KR20180022277A - System for measuring vehicle interval based blackbox - Google Patents

Abstract

The present invention relates to a system for measuring a distance between vehicles based on black box images. The system comprises: a black box image photographing unit (100) configured to photograph the front of a vehicle; a lane and vanishing point detecting unit (200) configured to detect a lane and a vanishing point in an image received from the image photographing unit; a vehicle detecting unit (300) configured to detect a vehicle in the image received from the image photographing unit; a detection correcting unit (400) configured to correct information of the lane and vanishing point detecting unit and information of the vehicle detecting unit; and a unit (500) for measuring a distance between vehicles, configured to measure a distance between vehicles by information received from the detection correcting unit.

Description

[0001] The present invention relates to a black box image-based vehicle distance measuring system,

[0001] The present invention relates to an image-processing-based inter-vehicle distance measuring system using an image obtained through a single camera (vehicle black box), more specifically, to more accurately recognize a vehicle and a lane through machine learning and vanishing point correction, ADAS (Advanced Driver Assistance Systems) capable of measuring the headway distance using branch information.

The ADAS (Advanced Driver Assistance Systems) market is being activated all over the world, and the existing ADAS system has high accuracy, such as radar and radar, but expensive sensors are used. Therefore, there is a need to develop a less expensive and efficient inter-vehicle distance system.

The distance measurement sensor of the existing ADAS vehicle depends on the laser, so it is necessary to develop a more affordable distance meter as the accuracy of price is high and the accuracy is guaranteed.

In addition, the curved lane has a problem in that the distance measurement error rate is increased compared with the straight lane, and it is required to develop a technique that can minimize the error rate by supplementing the error.

Patent Document 1: Korean Patent No. 1268473 Patent Document 2: Korean Patent Registration No. 2012-0021445 Patent Document 3: Korean Patent Publication No. 2004-0037429

It is an object of the present invention to develop an ADAS (Advanced Driver Assistance Systems) market around the world, and to develop an inexpensive and efficient inter-vehicle distance system because existing ADAS systems use high-accuracy sensors such as radar and radar, There is a need.

It is also an object of the present invention to provide a distance-measuring sensor of a conventional ADAS vehicle, which is dependent on a laser, so that it is necessary to develop a less costly distance-measuring device while ensuring an accurate level of accuracy and cost.

In order to accomplish the above object, the present invention provides a black-box image pickup unit 100 for photographing a front of a vehicle; A lane and vanishing point detector 200 for detecting a lane and a vanishing point in an image received by the image pickup unit; A vehicle detecting unit (300) for detecting a vehicle from an image received by the image capturing unit; A detecting and correcting unit (400) for correcting the information of the lane and vanishing point detecting unit and the vehicle detecting unit; And an inter-vehicle distance detecting unit (500) for measuring an inter-vehicle distance based on the information received from the detection and correction unit.

According to the present invention, the lane and vanishing point detecting unit is a part for detecting a lane and a vanishing point, and includes a lane detecting unit 201, a lane determining unit 202, and a vanishing point detecting unit 203. to provide.

According to the present invention, the lane detecting unit recognizes the lane of the lane as the current driving lane, and the lane viewed from the image regardless of the lane of the lane can be divided into a specific area of interest and a slope based on a common characteristic having a specific slope The present invention provides a black box image-based distance measurement system for recognizing and detecting a result satisfying two conditions as a currently running lane.

According to the present invention, the lane correction unit is a part that changes a straight road from a curved road, and can reduce an error rate in a curved road running situation where an error rate becomes larger than a straight road. A black box image-based distance measurement system for setting a plurality of regions of interest in a certain region and converting the error values out of the reliability of ± σ into a straight line through a reverse projection transformation, to provide.

According to the present invention, the vanishing point detecting unit detects a vanishing point by finding an intersection of straight lines detected one by one from the corrected lane information through the car bridge bridge, and provides a black box image-based vehicle distance measuring system.

