KR102295579B1 - Apparatus and mathod for detecting obstacle of a vehicle - Google Patents

Abstract

A vehicle obstacle detection apparatus and method for detecting various obstacles in a driving image, comprising: an image acquisition unit for acquiring an image; a characteristic map generation unit for generating a characteristic map of the acquired image; A scan line setting unit for setting at least one scan line, a spatiotemporal pattern map generating unit generating a spatiotemporal pattern map corresponding to the set scan line, an obstacle detecting unit detecting an obstacle by analyzing the generated spatiotemporal pattern map; It may include a control unit for controlling the operation of the vehicle according to the obstacle.

Description

Apparatus and method for detecting obstacles in a vehicle {APPARATUS AND MATHOD FOR DETECTING OBSTACLE OF A VEHICLE}

The present invention relates to a vehicle, and more particularly, to a vehicle obstacle detecting apparatus and method for detecting various obstacles in a driving image.

Obstacle detection technology in a general driving image is divided into using the characteristics of the obstacle, recognizing the shape of the obstacle, and analyzing the motion of the obstacle.

Methods using the characteristics of obstacles include methods using vehicle symmetry, shadows, or SIFT.

The problem with this method is that, in a frame in which the features of the obstacle are not fully displayed, the performance is greatly reduced, and the probability that the features effective for detecting the obstacles are not fully appeared is not large.

In addition, the method of recognizing the shape of the obstacle is to detect only the obstacle included in the target of the classifier using a classifier.

The problems of this method are that the calculation of the classifier is complicated, so that real-time calculation is difficult, and obstacles other than the type specified in the classifier cannot be recognized.

In addition, in the method of analyzing the motion, it is difficult to obtain motion information from the driving image, and at this time, a correspondence matching method such as optical flow is mainly used. Since detection information is often required, the first problem may appear at the same time, and if the motion is large, the error may be large and the detection performance may be reduced.

Accordingly, there is a demand for an obstacle detection method capable of accurately and quickly detecting an obstacle in a future driving image in a simple manner.

An object of the present invention is to provide an apparatus and method for detecting an obstacle in a vehicle capable of detecting a rapidly moving obstacle by generating a space-time pattern map corresponding to a scan line.

The technical problems to be achieved in the present invention are not limited to the technical problems mentioned above, and other technical problems not mentioned will be clearly understood by those of ordinary skill in the art to which the present invention belongs from the following description. will be able

In order to solve the above technical problem, an apparatus for detecting an obstacle for a vehicle according to an embodiment of the present invention includes an image acquisition unit for acquiring an image, a characteristic map generator for generating a characteristic map of the acquired image, and generation A scanline setting unit that sets at least one scanline corresponding to the set characteristic map, a spatiotemporal pattern map generating unit that generates a spatiotemporal pattern map corresponding to the set scanline, and an obstacle detection by analyzing the generated spatiotemporal pattern map It may include an obstacle detecting unit, and a control unit for controlling the operation of the vehicle according to the detected obstacle.

In addition, the method for detecting an obstacle in a vehicle according to an embodiment of the present invention includes acquiring an image, generating a characteristic map of the acquired image, and setting at least one scanline corresponding to the generated characteristic map. and generating a spatiotemporal pattern map corresponding to the set scan line, analyzing the generated spatiotemporal pattern map to detect an obstacle, and controlling the operation of the vehicle according to the detected obstacle. can

An apparatus and method for detecting an obstacle in a vehicle according to at least one embodiment of the present invention configured as described above is a technology for analyzing changes in an obstacle using a continuous image obtained by a front camera and detecting an obstacle showing a specific change, It provides the effect of detecting an obstacle moving faster than the own vehicle by using the space-time pattern map.

In addition, since the present invention detects an obstacle moving faster than the own vehicle, it provides an effect that the calculation speed is faster than the existing methods of analyzing the original driving image.

In addition, the present invention provides the effect that the processing time for detecting the obstacle is fast because it is possible to use the temporal change and the spatial change of the driving image at the same time.

In addition, the present invention provides an effect that is not affected by camera performance and image acquisition environment, such as lens distortion, because an obstacle is detected based on a spatio-temporal pattern map rather than a method of detecting an obstacle based on an original image. .

The effects obtainable in the present invention are not limited to the above-mentioned effects, and other effects not mentioned may be clearly understood by those of ordinary skill in the art to which the present invention belongs from the following description. will be.

