KR102371592B1 - Apparatus and method for estimating inter-vehicle distance - Google Patents

Abstract

The present invention relates to an apparatus and method for estimating an inter-vehicle distance, acquiring an image through a camera, detecting a vehicle and a lane from the image, and determining the position coordinates of the vehicle detected in the image and the lane angle of the detected lane in the image The lane width ratio is detected and the similarity is determined by comparing the fixed lane angle and the fixed lane width ratio, and the distance from the own vehicle to the detected vehicle is estimated by correcting the position coordinates of the detected vehicle according to the similarity.

Description

Apparatus and method for estimating inter-vehicle distance

The present invention relates to an apparatus and method for estimating an inter-vehicle distance for estimating a distance from a host vehicle to a vehicle in front through the use of lane information to correct shake in the vertical direction of an image due to road surface geometry or vehicle vibration.

In general, an autonomous driving system mainly uses a sensor capable of recognizing an object within a certain range, such as a radar and a laser. At this time, the value from the sensor itself is a distance value, and using this, the distance and speed of the object are calculated to determine the behavior of the vehicle. However, there is a problem in that the application of various sensors increases the manufacturing cost.

Accordingly, an algorithm for recognizing an object using only a camera in an autonomous driving system and measuring the distance to the recognized object is being developed. In the existing image-based distance measurement technology, the distance is estimated using the geometry information of the camera and the road surface. At this time, on the premise that the road surface is parallel to the vehicle, the distance to the recognized object is obtained using the relational expression between the road surface and the camera.

However, the actual road surface is not an ideal plane due to cracks, vertical curves, or bumps, and the vehicle itself vibrates. Therefore, since the premise that the existing road surface and the vehicle body are parallel does not hold, it is impossible to accurately estimate (measure) the distance.

KR 101463513 B1

The present invention acquires a front image of the own vehicle through a camera, detects lane information from the acquired image, and calculates a change in movement in the vertical direction in an image due to road surface geometry and vehicle vibration, etc. based on the detected lane information, and An object of the present invention is to provide an inter-vehicle distance estimation apparatus and method for estimating the distance from the host vehicle to the vehicle in front.

In order to solve the above problems, a method for estimating an inter-vehicle distance according to an embodiment of the present invention includes: acquiring an image through a camera; detecting a vehicle and a lane from the image; and detecting a lane angle and a lane width ratio of the detected lane in the image, comparing the lane angle and lane width ratio with a fixed lane angle and a fixed lane width ratio to determine a similarity, and the detected vehicle according to the similarity and estimating a distance from the host vehicle to the detected vehicle by correcting the position coordinates.

The position coordinate of the vehicle is characterized in that it is a y-coordinate of a point in which the detected wheel of the vehicle contacts the road surface in the image.

The lane width ratio may be a ratio between an actual lane width and the lane width.

The determining of the similarity may include calculating the similarity between the lane angle and the lane width ratio and the fixed lane angle and the fixed lane width ratio, and determining whether the calculated similarity is greater than or equal to a threshold value.

The estimating of the distance may include estimating the distance using the detected vehicle location coordinates if the similarity is greater than or equal to a threshold.

The estimating of the distance may include: if the similarity is less than a threshold, extracting a homography from the image; converting the detected position coordinates of the vehicle into coordinates in a plane view using the homography; converting into coordinates in a reference camera geometry by applying a fixed coordinate transformation determinant to coordinates in the plane view; and estimating the distance using coordinates in the reference camera geometry.

On the other hand, the apparatus for estimating the inter-vehicle distance according to an embodiment of the present invention detects a vehicle and a lane from a camera for acquiring an image, a memory storing a fixed lane angle and a fixed lane width ratio, and the image, and the position of the detected vehicle Detecting the coordinates and the ratio of the lane angle and the lane width of the detected lane, comparing the fixed lane angle and the fixed lane width ratio to determine the similarity, and correcting the position coordinates of the detected vehicle according to the similarity degree to detect the detection from the own vehicle It characterized in that it comprises a processor for estimating the distance to the vehicle.