According to the present invention, the vehicle detecting unit 300 detects a vehicle through machine learning, and the machine learning is performed by using a CNN (Convolutional Neural Network) And a black box image based vehicle distance measurement system that measures the accurate vehicle and position even in the external environment (light source and weather).

According to the present invention, the detection and correction unit 400 includes a vanishing point correction unit 401 and a vehicle position correction unit 402 as parts for reducing a detection error when the field of view of the camera changes or changes in the vertical and horizontal directions Based vehicle-based distance measurement system.

According to the present invention, the vanishing point correction unit calculates the angle of the vanishing point measured with respect to the reference vanishing point by using the trigonometric function using the vanishing point measured in real time during operation and the predetermined reference vanishing point, and then moves the entire image by the calculated angle The present invention provides a black box image-based vehicle distance measuring system.

According to another aspect of the present invention, the vehicle position correcting unit detects the corrected vehicle position by updating the new vehicle position by moving the vehicle position in the image shifted to the reference vanishing point through the vanishing point correction unit. Provides an inter-vehicle distance measurement system.

Further, according to the present invention, the inter-vehicle distance detecting section detects the vehicle-to-vehicle distance from the image information corrected through the detection correction section to the pixel value of the vehicle from the reference line (bonnet of the vehicle) The present invention provides a black box image-based distance measurement system for measuring an inter-vehicle distance in comparison with distance data.

According to the black box image-based headway distance measurement system of the present invention, it is possible to more accurately measure the vanishing point of the lane and the headway distance to the preceding vehicle based on the image of the front of the vehicle taken by the image capturing unit.

According to the black-box image-based headway distance measurement system according to the embodiment, the position of the vanishing point measured in real time using a trigonometric function is corrected to the position of the reference vanishing point set in advance, And correcting the position of the vehicle, it is possible to measure the headway distance more easily and accurately.

According to the black box image-based vehicle distance measuring system according to the present invention, the error rate with respect to the vanishing point and the inter-vehicle distance can be minimized by adjusting the region of interest for detecting the vanishing point of the lane or the vehicle ahead.

According to the black box image-based headway distance measurement system according to the present invention, in order to compensate for the problem that the distance measurement error rate increases in a curved lane, the error rate is minimized by converting the curved lane into a straight lane and measuring the distance.

According to the black box image-based headway distance measurement system according to the present invention, autocalibration can be easily implemented without any additional calibration operation.

1 is a view showing an image processing based inter-vehicle distance measuring system.
Fig. 2 shows a state in which a lane is detected and a vanishing point is detected using the detected lane.
Fig. 3 shows how the vehicle is detected in the image and the distance is measured using the pixel.
Fig. 4 shows a state in which a curved lane is converted into a straight lane to measure the distance.
Fig. 5 is a diagram for machine learning using the feature points of the vehicle when the vehicle is detected. Fig.
Fig. 6 is a diagram showing a region of interest set for determining a reference line (bonnet of a vehicle) that does not change during vehicle distance measurement.

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

First, in the drawings, it is noted that the same components or parts are denoted by the same reference numerals as possible. In the following description of the present invention, a detailed description of known functions and configurations incorporated herein will be omitted so as to avoid obscuring the subject matter of the present invention.

The terms "about "," substantially ", etc. used to the extent that they are used herein are intended to be taken to mean an approximation of, or approximation to, the numerical values of manufacturing and material tolerances inherent in the meanings mentioned, It is used to prevent unauthorized exploitation by an unscrupulous infringer from disclosing the exact or absolute numerical value to help.

FIG. 1 shows an image processing based automatic traveling system according to an embodiment, FIG. 2 shows a state where a vanishing point is detected through lane recognition, FIG. 3 shows a state where a vehicle is detected through machine learning, Fig. 4 shows a state in which a curve lane is converted into a linear lane.

