JPH04301571A - Apparatus for recognizing environment of moving vehicle - Google Patents

Abstract

PURPOSE:To reduce the amount of processing and to conduct extraction of a moving body and calculation of the amount of movement at a high speed. CONSTITUTION:A preceding vehicle or a white line on a road, which is taken as an object, is inputted as an image by a right image input element 1 and a left image input element 2 which are constructed of CCD cameras or the like installed so that center axes are parallel. Feature amount detecting elements 3 and 4 extract edges through filters from the images picked up by the right image input element 1 and the left image input element 2, respectively, and feature amounts extracted are sent to a differential device 5. In the differential device 5, comparison of the feature amounts obtained by the feature amount detecting elements 3 and 4 is conducted and, as the result, time-series data for the feature amounts are prepared. These time series data are stored in an image accumulating element 6 and, only prescribed parallel feature couples being observed finally in a feature amount tracing element 7, the amount of movement thereof is calculated.

Description

[Detailed description of the invention]

0001

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to an environment recognition device for a moving vehicle that recognizes moving objects by image processing.

0002

2. Description of the Related Art Conventional methods for extracting moving objects, particularly moving obstacles on roads such as people and vehicles, use stereoscopic parallax and movement vectors. In this method, the amount of movement is calculated using the results of viewing the object in stereo, or a movement vector is defined, and stereo viewing is used for the feature amount having the movement vector.

0003

[Problems to be Solved by the Invention] However, the above conventional method of determining the amount of movement using the results of stereo viewing, or defining a movement vector and using stereo viewing for the feature quantity having that movement vector, requires a large amount of calculation. The problem is that the amount of information becomes enormous, making it difficult to install it in actual vehicles from the viewpoint of processing speed.

0004

[Means for Solving the Problems] The present invention has been made for the purpose of solving the above-mentioned problems, and has the following configuration as a means for solving the above-mentioned problems. That is, a stereo image input means using left and right cameras arranged so that their central axes are parallel, a means for extracting edges of each input image from the left and right cameras, and a means for obtaining a difference image between the left and right edge extraction screens. , means for extracting a pair of parallel lines from the difference image, and means for extracting a moving object based on the movement of the pair of parallel lines.

[0005]

[Operation] With the above configuration, it is possible to extract a moving object and calculate the amount of movement at high speed.

[0006]

DESCRIPTION OF THE PREFERRED EMBODIMENTS Preferred embodiments of the present invention will now be described in detail with reference to the accompanying drawings. FIG. 1 is a block diagram showing the overall configuration of an environment recognition device for a moving vehicle (hereinafter referred to as the device) according to an embodiment of the present invention. In the figure, a right image input section 1 and a left image input section 2 are configured with, for example, a CCD camera or the like, and in this device, two cameras are used to input a target preceding vehicle and a white line on the road. These two cameras are installed so that their central axes are parallel to each other, minimizing changes in the left and right images on the imaging plane due to inclination.

The feature quantity detection units 3 and 4 perform Gaussian-Laplacian detection, for example, to perform feature quantity extraction such as edge extraction, which will be described later, from the images taken by the right image input unit 1 and left image input unit 2, respectively. It has a filter and sends the extracted feature amount to the difference device 5. The difference device 5 compares the feature amounts obtained by the feature amount detection units 3 and 4, and as a result, time series data for the feature amounts are created. Then, those time series data are stored in the image storage section 6. t to t+n in FIG. 1 indicate n pieces of time-series data stored in the image storage unit 6.

[0008] The image data stored and accumulated in the image storage unit 6 is finally focused on only predetermined feature quantities in the feature quantity tracking unit 7, and the amount of movement thereof is calculated. Next, a feature amount tracking method in the apparatus of this embodiment will be explained. FIG. 2 is a schematic flowchart showing the procedure of feature amount tracking in the apparatus of this embodiment. In the same figure, in step S1, a target image is input into the right image input section 1 and the left image input section 2, and in step S2, edge extraction is performed using the above-mentioned filters built into the feature quantity detection sections 3 and 4. Then, in step S3, the difference device 5 performs a process of comparing the left and right image feature quantities by calculating a difference between the feature quantities. In order to obtain the difference in feature amounts, here, we do not match the right image and the left image with a large difference in feature amount, or if there is no difference, but instead use parallel lines that satisfy those conditions. We focus only on features that can be approximated by (this is called the parallel hypothesis).

[0009] Normally, when an object moves, it means that the object moves in parallel when viewed from the observer's side, and attention is paid to the fact that even if the state of movement becomes complicated, this parallelism is not lost. Specifically, as shown in Fig. 3, the edges of the object obtained from the right image (indicated by the dashed line) and the edge of the object corresponding to the left image (indicated by the solid line) are parallel to each other. It is determined whether or not to perform matching based on whether the interval d between the parts is within a predetermined range. In other words, only those edges whose distances between parallel lines formed by the obtained edges are within a certain range are tracked as features. In other words,
The idea is to focus on objects whose parallax is within a certain range.For distant objects, when captured as images with two cameras, there is no parallax, or even if there is, the parallax is small, so it is ignored as a feature. That will happen.

In step S4 of FIG. 2, the above-mentioned spacing between the parallel parts is determined, and if the difference is within a predetermined value, that is, the spacing d between the parallel parts is within a certain range, it is converted into a parallel line. The image storage unit 6
Store in. However, if the interval between the parallel parts is not within a certain range, the parallel hypothesis does not hold, and the feature amount is deleted in step S5.

[0011] In step S7, it is determined whether the time-series data of the feature quantities created by tracking the feature quantities and stored in the image storage section 6 has reached a predetermined value. If so, return to step S1 again. However, when predetermined time-series data is obtained, the process proceeds to step S8, and the amount of movement of the feature amount is calculated based on the time-series data.

FIG. 4 is an example of time series data, and shows how pairs of parallel lines, which are feature quantities, move on the screen. By tracing this pair of parallel lines, the amount of movement of the target object can be calculated. As explained above, according to this embodiment, two cameras are installed in parallel, and a pair of parallel lines obtained by calculating the difference between the feature values of the left and right images extracted through a filter is tracked as a feature value. This has the effect of reducing the amount of processing and speeding up the extraction of moving objects and the calculation of the amount of movement.

[0013]

[Effects of the Invention] As explained above, according to the present invention,
This method has the effect of reducing the amount of processing by focusing only on feature quantities that can be approximated by parallel lines on the image, and speeding up the process of extracting a moving object and calculating its movement amount.

[Brief explanation of drawings]

FIG. 1 is a block diagram showing the overall configuration of an environment recognition device for a moving vehicle according to an embodiment of the present invention;

FIG. 2 is a schematic flowchart showing the procedure for feature quantity tracking in the device of the embodiment;

[Fig. 3] A diagram showing parallel line pairs corresponding to left and right images obtained by edge extraction,

FIG. 4 is a diagram showing an example of time-series data of feature amounts.

[Explanation of symbols]

1 Right image input section 2 Left image input section 3, 4 Feature detection unit 5 Difference device 6 Image storage section 7 Feature tracking unit

Claims (1)

[Claims] 1. A stereo image input means using left and right cameras arranged so that their central axes are parallel, means for extracting edges of each input image from the left and right cameras, and a difference image between the left and right edge extraction screens. 1. An environment recognition device for a moving vehicle, comprising: means for obtaining a pair of parallel lines from the difference image; and means for extracting a moving object based on the movement of the pair of parallel lines.