The position coordinate of the vehicle is characterized in that it is a y-coordinate of a point where the wheel of the vehicle detected in the image is in contact with the road surface.

The lane width ratio may be a ratio between an actual lane width and the lane width.

The processor may calculate a similarity between the ratio of the lane angle and the lane width and the ratio of the fixed lane angle and the fixed lane width, and check whether the calculated similarity is equal to or greater than a threshold value.

The processor may estimate the distance using the detected vehicle position coordinates if the similarity is greater than or equal to a threshold value.

If the degree of similarity is less than a threshold, the processor extracts a homography from the image, converts the detected position coordinates of the vehicle into coordinates in a plane viewpoint using the homography, and converts the coordinates in the plane viewpoint to the coordinates in the plane viewpoint. It is characterized in that it is converted into coordinates in a reference camera geometry by applying a fixed coordinate transformation determinant, and the distance is estimated using the coordinates in the reference camera geometry.

The present invention acquires a front image of the own vehicle through a camera, detects lane information from the acquired image, calculates a change in movement in the vertical direction in the image due to road surface geometry and vehicle vibration, etc., based on the lane information, and considers this Therefore, since the distance from the own vehicle to the vehicle in front is estimated, the distance from the own vehicle to the vehicle in front can be accurately estimated using only the camera.

1 is a block diagram of an apparatus for estimating an inter-vehicle distance according to an embodiment of the present invention;
2 is a diagram for explaining a fixed coordinate change determinant;
3 is a diagram for explaining derivation of a distance estimation function;
4 is a view for explaining the detection of a vehicle position coordinates, a lane angle, and a lane width ratio;
5 is a view for explaining the position coordinate correction of the vehicle.
6 is a flowchart illustrating a method for estimating an inter-vehicle distance according to an embodiment of the present invention.

Terms such as "include", "comprise", "have" and the like described in this specification mean that the corresponding component may be embedded unless otherwise stated, so it does not exclude other components but other components It means that it can contain more elements.

In addition, terms such as “…unit”, “…group”, and “module” described in this specification mean a unit that processes at least one function or operation, which may be implemented by hardware or software or a combination of hardware and software. can In addition, articles such as "a", "an" and "the" in the context of describing the present invention are meant to include both the singular and the plural unless otherwise indicated herein or otherwise clearly contradicted by the context. can be used

Hereinafter, embodiments of the present invention will be described in detail with reference to the accompanying drawings.

The present invention obtains an image through a camera and the vehicle in the image in order to solve the problem that the distance of the object detected in the image cannot be accurately estimated (measured) when the camera geometry is changed due to the road surface and vibration, etc. And using the lane information detected by detecting the lane, calculates the vertical change in the image due to road surface geometry and vehicle vibration, etc., and estimates the distance from the own vehicle to another vehicle by considering the calculated vertical change. .

1 is a block diagram of an inter-vehicle distance estimation apparatus according to an embodiment of the present invention, FIG. 2 is a diagram for explaining a fixed coordinate change determinant, and FIG. 3 is a diagram for explaining derivation of a distance estimation function, 4 is a diagram for explaining the detection of a vehicle position coordinate, a lane angle, and a lane width ratio, and FIG. 5 is a diagram for explaining a vehicle position coordinate correction.

As shown in FIG. 1 , the apparatus for estimating the inter-vehicle distance includes a camera 110 , a memory 120 , and a processor 130 .

The camera 110 is installed in the vehicle to acquire a front image of the own vehicle in real time. In this case, the camera 110 acquires an image from a perspective view. The camera 110 transmits the acquired image to the processor 130 . In this embodiment, the camera 110 is installed in the front of the vehicle to obtain only the front image of the own vehicle as an example, but the present invention is not limited thereto. It may be implemented to acquire a front image, a rear image, and a side image of the vehicle.

The camera 110 includes images such as a charge coupled device (CCD) image sensor, a complementary metal oxide semi-conductor (CMOS) image sensor, a charge priming device (CPD) image sensor, and a charge injection device (CID) image sensor. It may be implemented as an image sensor of any one of the sensors.