Referring to FIG. 1, a black box image-based vehicle distance measuring system 10 according to an embodiment includes a black box image capturing unit 100, a lane and vanishing point detecting unit 200, a vehicle detecting unit 300, The lane and vanishing point detecting unit 200 includes three lane detecting units 201, a lane determining unit 202 and a vanishing point detecting unit 203. The detecting and correcting unit 400 Includes a vanishing point correction unit 401 and a vehicle position correcting unit 402. [

The black box image capturing unit 100 photographs the front of the vehicle. In the present invention, the black box camera refers to a black box camera that photographs a traveling image necessary for a vehicle distance from a black box camera.

Hereinafter, detection of vanishing point or inter-vehicle distance through image processing will be described.

The lane detecting unit 201 can detect a lane from the image received by the image capturing unit 100. The disappearance detecting unit 203 detects the lane detected by the lane detecting unit 201 and the lane correcting unit 202, The vanishing point of the lane can be detected.

In particular, referring to FIG. 2, in the lane calibration unit 300, an image received by the image capturing unit 100 by an inverse perspective transformation, for example, may be converted from 3D coordinates to 2D coordinates.

In particular, the inverse perspective transformation may induce homography using an index of a corner portion of a lane that looks like a trapezoid in an image received by the image capturing unit 100.

The homography is defined in a homogeneous coordinate system, and its general formula is as follows.

In this case, c is a nonzero constant, (u, v, 1) T represents x ', and (x, y, 1) T represents x.

이다.And,

to be.

In relation to H, the factors h1, h2, h4 and h5 are the rotation / scale factor, the h3 and h6 factors are the translation factors, and the h7 and h8 factors are the perspective factors.

Thus, the lane correcting unit 203 can transform a straight lane or a curved lane that looks like a trapezoid into a straight line by modifying the rotation / scale factor, the translational factor, and the perspective factor in the matrix for H in the homography equation.

The image received by the image capturing unit 100 is converted from the 3D coordinate to the 2D coordinate by the back projection transformation in the lane correcting unit 300 and then the linear component of the lane is extracted by Hough Transform .

The Hough transform extracts an edge from an image for a straight line detection of a lane and then detects a myriad of points that become an edge so that the value of r (the vertical distance from the reference point to the edge) and theta (the distance between the horizontal axis and the extension line of r Angle) of the lane becomes equal to the threshold value, the straight line of the lane is detected.

However, the manner of detecting lanes in the lane detecting section 201 and the lane determining section 202 is not limited to the above-described method, and any of them can be used as long as the lane can be detected more accurately.

3, the vanishing point detector 203 can detect the vanishing point of the lane from the lane detected through the lane detecting unit 201 and the lane determining unit 202. [

The vanishing point of the lane is the point where the line and the line meet, for example, the configuration of the lane can be the point where the two lines meet far away.

At this time, the vanishing point detected by the vanishing point detecting unit 203 is the current vanishing point reflecting the position or angle of the image capturing unit 100, and an incorrect vanishing point can be detected by the external condition. For example, the external condition may include lane spacing.

Accordingly, the vanishing point of the lane detected by the vanishing point detecting unit 203 may be transmitted to the vanishing point corrector 401 belonging to the detecting and correcting unit 400 for more accurate vanishing point detection.

At this time, the vanishing point detecting unit 203 detects the vanishing point of the lane in real time in accordance with the running of the vehicle, and at the same time, the disappearing point correcting unit 401 can correct the vanishing point of the lane in real time.

For this, a vanishing point on the image receiving unit, that is, a vanishing vanishing point to be a reference, may be input in advance to the vanishing point corrector 401.

Thereby, the vanishing point correction unit 401 can correct the vanishing point by matching the vanishing point of the lane detected by the vanishing point detecting unit 203 to the previously inputted reference vanishing point.

Specifically, the vanishing point on the image receiving unit previously input may be the center of the image capturing unit 100, and the vanishing point correcting unit 401 may calculate the vanishing point of the lane detected by the vanishing point detecting unit 203, Can be moved to the center.