1 is a block diagram illustrating an apparatus for detecting an obstacle for a vehicle according to the present invention.
FIG. 2 is a block diagram illustrating the scan line setting unit of FIG. 1 .
FIG. 3 is a block diagram illustrating the spatiotemporal pattern map generator of FIG. 1 .
4 to 7 are diagrams for explaining a process of setting a scan line.
8 is a diagram for explaining a process of generating a characteristic map.
9 is a view showing a scan line set in an original image.
10 is a diagram for explaining a process of generating a spatiotemporal pattern map.
11 is a diagram showing an accumulated spatiotemporal pattern map and its sub-units.
12 is a diagram illustrating a trend line slope of a spatio-temporal pattern map.
13 is a diagram illustrating an obstacle moving speed according to a slope of a trend line.
14 is a diagram for explaining a micro-gradient calculation process.
15 to 17 are views showing a spatiotemporal pattern map according to an obstacle detection result.
18 is a flowchart illustrating a method for detecting an obstacle in a vehicle according to the present invention.

Hereinafter, with reference to the accompanying drawings, embodiments of the present invention will be described in detail so that those of ordinary skill in the art to which the present invention pertains can easily implement them. However, the present invention may be embodied in various different forms and is not limited to the embodiments described herein. And in order to clearly explain the present invention in the drawings, parts irrelevant to the description are omitted, and similar reference numerals are attached to similar parts throughout the specification.

Throughout the specification, when a part "includes" a certain component, it means that other components may be further included, rather than excluding other components, unless otherwise stated. In addition, terms such as “…unit”, “…group”, and “module” described in the 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. have.

Throughout the specification, when a part "includes" a certain component, it means that other components may be further included, rather than excluding other components, unless otherwise stated. In addition, parts indicated with the same reference numerals throughout the specification mean the same components.

Hereinafter, an apparatus and method for detecting an obstacle in a vehicle that can be applied to embodiments of the present invention will be described in detail with reference to FIGS. 1 to 18 .

1 is a block diagram illustrating an obstacle detecting apparatus for a vehicle according to the present invention, FIG. 2 is a block diagram illustrating a scan line setting unit of FIG. 1, and FIG. 3 is a spatiotemporal pattern map generation unit of FIG. It is a block diagram for explanation.

1 to 3 , the apparatus 1000 for detecting an obstacle for a vehicle according to the present invention includes an image acquisition unit 100 , a characteristic map generation unit 200 , a scan line setting unit 300 , and a space-time pattern. It may include a map generator 400 , an obstacle detector 500 , and a controller 600 .

Here, the image acquisition unit 100 may be a camera that acquires an image.

For example, the camera may be mounted on the front of the vehicle to capture an image of the front of the vehicle in real time.

In addition, the characteristic map generator 200 may generate a characteristic map of the acquired image.

Here, the characteristic map generator 200 may generate the characteristic map based on the edge of the acquired image.

The reason for creating a feature map is that when a spatio-temporal pattern map is generated based on an acquired original image, the amount of data computation is too large. Therefore, it is possible to quickly detect an obstacle by reducing the processing time.

Also, the feature map generator 200 may remove noise from the feature map when the feature map is generated based on the edge of the acquired image.

The reason is that, if the noise of the feature map is removed, only the necessary components are more clearly seen.

In some cases, the feature map generator 200 may further perform a post-processing process of applying smoothing after binarizing the generated feature map.

Next, the scanline setting unit 300 may set at least one scanline corresponding to the generated characteristic map.

Here, as shown in FIG. 2 , the scanline setting unit 300 includes a front direction vector calculator 310 , a first scanline generator 320 , a second scanline generator 330 , and a scanline corrector. (340) may be further included.

In this case, the front direction vector calculator 310 may calculate a front direction vector of the camera from the characteristic map.

For example, the front direction vector calculator 310 may calculate a ground coordinate corresponding to the main point of the camera and a ground coordinate of the camera, and calculate a front direction vector based on a difference between the two calculated coordinates.

In addition, the first scanline generator 320 may generate a first scanline in the ground coordinate system based on the calculated front direction vector.

