracted infinity point direction. The moving object detecting means is used.

[0010]

DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS As shown in FIG. 1, a moving object detecting apparatus according to the present invention is mounted on a vehicle, and an image input section 1 for continuously obtaining a moving image by continuously picking up an area in the traveling direction by a video camera. Each continuously captured image is sampled on a pixel-by-pixel basis to obtain density-related data, and the distribution state of velocity vectors on the image according to the spatial density gradient and temporal density gradient of the moving image. , An optical flow calculation unit 2 for calculating an optical flow, a smoothing processing unit 3 for spatially smoothing the vector field of the calculated optical flow, and a horizontal velocity vector from the smoothed optical flow. Based on the infinity point direction extraction unit 4 that finds the minimum value of the component size and finds the infinity point direction, and the calculated infinity point direction,
The moving object detection unit 5 detects a moving object from the polarity of the velocity vector component in the horizontal direction of the optical flow.

The calculation of the optical flow in the optical flow calculation unit 2 is performed, for example, in the document “Art.
"optical Intelligence" Vol. 1
7. No. 1-3 (1981). pp. 185-20
3. Berthold K. P. Horn and B
rian G. Schunkk "Determinin
The conventional method described in “g Optical Flow” is used as it is.

FIG. 3 shows an example of the optical flow. As shown in FIG. 2, an input image obtained by imaging a region in the traveling direction of the vehicle is shown to cross the traveling direction of the vehicle. It is a field of a local two-dimensional velocity vector in a moving image when there is an object O moving in the direction. In the figure, P indicates an infinite point corresponding to the traveling direction of the vehicle.

The smoothing processing section 3 smoothes the vector field of the optical flow with a spatial filter such as a Gaussian filter to reduce the noise component contained in the vector field.

The Gaussian filter f (r) is represented by the following equation (1) according to the distance r from the center and the standard deviation σ.

[0015]

[Equation 1]

The infinity point direction extraction unit 4 calculates the velocity in the horizontal (X) direction in an arbitrary range in the vertical (Y) direction by the following equation (2) in the vector field V (X, Y) of the optical flow. Sum of squares of x component of vector E
(X) is calculated, x = x _C that minimizes E (x) is obtained, and the infinity point direction D is determined as shown in FIG.

[0017]

[Equation 2]

The moving object detecting unit 5 detects the background velocity vector V _B given by the following equation (3), which is generated by the self-motion of the mounted vehicle traveling, and the traveling direction of the vehicle given by the following equation (4). the polarity of the x component of the velocity vector V _O of the object O to be moved across for by using the difference, to detect the moving object O.

[0019]

[Equation 3]

[0020]

[Equation 4]

At this time, for example, if the velocity vector V _B while the vehicle is traveling is a condition of V _B <0 x <x _C or V _B > 0 x> x _C , a certain threshold value ν _O (ν _O > 0) relative to, satisfy the condition of x _{<x C} and V _O> 0 or x> _{x C} and V _O <0 seek (x, y), as shown in FIG. 5,
The position of the moving object O is detected.

As described above, according to the present invention, the object O moving transversely to the traveling direction of the vehicle is selected from the input images shown in FIG. 2 obtained by imaging the region in the traveling direction of the vehicle. Will be able to be detected.

In this embodiment, it is not possible to detect an object moving parallel to the traveling direction of the vehicle, but the information for the driver of the vehicle is to detect a moving body approaching the vehicle. It ’s important that this information
For example, it is possible to quickly avoid a collision.

[0024]

As described above, in the moving object detecting device according to the present invention, the minimum value of the magnitude of the velocity vector component in the horizontal direction is obtained from the optical flow obtained from the input image, and the infinity point direction is determined. Then, the moving object is detected based on the polarity of the velocity vector component in the horizontal direction of the optical flow with reference to the direction of the infinity point thus extracted, and under the condition that the image input side is moving, It has an advantage that an object that moves so as to cross the moving direction can be reliably detected.

[Brief description of drawings]

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

FIG. 2 is a diagram showing an example of an input image.

FIG. 3 is a diagram showing an optical flow corresponding to an input image.

FIG. 4 is a diagram showing an infinity point direction in an input image.

FIG. 5 is a diagram showing a position of a moving object detected from an input image.

[Explanation of symbols]

 1 Image Input Section 2 Optical Flow Calculation Section 3 Smoothing Processing Section 4 Infinity Point Direction Extraction Section 5 Moving Object Detection Section O Moving Object P Infinity Point D Infinity Point Direction

Claims (1)

[Claims] 1. A means for inputting a moving image, an optical flow calculating means for calculating an optical flow indicating a distribution state of velocity vectors on the image from the inputted moving image, and a horizontal direction from the calculated optical flow. Direction of the optical flow in the horizontal direction based on the infinity point direction extraction means that finds the minimum value of the velocity vector component in the direction and finds the infinity point direction. A moving object detecting device comprising a moving object detecting means for detecting a moving object from the polarities of vector components.