The vanishing point correction may be determined by the following equation by using the following equation: Difference between the reference vanishing point input beforehand and the vanishing point of the lane detected by the vanishing point detecting unit 203.

Here, θ1 is the horizontal angle difference between the vanishing point on the image receiving unit 200 and the vanishing point of the lane detected by the vanishing point detecting unit 400, w is the width of the image receiving unit 200, f is the image capturing unit 100 ). And x is the x-axis coordinate value of the vanishing point of the lane detected by the vanishing point detecting unit 400. [

Therefore, by moving the vanishing point of the lane detected by the vanishing point detecting unit 400 by the horizontal angle difference determined by the above equation on the x axis, it can be matched with the vanishing point on the image receiving unit 200 previously input.

Further, the difference in vertical angle between the previously entered reference vanishing point and the vanishing point of the lane detected by the vanishing point detecting unit 203 can be determined by the following equation.

2 is a vertical angle difference between a reference vanishing point input in advance and a vanishing point of a lane detected by the vanishing point detecting unit 400. H is a height of the image receiving unit 200 and f is a focal length of the image pickup unit 100 to be. And y is the y-axis coordinate value of the vanishing point of the lane detected by the vanishing point detecting unit 400. [

Therefore, by moving the vanishing point of the lane detected by the vanishing point detecting unit 203 by the vertical angle difference determined by the above equation on the y axis, it can be matched with the previously inputted reference vanishing point.

As described above, the vanishing point correction unit 401 is coded with two formulas, so that the vanishing point of the lane detected by the vanishing point detecting unit 203 can be easily corrected in real time. By correcting the vanishing point, 100 can be corrected.

In other words, the vanishing point serves as a reference point for image correction, and images included in the image received by the image pickup unit 100 by the correction of the vanishing point move in the same manner.

Meanwhile, the black box image-based headway distance measurement system 10 according to the embodiment can adjust the ROI set in the image received by the image pickup unit 100 to more accurately measure the vanishing point.

The region of interest is for vanishing point detection, and as can be seen in FIG. 4, the region of interest may be an area within the dotted line denoted by A.

At this time, the size of the region of interest for detection of the vanishing point may be adjusted based on the image received by the image receiving unit. In other words, the size of the region of interest for the vanishing point detection may be increased or decreased based on the image received by the image receiving unit.

For example, the slope of the lane can be detected from the image received by the image pickup section 100, and the size of the region of interest can be adjusted based on the slope of the lane.

Specifically, in the case of a straight lane in which the slope of the lane is 0, the size of the region of interest for detecting the vanishing point is increased from the image received by the image photographing unit 100, and the slope of the lane from the image received by the image photographing unit 100 Is smaller than 0, it is possible to reduce the size of the region of interest for the vanishing point detection.

As a result, the area of interest is automatically adjusted to match the situation of the running road, such as a highway or a curved road, so that the vanishing point can be detected more accurately.

This adjustment of the area of interest is important in relation to the automatic running of the vehicle, and the detection of the vanishing point may be different depending on the area of interest.

In addition, the region of interest can be adjusted to more accurately detect lane detection, which is a preliminary task of detecting vanishing points.

For example, an edge of a lane may be detected from an image received by the image capturing unit 100, and a region of interest may be set on the edge of the lane.

Specifically, a plurality of edges of the lane can be detected from the image received by the image capturing unit 100, and the respective regions of interest can be set on the edges of the lane.

Accordingly, the image received by the image capturing unit 100 can be divided into a plurality of segments, the edges of the lane are individually linearized by the lane detecting unit 201, and the vanishing point detecting unit 203 detects vanishing points It is possible to detect a more accurate vanishing point.

Referring to FIG. 5, the vehicle detecting unit 300 can detect a vehicle positioned in front of the vehicle from the image received by the image capturing unit 100. FIG.