Here, the first scan line generation unit 320 obtains a straight line orthogonal to the calculated front direction vector, divides the selected region into a plurality of line segments based on the obtained straight line, and determines the midpoint of the divided line segment as a reference point, A first scanline parallel to the front direction vector may be generated while passing through a predetermined reference point.

For example, the area selected based on the obtained straight line may be an area for detecting an obstacle.

In addition, the first scan line may include a front direction vector and a corresponding reference point.

For example, there may be a plurality of first scan lines in the ground coordinate system, and the plurality of first scan lines may be parallel to each other.

Next, the second scanline generator 330 may generate a second scanline in the image coordinate system based on the generated first scanline.

For example, the second scanline generating unit 330 obtains a horizon based on the generated first scanline, obtains a vanishing point based on a straight line converted from the first scanline into image coordinates, and then obtains a vanishing point on the converted straight line. A second scan line passing through the reference point and the vanishing point may be generated.

Here, the second scan line generator 330 may set a straight line passing through the vanishing point as the horizon when the first scan lines in the image coordinate system gather at one vanishing point when obtaining the horizon.

In addition, the second scanline generator 330 may convert the first scanline in the ground coordinate system into a straight line in the image coordinate system when calculating the vanishing point, and set an intersection point between the converted straight line and the horizon as the vanishing point.

Next, the scan line corrector 340 may correct an error of the generated second scan line.

Here, the scan line corrector may correct an error of the second scan line by removing a portion of the second scan line located near the horizon.

Meanwhile, the spatiotemporal pattern map generation unit 400 may generate a spatiotemporal pattern map corresponding to the set scan line.

Here, the spatiotemporal pattern map generation unit 400 may include a first generation unit 410 , a storage unit 420 , and a second generation unit 430 , as shown in FIG. 3 .

In this case, the first generator 410 may generate a row image corresponding to each scan line for each frame.

In addition, the storage unit 420 may accumulate and store one or more frames of the generated row images of each scan line.

Next, the second generator 430 may generate a spatiotemporal pattern map in which row images of each scan line are accumulated, based on the row images stored in the storage 420 .

For example, the spatiotemporal pattern map may include a plurality of sub-units, and the plurality of sub-units may include a row image in which at least three frames are accumulated.

Next, the obstacle detection unit 500 may detect the obstacle by analyzing the generated space-time pattern map.

Here, the obstacle detection unit 500 may analyze the generated space-time pattern map to obtain a gradient of a trend line included in the space-time pattern map, and may detect an obstacle based on the gradient.

Also, when the slope is positive, the obstacle detection unit 500 may recognize an obstacle faster than the own vehicle, and if the slope is negative, it may recognize an obstacle slower than the own vehicle.

And, when there are a plurality of trend lines included in the spatio-temporal pattern map, the obstacle detection unit 500 may select the brightest trend line as the representative trend line and obtain the slope of the selected representative trend line.

In some cases, when there are a plurality of representative trend lines, the obstacle detection unit 500 may select one representative trend line based on the most constant rule and obtain the slope of the selected representative trend line.

Next, the controller 600 may control the operation of the vehicle according to the detected obstacle.

As described above, the present invention is a technology for analyzing changes in obstacles using continuous images obtained by a front camera and detecting obstacles showing a specific change, and it is possible to detect obstacles moving faster than the own vehicle using a spatiotemporal pattern map.

In addition, since the present invention detects an obstacle moving faster than the own vehicle, the calculation speed may be faster than the existing methods of analyzing the original driving image.

In addition, since the present invention can use the temporal change and the spatial change of the driving image at the same time, the processing time for detecting the obstacle can be fast.

In addition, since the present invention detects an obstacle based on a spatio-temporal pattern map rather than a method of detecting an obstacle based on an original image, it may not be affected by camera performance and image acquisition environment such as lens distortion.

4 to 7 are diagrams for explaining a process of setting a scan line.

4 to 7 , in the present invention, in order to generate a space-time pattern map (STP Map) for detecting a motion direction, information about the vehicle front direction is required.

The present invention assumes that since the camera of the vehicle is disposed toward the front of the vehicle, the front direction of the camera coincides with the front direction of the vehicle.

Therefore, in the present invention, as shown in FIG. 7 , the ground coordinates corresponding to the main point of the camera and the ground coordinates of the camera are calculated using a projective transformation, and the camera front direction vector 510 can be obtained through the difference. .