The memory 120 stores a fixed coordinate transformation determinant (inverse homography, iH), a fixed lane width ratio, and a fixed lane angle. The memory 120 may store a program for the operation of the processor 130 , and may temporarily store input and/or output data.

The fixed coordinate transformation determinant is obtained in advance through the following process. First, a perspective view image Img1 and a top-view image Img2 are acquired using a camera when the vehicle V is positioned in the center of a general road (a road with an ideal road surface). . Here, the perspective image Img1 is an image acquired in a reference camera geometry situation. A vehicle V and lanes L1, L2, L3, and L4 are detected from the acquired perspective image Img1 and the plane image Img2, respectively. Then, the intersections (P ₁ , P ₂ , P ₃ , P ₄ ) of the lanes L1 and L2 detected in the perspective image Img1 and y=y1 and y=y2 are detected, and detected in the plane image Img2 The intersections (P' ₁ , P' ₂ , P' ₃ , P' ₄ ) of the traffic lanes L3 and L4 and y=y1 and y=y2 are detected. Homography the relationship between the intersection (P ₁ , P ₂ , P ₃ , P ₄ ) of the perspective image (Img1) and the intersection (P' ₁ , P' ₂ , P' ₃ , P' ₄ ) of the planar image (Img2) It can be expressed as (homography) H. That is, the relational expression for converting the perspective image Img1 into the planar image Img2 is obtained as homography H. Homography H is a 3×3 matrix. A fixed coordinate transformation determinant (iH) is obtained from the inverse of the obtained homography H.

The fixed lane angle is the angle θ between lanes detected in the perspective image Img1, and the fixed lane width ratio α is the ratio between the actual lane width d and the lane width p in the perspective image Img1. am. The equation for calculating the lane width ratio α can be expressed as [Equation 1].

Also, the memory 120 may store a distance estimation function f. The distance estimation function models the y-coordinate (the y-axis position of the bottom of the vehicle in the image) for each actual distance based on the least-squares method (refer to the graph of FIG. 3 ). Through this modeling, a distance estimation function based on the denominator of [Equation 2] is derived.

Here, x is the actual distance, and y is the location coordinates of the vehicle in the image.

Memory 120 is a flash memory (flash memory), hard disk (hard disk), SD card (Secure Digital Card), RAM (Random Access Memory, RAM), ROM (Read Only Memory, ROM), web storage (web storage) ) may be implemented as any one or more of the storage media.

The processor 130 detects a vehicle and a lane from the image (perspective image) received from the camera 110 . In this case, the processor 130 uses a learning-based vehicle and lane detection technology in the image.

When a vehicle in the image is detected, the processor 130 extracts (detects) the position coordinates of the detected vehicle (a position where the wheels of the vehicle and the road surface are in contact). In other words, the processor 130 detects the y-coordinate at which the lower end of the vehicle is located in the image. Here, the y-coordinate means a y-axis value in the image coordinate system. For example, the processor 130 detects the y-axis value y _V in which the lower end of the detected vehicle is located when the vehicle V in front is detected in the image as shown in FIG. 4 . Here, the front vehicle V means a vehicle located in front of a lane on which the host vehicle travels.

The processor 130 detects an angle θ' between lanes detected in the image. For example, the processor 130 detects an angle θ′ formed by the lanes L5 and L6 detected in the image shown in FIG. 4 .

The processor 130 detects a lane width ratio α′ using the lanes L5 and L6 detected in the image. The processor 130 calculates the lane width p in the image, assuming that the actual lane width d is the same as the actual lane width used when calculating the fixed lane width ratio (see FIG. 4 ). The processor 130 calculates the ratio of the actual lane width d to the obtained lane width p using [Equation 2].

When the processor 130 detects the ratio of the lane angle and the lane width from the received image, the processor 130 compares the detected lane angle and the lane width ratio with the fixed lane angle and the fixed lane width ratio stored in the memory 120 to determine the similarity.

The similarity μ is calculated using [Equation 3].

Here, ρ is a lane angle gain, and σ is a lane width ratio gain.

If the similarity is greater than or equal to a threshold (eg, 0.7), the processor 130 estimates the distance f(y _V ) from the own vehicle to the vehicle in front by using the coordinates (y _V ) of the lower end of the vehicle in the image.

