JPH07220194A - Road environment recognition device - Google Patents

Abstract

(57) [Abstract] [Object] The present invention relates to a road environment recognition device for controlling autonomous running of a vehicle, and accurately recognizes the position of a sign line even if a shadow, a sign character, or a vehicle ahead is present on the road. In addition, it is an object of the present invention to provide a device that can separately recognize a plurality of front vehicles. [Structure] In order to recognize a marker line, only an edge (outline of an object) is extracted from a front image obtained by an image pickup means attached to the front of the vehicle, and the edge (or outline) of the vehicle is obliquely seen to the left or right of the image. The edge that appears to some extent continuously is recognized as the left (or right) marker line in the vehicle lane. Further, in order to identify the position of the vehicle in front, a plurality of image pickup means in the front of the traveling direction are provided, and the rectangles having substantially the same graphical features among the rectangles extracted on each image are associated with each other, and the rectangles are associated with each other. Recognize as a vehicle ahead. Further, the distance is calculated from the position shift (parallax) between the plurality of images for the associated rectangle using the principle of triangulation.

Description

Detailed Description of the Invention

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a road environment recognition apparatus having an image processing system for recognizing a road environment, and more specifically, it is installed in front of a vehicle in an autonomously traveling vehicle and From the image from the TV camera that captures the front of the traveling direction, the marker lines provided on the road surface are recognized from the area near the vehicle to the area far away, and at the same time, the vehicle / obstacle traveling in front of the vehicle is identified. The present invention relates to a road environment recognition device that can be recognized.

[0002]

2. Description of the Related Art In recent years, a self-sustained travel control device has been developed which recognizes a road environment by an image processing portion and controls self-sustained travel of a vehicle based on the recognition result. This self-sustained travel control device is installed in a vehicle running in the traveling lane. For example, if the self-vehicle departs from the lane, the driver is automatically warned, and based on the information, the vehicle Control is realized. Realization of such a device is extremely important because it can reduce the driver's driving burden and prevent accidents.

More specifically, from the image from the TV camera installed in front of the vehicle and capturing the front of the traveling direction of the vehicle, the marker line provided on the road surface is distant from the area near the vehicle. It recognizes vehicles and obstacles traveling in front of the vehicle at the same time, and in the steady running state, the marker line such as the white line visible on the left side is used as a guideline along which the vehicle travels in front of the vehicle. The distance to the other vehicle is calculated and the own vehicle is steered so as not to collide with the preceding vehicle.

As a system for recognizing the traveling lane of the own vehicle by using the image processing technology from the image of the front of the own vehicle and simultaneously recognizing another vehicle traveling in front of the own vehicle, the applicant of the present application has "Road environment recognition device" (Japanese Patent Application No. 5-15
No. 5138). This Japanese Patent Application No. 5-155138
The number of positive and negative vertical edges existing in the image is extracted, and the positive and negative edges extracted using the white line width and edge continuity are verified to be the edges of the white line, and they extend mainly in the vertical direction. Marker line recognition is performed by extracting a pair of positive and negative vertical edges as marker line candidates. In front vehicle recognition, within the vehicle lane extracted as a result of white line recognition,
The area where many horizontal edges appear was extracted as one vehicle in front.

As a marking line recognition method other than the above-mentioned prior application,
For example, "Autonomous vehicle direction control device" (Japanese Patent Laid-Open No. 3-22
No. 5513), and the publication of the patent application discloses that an image is used to search for a white line in an intermediate region or a region close to the own vehicle by utilizing the recognition result of a marker line in a region farther than the own vehicle. There is.

[0006]

However, each of the above known examples has the following problems. First, Japanese Patent Application No. 5-
In marker line recognition in No. 155138, by utilizing local properties such as white line width and edge continuity, it is verified whether the extracted positive and negative edges are white line edges, and a pair of vertically extending positive and negative lines is used. Since the vertical edge is used as a white line candidate, there is a possibility of erroneous recognition when, for example, a white line smoothly extends in a direction different from the main line in a branch road. Also,
In the preceding vehicle recognition of the conventional example, a region in which many horizontal edges appear in the own vehicle lane is extracted as one preceding vehicle, so that, for example, the preceding vehicle A running just before the own vehicle is on the left side of the lane. When a car B traveling in front of A is traveling on the right side of the lane, A and B are shown as overlapping in the captured image. In this case, The presence of one large forward vehicle could lead to false positives.