That is, the front direction vector calculator of the present invention may calculate the ground coordinates corresponding to the main points of the camera and the ground coordinates of the camera, and calculate the front direction vector 510 based on the difference between the two calculated coordinates.

Next, in the present invention, as shown in FIG. 7 , a straight line 520 orthogonal to the camera front direction vector 510 is obtained, and then an area 550 to detect a cut-in obstacle on the straight line 520 . is selected, and the selected area 550 is uniformly divided.

And, in the present invention, as shown in FIG. 7 , with the midpoint of the divided line segment 530 as the reference point 540 , a straight line passing through the reference point 540 and parallel to the camera front direction vector 510 is a virtual scan in the ground coordinate system. Select a line (virtual scanline) 530 .

Accordingly, as shown in FIG. 7 , a virtual scanline 530 is expressed and stored as a front direction vector 510 and a corresponding reference point 540 .

And, as shown in FIG. 4 , since all of the virtual scan lines of the ground coordinate system are parallel, they are gathered at one vanishing point P1 in the image coordinate system, and this vanishing point P1 exists on the horizon C.

Next, in the present invention, after obtaining the horizon (C) in the projective transformation, one of the virtual scan lines (A) in the ground coordinate system is converted into image coordinates. Find the vanishing point (P1) through the intersection.

And, when the image distortion is removed, since the virtual scan line (A) is converted into a straight line in the image coordinate system, each scan line (B) is formed by connecting the point on the image corresponding to the reference point (P2) and the vanishing point (P1). create

At this time, as shown in FIG. 4 , the scan line starts from the vanishing point P1 and passes the reference point P2 , so it is formed below the horizon C.

Further, in the present invention, in the vicinity of the horizon C of the scan line B, since the motion of the detection target is small and an error occurs, a process of removing a certain length near the horizon C is performed.

That is, as shown in FIG. 7 , the first scanline generator of the present invention obtains a straight line 520 orthogonal to the calculated front direction vector 510 , and selects a plurality of regions 550 based on the obtained straight line 520 . A virtual scan line 540 parallel to the front direction vector 510 may be generated by dividing the segment into line segments of .

Here, the area 550 selected based on the obtained straight line may be an area for detecting an obstacle.

In addition, the virtual scan line 530 of the present invention may include a front direction vector 510 and a corresponding reference point 540 .

Also, as shown in FIG. 7 , a plurality of virtual scan lines in the ground coordinate system may be provided, and the plurality of virtual scan lines may be parallel to each other.

Next, as shown in FIG. 4 , the second scanline generator of the present invention obtains a horizon C based on the generated virtual scanline A, and based on a straight line converted from the virtual scanline A into image coordinates. , after obtaining the vanishing point P1, a scan line B passing through the reference point P2 and the vanishing point P1 on the converted straight line may be generated.

Here, when the second scan line generator of the present invention obtains the horizon C, if virtual scan lines A in the image coordinate system are gathered at one vanishing point P1, a straight line passing through the vanishing point P1 is drawn on the horizon ( C) can be set.

In addition, the second scan line generator of the present invention converts the virtual scan line (A) in the ground coordinate system into a straight line in the image coordinate system when calculating the vanishing point (P1), and calculates the intersection of the converted straight line and the horizon (C) It can be set as the vanishing point (P1).

Next, the scan line corrector of the present invention may correct the error of the scan line B by removing a part of the scan line B positioned near the horizon C.

In addition, the present invention, as shown in FIG. 5 , in the forward driving image, a method of detecting a movement direction in order to detect a cut-in obstacle is required.

In the case of a cut-in obstacle, since it must enter the vehicle from the rear or side of the vehicle, it must be overtaken, which means that the cut-in obstacle must have a positive velocity component in the front direction of the vehicle.

In addition, if a virtual scan line is formed in the vehicle front direction in order to detect the sign of the vehicle front direction velocity component of the object, the motion in which the vehicle front direction speed component is positive moves forward one-dimensionally on the virtual scan line. appear (b)

Conversely, in the case of an object stationary on the ground, it appears to be pushed back on the virtual scanline. (a)

Also, even when the vehicle rotates, this phenomenon is preserved in the virtual scan line in which the following Equation (1) holds.

Equation 1

d < V/w

In this case, d denotes the distance between the virtual scan line and the vehicle's camera, V denotes the speed of the vehicle, and w denotes the angular velocity of the vehicle.