On the other hand, if the similarity is less than the threshold, the processor 130 determines that a large change has occurred in the vertical direction of the image and extracts homography H' from the image input from the camera 110 . For example, when the camera geometry is changed by the road surface, the processor 130 extracts the homography H' from the image acquired through the camera 110 .

The processor 130 converts the position coordinate y of the vehicle into the coordinate y' in the plane view using the extracted homography H'. That is, as shown in FIG. 5 , the processor 130 calculates the position coordinate y' of the vehicle in a plan view by multiplying the position coordinate y of the vehicle by the homography H'. The processor 130 obtains the transformed coordinate y' in the reference camera geometry by using the fixed coordinate transformation determinant (iH). As shown in FIG. 5, the processor 130 is the transformed coordinate y' By multiplying the fixed coordinate transformation determinant (iH), the corrected vehicle position coordinate y" is obtained.

6 is a flowchart illustrating a method for estimating an inter-vehicle distance according to an embodiment of the present invention.

The processor 130 of the inter-vehicle distance estimation apparatus receives an image from the camera 110 (S110). The camera 110 acquires the front image of the own vehicle and transmits it to the processor 130 . Here, the front image is an image obtained from a perspective view.

The processor 130 detects a vehicle and a lane from the received image (S120). The processor 130 detects a vehicle in the image using a learning-based vehicle detection algorithm. Also, the processor 130 detects a lane in the image using a learning-based lane detection algorithm.

The processor 130 extracts (calculates) the position coordinates of the detected vehicle and the ratio of the lane angle and the lane width of the detected lane ( S130 ). The location coordinates of the vehicle are the values of the y-axis (y-coordinates) where the lower end of the vehicle in the image is located. The lane angle is the angle formed by the detected two lanes (both lanes of the lane in which the own vehicle is traveling), and the lane width ratio is the ratio of the actual lane width to the lane width in the image.

The processor 130 compares the extracted lane angle and lane width ratio with the fixed lane angle and fixed lane width ratio stored in the memory 120 to determine similarity ( S140 ). In other words, the processor 130 calculates the similarity using the extracted lane angle and lane width ratio and the fixed lane angle and fixed lane width ratio, and checks whether the calculated similarity is greater than or equal to a threshold value.

If the calculated similarity is equal to or greater than the threshold, the processor 130 estimates the distance d' from the host vehicle to the detected vehicle using the position coordinate y of the vehicle detected from the image (S150). That is, the processor 130 estimates the distance d' by applying the location coordinates y of the vehicle detected from the image to the distance estimation function f.

In S140, if the similarity is less than the threshold, the processor 130 extracts the homography H' from the image (S170).

The processor 130 corrects the position coordinate (y) of the vehicle in the image using the extracted homography (H') and the fixed coordinate transformation determinant (iH) ( S180 ). The processor 130 converts the location coordinates (y) of the vehicle in the image by multiplying the location coordinates (y) of the vehicle in the image by the extracted homography (H') into the location coordinates (y') of the vehicle in the plan view. The processor 130 applies the fixed coordinate transformation determinant (iH) to the position coordinates (y') of the vehicle in a plane view to obtain the position coordinates (y") of the vehicle in the reference camera geometry situation. In other words, the processor ( 130) is to convert the position coordinates of the vehicle in the image into the position coordinates of the vehicle in the image in the perspective view when the vehicle is located on an ideal plane road surface. Corrects vertical shake caused by road surface shape and vibration included in the image.

The processor 130 estimates the distance d' from the host vehicle to the detected vehicle using the corrected position coordinates y" of the vehicle (S190).

According to an embodiment of the present invention, the inter-vehicle distance estimation technology of the present invention can measure a precise distance of an object detected in an image during image-based autonomous driving.

In addition, the inter-vehicle distance estimation technique of the present invention may be applied to a vehicle detection algorithm and a pedestrian detection algorithm to estimate the distance from the own vehicle to another vehicle detected and/or the detected distance to the pedestrian.