For example, the vehicle detecting unit 300 can detect a forward vehicle using a CNN (Convolutional Neural Network).

The circuit neural network technology basically abstracts an image of an object (vehicle) to be detected, compresses the abstracted image so that only the most important information is left, abstracts the compressed image again, and compresses the compressed image repeatedly, It is a technology to classify whether or not the vehicle is accumulated by accumulating information accumulated in the image layer.

Therefore, information on a plurality of vehicles may be input in advance to the vehicle detection unit 300. For example, information on a plurality of vehicles may include a plurality of features for a plurality of vehicle types

The vehicle can be detected by matching the input information of the plurality of vehicles and the image received by the image capturing unit 100.

The detection of the vehicle by the vehicle detection unit 300 can be performed simultaneously and separately with the detection of the lane by the lane detecting unit 201 and the detection of the vanishing point by the vanishing point detecting unit 203. [ In other words, the lane, the vanishing point, and the vehicle information can be obtained simultaneously from the image received by the image capturing unit 100.

As described above, the information about the vehicle detected by the vehicle detecting section 300, in particular, the position information of the vehicle, can be corrected by the vehicle position correcting section 402. [

The vehicle position correcting unit 402 can be performed together with the correction of the vanishing point in the vanishing point correcting unit 401 and can move the position of the vehicle detected by the vehicle detecting unit 300 by correcting the vanishing point. Accordingly, the position of the vehicle can be corrected together with the vanishing point of the image received by the image capturing unit 100, and finally the image capturing unit 100 is provided with the vanishing point correction unit 401 and the vehicle position correction unit 402 The corrected image can be displayed.

The inter-vehicle distance from the preceding vehicle in the inter-vehicle distance calculating unit 500 can be calculated after the position correction of the vehicle performed by the vehicle position correcting unit 402 is performed as described above.

At this time, the inter-vehicle distance calculating unit 500 can utilize the number of pixels of the image according to the inter-vehicle distance. (See Fig. 6)

For example, in the inter-vehicle distance calculating unit 500, the number of pixels of the image according to the reference inter-vehicle distance can be input in advance, and the number of pixels of the image corrected by the vanishing point correction unit 401 and the vehicle position correcting unit 402 The headway distance can be calculated.

Specifically, the number of pixels in an image can be input in advance for each distance that is a reference by using distance data measured on an actual road. For example, when the reference inter-vehicle distance is 1 km, if the number of pixels of the image is input in advance of 100 pixels, the number of pixels of the forward vehicle in the image corrected by the vanishing point correction unit 401 and the vehicle position correction unit 402 is 100 pixel, it can be inferred that the distance between the vehicle and the preceding vehicle is 1 km.

Alternatively, it is a matter of course that the number of pixels of the image according to the reference inter-vehicle distance is input in advance to the inter-vehicle distance calculating unit 500, and the inter-vehicle distance can be derived from the proportional relation. When the reference inter-vehicle distance is 1 km, if the number of pixels of the image is input in advance of 100 pixels, if the number of pixels of the preceding vehicle is 50 pixels in the image corrected by the vanishing point correction unit 401 and the vehicle position correcting unit 402, The distance between the vehicle and the vehicle is 2 km.

After the inter-vehicle distance is calculated in the inter-vehicle distance calculating unit 500 as described above, the lower inter-vehicle distance of the preceding vehicle can be displayed in the image received by the image photographing unit 100, as shown in FIG. Thereby, the driver of the vehicle can easily confirm the distance between the vehicle and the preceding vehicle.

In addition, by correcting the position of the vanishing point and the position of the vehicle before calculating the inter-vehicle distance, it is possible to accurately calculate the inter-vehicle distance even when the image pickup section 100 such as a camera is shaken.