Even in the case of a low-speed U-turn, since the value of the right side of Equation 1 is about 10 m, it includes all virtual scan lines of sufficient distance.

In order to transform a virtual scanline existing on the ground coordinate system into an adaptive scanline in an image, a projective transformation must be used.

Here, since occlusion exists, points on the virtual scanline cannot be equally observed in the converted adaptive scanline, but as shown in FIG. 6, the point d observed in the adaptive scanline ) and the point (c) on the virtual scan line are points on the same object, since the direction of motion is the same, information about the motion direction can be obtained in the same way.

Accordingly, the present invention detects an obstacle overtaking by detecting an object moving forward in the adaptive scanline.

Then, according to the present invention, if an obstacle is detected in the scanline of a specific area, it is determined as a cut-in obstacle, otherwise it is determined as an obstacle to be simply overtaken.

FIG. 8 is a view for explaining a process of generating a characteristic map, FIG. 9 is a view showing a scan line set in an original image, and FIG. 10 is a view for explaining a process of generating a spatiotemporal pattern map.

8 and 9, the present invention simplifies data to be configured in a space-time pattern map (STP Map) through edge detectors such as Sobel and SUSAN.

As shown in FIG. 8 , according to the present invention, a feature map (edge map) may be generated based on an edge instead of a pixel value of an original image, and the generated feature map may be post-processed.

Here, the reason for the post-processing is to remove the data noise of the feature map and to reveal the necessary components well.

Also, according to the present invention, smoothing may be applied after binarizing the post-processed feature map.

That is, the characteristic map generator of the present invention may generate the characteristic map based on the edge of the acquired image.

In addition, the feature map generator of the present invention may remove noise from the feature map when the feature map is generated based on the edge of the acquired image.

Meanwhile, as shown in FIG. 9 , a plurality of scan lines may be set in the original image, and the length of the scan lines for each position of the original image may vary.

Therefore, in the present invention, as shown in FIG. 10 , in order to generate a spatiotemporal pattern map of a certain size, interpolation is performed to construct a row image of a scanline 1 having a constant length, and the next frame Similarly, the scan line (1) at the same position is configured as a row image with a constant length and stacked on top of the previous row image is repeated.

The image in which the scanlines constructed in this way are stacked is called a space-time pattern map.

Similarly, the same operation is repeated for the other selected scanlines (2, 3, 4).

In the present invention, as an example, since there are four selected scan lines 1, 2, 3, and 4, about four spatiotemporal pattern maps may exist.

That is, the spatiotemporal pattern map generation unit of the present invention generates a row image corresponding to each scanline for each frame, accumulates and stores one or more frames of the generated row images of each scanline, and based on the stored row images, each A space-time pattern map in which row images of scanlines are accumulated can be generated.

Here, the spatiotemporal pattern map includes a plurality of sub-units, and the plurality of sub-units may include a row image in which at least three frames are accumulated, but is not limited thereto.

11 is a view showing the accumulated spatiotemporal pattern map and its sub-units, FIG. 12 is a view showing the inclination of the trend line of the spatiotemporal pattern map, FIG. 13 is a view showing the obstacle movement speed according to the inclination of the trend line, and FIG. It is a diagram for explaining a process of calculating a small gradient.

As shown in FIG. 11, when the cumulative size of row images is greater than 3, that is, when a spatiotemporal pattern map in which at least 3 frames of scanlines are accumulated, a spatiotemporal pattern map composed of only three recent row images? A sub-unit 620 of the spatio-temporal pattern map 610 using the map as the minimum unit of obstacle detection is generated.

This is to prevent the memory capacity from being enlarged more than necessary due to the accumulation of spatiotemporal patterns, and to recognize obstacles in real time using only the time of the previous two frames including the current scan line.

And, as shown in FIG. 12 , when the sub-unit 620 of the spatiotemporal pattern map 610 is configured, the slope of the trend line 630 that appears the brightest in this is obtained.

A general data fitting method may be used, but as in the present invention, to analyze a spatiotemporal pattern of a short time interval, it is appropriate to implement the Recursive & Feedback method, and when there are a plurality of trend lines 630, the most constant A representative trend line 630 and one slope thereof are selected based on the rule.

As shown in FIG. 12 , a phenomenon in which the background is pushed back on the screen and the cut-in obstacle moves relatively forward appears as an inclination on the scan line.