Further, in Japanese Patent Laid-Open No. 3-225513, a white line in an intermediate area or an area close to the own vehicle is searched for by utilizing the recognition result of a white line in a distant area, but the resolution in distant recognition is not sufficient. Therefore, there was a possibility that an error would occur. The present invention has been made in order to solve such a problem, and to easily and accurately recognize the position of the marking line even if there is a shadow or a sign character on the road, or a vehicle ahead,
An object of the present invention is to provide a road environment recognition device that can separately recognize two or more front vehicles even if they overlap each other on an image.

[0008]

In order to recognize the position of a marker line, the road environment recognition apparatus of the present invention creates an edge image in which only the contour of an object is extracted from the front image obtained by the image pickup means, and the image is obtained. At the top left of your vehicle (or diagonally right)
In the region where the image of the front is captured, the edges that appear to some extent are extracted, and they are extended far in the image,
Recognize as the left marking line (or right marking line) in your own lane.

Prior to recognition, parameters (depression angle, angle of view, resolution, etc.) peculiar to the image pickup means (camera, etc.)
And the range of possible infinity points is calculated based on knowledge information (curvature etc.) related to the road that is passing, and if the extension line does not pass through this range, it is determined that it was erroneously recognized. . Furthermore, parameters specific to the traveling lane width of the road and the imaging means (camera, etc.) (depression angle, angle of view, resolution, etc.)
Configured to estimate the right marker line (or left marker line) of the vehicle's driving lane from the infinity point, which is an extension of the left marker line (or right marker line), using the knowledge of the infinity point calculated from did.

Further, the road environment recognition apparatus of the present invention is provided with a plurality of image pickup means for picking up an image in front of the traveling direction in order to recognize the position of the vehicle ahead, and from a plurality of front images obtained from the plurality of image pickup means. , The rectangles extracted on the respective images obtained from the pair of image pickup devices are obtained as a result of picking up the same object with rectangles having the same area, horizontal and vertical line length and other graphical features. Assuming that the rectangles are rectangles, the positions of the vehicles in front are recognized by associating the rectangles with each other, and the principle of triangulation is used from the positional deviation (parallax) of the associated rectangles on the image. Thus, the distance to the object in front is calculated.

In the above-described rectangle extraction processing, horizontal and vertical edges in the vehicle lane are extracted separately, and the respective projections are calculated to calculate the horizontal and vertical lines in which the accumulated projection value exceeds a certain level in the image. We left it as a candidate for a line with a horizontal or vertical edge, and if the rectangle was composed of horizontal and vertical edges on that line, that part was extracted as a rectangle.

Further, among the rectangles, the one within the range of the two previously recognized marking lines is extracted to extract the preceding vehicle in the own vehicle lane.

[0013]

The operation and effect of the road environment recognition apparatus of the present invention described above will be described below with reference to FIG. This Figure 7
Is an example of an image obtained by capturing an image of the front from an image capturing unit (TV camera or the like) installed in front of the vehicle traveling in the traveling lane. In addition, the technology to be applied will be described on the assumption that the left-hand traffic of a vehicle is adopted as in Japan in the following action section and the embodiment section. Therefore, if it is assumed that the vehicle is traveling on the right side, the left and right of the following description may be reversed.

As is apparent from FIG. 7, the image taken by the camera is a three-dimensional scene projected on a two-dimensional plane with the focus of the camera as the center, and the line parallel to the optical axis of the camera is on the image. And gather at one point. This point is called the point at infinity.
In the figure, marking lines such as white lines that define the driving lane and the passing lane are the radiation that converges to the point at infinity on the image. This point at infinity can be calculated from camera-specific parameters such as the angle of view, depression angle, and resolution of the camera. Therefore, the left side of the running line on the image is difficult to be hidden by other running vehicles and obstacles, so a part of the sign line on the left side of this running line is extracted, and it is extended to the point at infinity and the sign on the left side is extracted. Recognize lines. Furthermore, since there are generally no sharp curves on a motorway, radiation that is shifted by the lane width from the left marker line extending to the point at infinity is drawn on the road surface to estimate the right marker line. According to the present invention, the vehicle lane can be correctly specified by the method described above even if the marker line is hidden in the distance or the left marker line is a broken line. It is possible to improve the recognition accuracy by calculating the possible range of the point at infinity based on the curvature of the road and the camera parameters in advance and checking whether the extension line passes through this range. it can.