In the above, even though all the components constituting the embodiment of the present invention are described as being combined or operated in combination, the present invention is not necessarily limited to this embodiment. That is, within the scope of the object of the present invention, all the components may operate by selectively combining one or more. In addition, although all the components may be implemented as one independent hardware, some or all of the components are selectively combined to perform some or all of the functions of the combined hardware in one or a plurality of hardware program modules It may be implemented as a computer program having Codes and code segments constituting the computer program can be easily inferred by those skilled in the art of the present invention. Such a computer program is stored in a computer readable storage medium (Computer Readable Media), read and executed by the computer, thereby implementing the embodiment of the present invention.

110: camera
120: memory
130: handler

Claims (12)

acquiring an image through a camera;
detecting a vehicle and a lane from the image;
detecting a ratio of a lane angle and a lane width between a position coordinate of a vehicle detected in the image and a lane detected in the image;
determining a similarity by comparing the ratio of the lane angle and the lane width with the ratio of the fixed lane angle and the fixed lane width; and
and estimating a distance from the own vehicle to the detected vehicle by correcting the position coordinates of the detected vehicle when it is determined that a movement change in the vertical direction occurs when the image is acquired based on the similarity A method of estimating the inter-vehicle distance.
According to claim 1,
The position coordinates of the vehicle are,
The method for estimating the inter-vehicle distance, characterized in that the y-coordinate of a point where the detected wheel of the vehicle contacts the road surface in the image.
According to claim 1,
The lane width ratio is
An inter-vehicle distance estimation method, characterized in that it is a ratio between the actual lane width and the lane width.
According to claim 1,
The step of determining the similarity is:
The method for estimating an inter-vehicle distance, comprising calculating a similarity between the lane angle and the lane width ratio and the fixed lane angle and the fixed lane width ratio, and checking whether the calculated similarity is equal to or greater than a threshold value.
5. The method of claim 4,
The step of estimating the distance is
If the similarity is greater than or equal to a threshold value, the inter-vehicle distance estimation method of claim 1, wherein the distance is estimated using the detected vehicle location coordinates.
5. The method of claim 4,
The step of estimating the distance is
if the similarity is less than a threshold, extracting a homography from the image;
converting the detected vehicle position coordinates into coordinates in a plane view using the homography;
converting into coordinates in a reference camera geometry by applying a fixed coordinate transformation determinant to coordinates in the plane view; and
and estimating the distance using coordinates in the reference camera geometry.
a camera for acquiring an image;
a memory storing a fixed lane angle and a fixed lane width ratio; and
A vehicle and a lane are detected from the image, the position coordinates of the detected vehicle and the ratio of the lane angle and the lane width between the detected lanes are detected, and the degree of similarity is determined by comparing the fixed lane angle and the fixed lane width ratio to the similarity. and a processor for estimating the distance from the own vehicle to the detected vehicle by correcting the position coordinates of the detected vehicle when it is determined that a movement change in the vertical direction has occurred when acquiring the image based on the estimation device.
8. The method of claim 7,
The position coordinates of the vehicle are,
The inter-vehicle distance estimation apparatus, characterized in that the y-coordinate of a point where the detected wheel of the vehicle contacts the road surface in the image.
8. The method of claim 7,
The lane width ratio is
The inter-vehicle distance estimating device, characterized in that it is a ratio between the actual lane width and the lane width.
8. The method of claim 7,
The processor is
and calculating a similarity between the ratio of the lane angle and the lane width and the ratio of the fixed lane angle and the fixed lane width, and checking whether the calculated similarity is equal to or greater than a threshold value.
11. The method of claim 10,
The processor is
If the similarity is greater than or equal to a threshold, the inter-vehicle distance estimation apparatus according to claim 1, wherein the distance is estimated using the detected vehicle position coordinates.
11. The method of claim 10,
The processor is
If the similarity is less than the threshold, a homography is extracted from the image, the detected vehicle position coordinates are converted to coordinates in a plane viewpoint using the homography, and coordinates are fixed to coordinates in the plane viewpoint. Converting to coordinates in the reference camera geometry by applying , and estimating the distance using the coordinates in the reference camera geometry.

Images