13, if the slope of the trend line (Estimated line) 630 is positive 730, it is determined as an obstacle faster than the own vehicle, and if it is negative 710, it is determined as a slow obstacle. , it can be judged that the speed is similar to that of the own car.

One of the features of the present invention is that accurate calculation of the slope is not included because it is based on short-time spatiotemporal pattern data.

This is because the present invention does not have any meaning in the exact value of the inclination, but uses only the sign of the inclination to detect the obstacle.

Therefore, it can be said that the more important part is to find a reasonable line segment for which the slope is to be obtained on the spatio-temporal pattern.

As shown in FIG. 14 , in the case of erroneous detection as a moving obstacle, the erroneous detection point will move differently from the point on the moving obstacle (because it is not an actual obstacle), if the relationship between the before and after processing is used. ), so the possibility that it will not be re-detected increases, nevertheless, the point detected as a moving obstacle has reliability, so it takes advantage of the fact that there is a high probability that it is a moving obstacle in the next interpolation.

15 to 17 are views showing a spatiotemporal pattern map according to an obstacle detection result.

As shown in FIGS. 15 to 17 , the present invention is a technology for analyzing changes in obstacles using continuous images obtained with a front camera and detecting obstacles showing a specific change, faster than the own vehicle using a spatiotemporal pattern map. Moving obstacles can be detected.

In addition, since the present invention detects an obstacle moving faster than the own vehicle, the calculation speed is faster than the existing methods of analyzing the original driving image.

In addition, according to the present invention, since the temporal change and the spatial change of the driving image can be used simultaneously, the processing time for detecting the obstacle is fast.

In addition, since the present invention detects an obstacle based on a spatio-temporal pattern map rather than a method of detecting an obstacle based on an original image, it is not affected by camera performance and image acquisition environment such as lens distortion.

In addition, the present invention proposes a method for adaptively setting a scan line in the front direction of a camera based on driving information of a vehicle in a driving image, and a space-time pattern corresponding to a very short time within one frame for real-time processing In order to compensate for the shortcomings of real-time processing, we propose a method for detecting obstacles by analyzing the currently generated spatiotemporal pattern in a feedback method using previous results.

The present invention detects moving obstacles by calculating the gradient of the pattern on the spatio-temporal pattern, so the calculation speed is faster than the existing methods of analyzing image information at all positions of the original image, and the performance is stable without depending on the camera such as lens distortion. It has the advantage of being applicable regardless of wide angle/narrow angle.

18 is a flowchart illustrating a method for detecting an obstacle in a vehicle according to the present invention.

18, the present invention acquires an image. (S10)

And, the present invention generates a characteristic map of the acquired image. (S20)

Next, the present invention sets at least one scanline corresponding to the generated characteristic map. (S30)

Next, the present invention generates a spatiotemporal pattern map corresponding to the set scan line (S40).

And, in the present invention, an obstacle is detected by analyzing the generated space-time pattern map. (S50)

Next, the present invention controls the operation of the vehicle according to the detected obstacle. (S60)

Next, according to the present invention, it is checked whether an obstacle detection end request is received ( S70 ), and when there is a detection termination request, the obstacle detection is terminated.

The present invention described above can be implemented as computer-readable code on a medium in which a program is recorded. The computer-readable medium includes all types of recording devices in which data readable by a computer system is stored. Examples of computer-readable media include Hard Disk Drive (HDD), Solid State Disk (SSD), Silicon Disk Drive (SDD), ROM, RAM, CD-ROM, magnetic tape, floppy disk, optical data storage device, etc. There is also a carrier wave (eg, transmission over the Internet) that is implemented in the form of.

Accordingly, the above detailed description should not be construed as restrictive in all respects but as exemplary. The scope of the present invention should be determined by a reasonable interpretation of the appended claims, and all modifications within the equivalent scope of the present invention are included in the scope of the present invention.