As described above, the image processing based automatic traveling system according to the embodiment can easily measure the vanishing point of the lane and the inter-vehicle distance with respect to the preceding vehicle based on the image of the front of the vehicle photographed by the image photographing unit, The vehicle distance can be more accurately measured by correcting the vanishing point and the position of the vehicle in real time using a function (trigonometric function). In addition, it is possible to realize auto calibration which is performed simultaneously with running without a separate camera calibration. Also, by processing images taken by a video shooting unit provided in a vehicle such as a black box camera, Various information can be obtained when driving.

Although the present invention has been described in connection with what is presently considered to be practical exemplary embodiments, it is to be understood that the invention is not limited to the disclosed embodiments, but, on the contrary, And various modifications and changes may be made thereto without departing from the scope of the present invention. Accordingly, the spirit of the present invention should not be construed as being limited to the embodiments described above, and all of the equivalents or equivalents of the claims, as well as the claims, will fall within the scope of the present invention.

10: Black box image-based distance measurement system
100:
200: lane and vanishing point detector
201: lane detecting section
202: lane departure government
203: Vanishing point detector
300:
400:
401: Vanishing point correction unit
402: vehicle position correcting section
500: inter-vehicle distance detector

Claims (10)

A black box image capturing unit (100) for photographing the front of the vehicle;
A lane and vanishing point detector 200 for detecting a lane and a vanishing point in an image received by the image pickup unit;
A vehicle detecting unit (300) for detecting a vehicle from an image received by the image capturing unit;
A detecting and correcting unit (400) for correcting the information of the lane and vanishing point detecting unit and the vehicle detecting unit;
And an inter-vehicle distance detector (500) for measuring an inter-vehicle distance using information received from the detection and correction unit.
The method according to claim 1,
The lane and vanishing point detecting unit 200 includes a lane detecting unit 201, a lane determining unit 202, and a vanishing point detecting unit 203, which detect a lane and a vanishing point.
3. The method of claim 2,
The lane detecting unit is a part for recognizing a lane of a lane as a current driving lane. A lane viewed from an image regardless of which lane is used satisfies both conditions by using a specific area of interest and a slope based on a common characteristic having a specific slope Based vehicle distance measurement system that recognizes and detects the result of the measurement as a currently running lane.
3. The method of claim 2,
The lane correction unit changes the straight road from a curved road to a curved road where the error rate is larger than that of the straight road, thereby reducing the error rate. A black box image-based distance measurement system for setting a plurality of ROIs in a certain region and converting the error values out of the reliability of ± σ into a straight line through an inverse projection transformation,
3. The method of claim 2,
Wherein the vanishing point detecting unit detects a vanishing point by finding an intersection of straight lines detected one by one from the corrected lane information through the car bridge bridge.
The method according to claim 1,
The vehicle detecting unit 300 detects a vehicle through machine learning, and the machine learning is performed by using a Convolutional Neural Network (CNN), a vehicle that does not fit the data used for the initial learning (Vehicle type) and external environment (light source and weather).
The method according to claim 1,
The detection and correction unit 400 includes a vanishing point correction unit 401 and a vehicle position correction unit 402. The vanishing point correction unit 401 and the vehicle position correction unit 402 are provided for reducing a detection error when a view is changed in vertical, Distance measurement system.
8. The method of claim 7,
Wherein the vanishing point correction unit calculates the angle of the vanishing point measured with respect to the reference vanishing point using a vanishing point measured in real time during operation and a predetermined reference vanishing point using a trigonometric function and then moves the entire image by the obtained angle. Image Based Vehicle Distance Measurement System.
9. The method of claim 8,
Wherein the vehicle position correcting unit detects the corrected vehicle position by updating the new vehicle position by moving the vehicle position from the image moved in accordance with the reference vanishing point through the vanishing point correcting unit.
The method according to claim 1,
The vehicle-to-vehicle distance detecting unit compares the pixel value of the corrected image information from the reference line (bonnet of the vehicle) to the vehicle detected through machine learning (lower wheel portion) with the distance data per pixel input in advance in the system An inter - vehicle distance measurement system based on a black box measuring the inter - vehicle distance.

Images