100: image acquisition unit
200: characteristic map generation unit
300: scan line setting unit
400: space-time pattern map generation unit
500: obstacle detection unit
600: control unit

Claims (20)

an image acquisition unit for acquiring an image;
a feature map generator generating a feature map of the acquired image;
a scan line setting unit for setting at least one scan line corresponding to the generated characteristic map;
a spatiotemporal pattern map generating unit that generates a spatiotemporal pattern map corresponding to the set scan line;
an obstacle detection unit for detecting an obstacle by analyzing the generated space-time pattern map; and,
and a controller for controlling the operation of the vehicle according to the detected obstacle. According to claim 1,
The characteristic map generator is configured to generate a characteristic map based on an edge of the acquired image. 3. The method of claim 2,
and the characteristic map generator removes noise from the characteristic map when the characteristic map is generated based on the edge of the acquired image. According to claim 1,
The scan line setting unit,
a front direction vector calculation unit for calculating a front direction vector of the camera from the characteristic map;
a first scanline generator that generates a first scanline in a ground coordinate system based on the calculated front direction vector;
a second scanline generator for generating a second scanline in an image coordinate system based on the generated first scanline;
and a scan line correcting unit for correcting an error of the generated second scan line. 5. The method of claim 4,
The front direction vector calculation unit calculates a ground coordinate corresponding to the main point of the camera and a ground coordinate of the camera, and calculates a front direction vector based on a difference between the calculated two coordinates. detection device. 5. The method of claim 4,
The first scanline generating unit obtains a straight line orthogonal to the calculated front direction vector, divides a selected region based on the obtained straight line into a plurality of line segments, sets a midpoint of the divided line segment as a reference point, and An obstacle detecting apparatus for a vehicle, characterized in that it generates a first scan line parallel to the front direction vector while passing through a reference point. 7. The method of claim 6,
The area selected based on the obtained straight line is an area for detecting an obstacle. 7. The method of claim 6,
The first scan line may include the front direction vector and a corresponding reference point. 5. The method of claim 4,
A plurality of first scan lines in the ground coordinate system,
The plurality of first scan lines are an obstacle detecting apparatus for a vehicle, characterized in that parallel to each other. 5. The method of claim 4,
The second scanline generating unit obtains a horizon based on the generated first scanline, obtains a vanishing point based on a straight line converted from the first scanline into image coordinates, and then obtains a reference point on the converted straight line and the An obstacle detecting apparatus for a vehicle, characterized in that generating a second scan line passing through the vanishing point. 11. The method of claim 10,
The second scan line generator, when obtaining the horizon, sets a straight line passing through the vanishing point as a horizon when the first scan lines gather at one vanishing point in the image coordinate system. . 11. The method of claim 10,
The second scan line generator, when obtaining the vanishing point, converts the first scan line in the ground coordinate system into a straight line in the image coordinate system, and sets an intersection point between the converted straight line and the horizon as a vanishing point. obstacle detection device. 11. The method of claim 10,
and the scan line correcting unit corrects an error of the second scan line by removing a portion of the second scan line located near the horizon. According to claim 1,
The spatiotemporal pattern map generation unit,
a first generation unit generating row images corresponding to the respective scan lines for each frame;
a storage unit for accumulating and storing one or more frames of the generated row images of each scan line;
and a second generator for generating a spatiotemporal pattern map in which row images of each scan line are accumulated, based on the row images stored in the storage unit. 15. The method of claim 14,
The spatiotemporal pattern map includes a plurality of sub-units,
The plurality of sub-units, the obstacle detecting apparatus for a vehicle, characterized in that it includes a row image in which at least three frames or more are accumulated. According to claim 1,
The obstacle detecting unit is configured to analyze the generated space-time pattern map to obtain a gradient of a trend line included in the space-time pattern map, and to detect the obstacle based on the gradient. 17. The method of claim 16,
The obstacle detecting unit, when the inclination is positive, recognizes an obstacle faster than the own vehicle, and when the inclination is negative, recognizes it as an obstacle slower than the own vehicle. 17. The method of claim 16,
The obstacle detecting unit, when there are a plurality of trend lines included in the spatiotemporal pattern map, selects a brightest trend line as a representative trend line, and obtains a slope of the selected representative trend line. 19. The method of claim 18,
and the obstacle detecting unit selects one representative trend line based on a most constant rule when there are a plurality of representative trend lines, and obtains a slope of the selected representative trend line. acquiring an image;
generating a characteristic map of the acquired image;
setting at least one scanline corresponding to the generated characteristic map;
generating a spatiotemporal pattern map corresponding to the set scanline;
detecting an obstacle by analyzing the generated space-time pattern map; and,
and controlling the operation of the vehicle according to the detected obstacle.

Images