On the other hand, in the image of the back of the vehicle in front in the image of the road ahead, there are many horizontal and vertical edges such as vehicle body, license plate, rear window, bumper, tail lamp, brake lamp, etc. There are many. (However, the resolution is poor in the image taken by a normal TV camera, and the characters such as the vehicle number existing on the license plate cannot be extracted and recognized.) Therefore, the horizontal and vertical edges in the vehicle lane are not recognized. Extract, and the part surrounded by them is extracted as a rectangle,
It is a feature of the vehicle ahead. Here, vertical edges such as character signs and shadows displayed on the road surface are less likely to be extracted by vertical lines due to the same perspective effect (convergence to the point at infinity) as white lines.

As described above, for example, a rectangular portion is extracted from the images picked up by the left and right binocular cameras, the rectangles having the same graphic feature are associated with each other, and the position shift (parallax) between the binocular images is made. Then, the distance to the vehicle ahead is calculated by the triangulation principle. Further, when a plurality of combinations can be considered in associating the rectangular portions, the one with a larger parallax can be adopted to extract a closer forward vehicle.

Further, by extracting the rectangle within the range between the two previously recognized marking lines, it is possible to extract the preceding vehicle traveling in the own vehicle lane. Therefore, in the present invention, the vehicle in front of the own vehicle runs on the left side in the lane, and the vehicle in front of it travels on the right side in the own vehicle lane. Even in such a case, the distance between the two vehicles can be extracted separately.

[0018]

DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS An embodiment of the road environment recognition device of the present invention will be described below. FIG. 1 is a block diagram showing the overall configuration of this embodiment. In the figure, 1 is an imaging means. In the present invention, the respective image pickup means may be installed vertically or horizontally in parallel with each other in the vehicle, but in the present embodiment, as shown in FIG. 1, the left image pickup means 1L is on the front left side of the vehicle and the right image pickup means is on the right side. Assume that 1R is installed.

Reference numerals 2L and 2R denote vehicle lane recognition sections for recognizing the vehicle lane from the image obtained by the image pickup means 1.
L and 3R are images obtained by the image pickup means 1L and 1R and the vehicle lane recognition unit 2L. It is a rectangle extraction unit that extracts a rectangle based on the information obtained in 2R. 4 is a rectangle extraction unit 3L, 3R
It is a rectangular matching / distance calculating unit that collates rectangular information obtained from each of the left image and the right image, which is output from the vehicle, recognizes a vehicle ahead, and calculates the distance of the vehicle ahead.

The operation of each unit shown in FIG. 1 will be described below. In the following description, it is assumed that the vehicle is traveling on the left side, and the marker line is a white line as an example. Further, reference numerals of the drawings are shown in parentheses for each means and processing. When the claims of the present embodiment are different from the claims, the corresponding names in the claims are described together. 1) Imaging means (1L, 1R) This is mounted on the left and right of the front of the vehicle as described above, and images the front of the traveling direction of the vehicle. As shown in FIG. 4, the left and right cameras are attached so as to incline downward from the horizon by a depression angle Θ, so that the range (field of view) of the road surface that can be imaged is a quadrangle A _R B _R C _R D
Become _R.

Further, camera parameters are set based on the camera used. These camera parameters include
There are the following. f: focal length Θ: depression angle R _x : x (horizontal) direction resolution R _y : y (vertical) direction resolution Φ _x : x direction angle of view Φ _y : y direction angle of view Further, in FIG. , Z) and the corresponding screen coordinates (x, y) using the above camera parameters, the following relationship is established.

(1) Conversion from screen coordinates to ground coordinates

[0023]

[Equation 1]

(2) Conversion from ground coordinates to screen coordinates

[0025]

[Equation 2]

2) Own vehicle lane recognition section (2L, 2R) FIG. 2 shows a detailed flow of the own vehicle lane recognition section. ◇ Edge extraction processing (21: edge image creating means) First, vertical edge extraction is performed on the image of the front. In this method, an edge detection filter of 3 × 3 or the like as shown in FIG. 6A is scanned over the entire screen for extraction. Since such an edge detection method is already known, a detailed description will be omitted. ◇ Left white line verification processing near the own vehicle (22: marker line recognition means) First, after removing edges with a certain length or less as noise from the extracted edge image, even if there is a vehicle ahead, Even if the left white line of the driving lane exists, the edge that extends toward the existence range of the point at infinity is detected in the lower left part of the image, and the edge is verified using knowledge about the white line width.

Specifically, when no such edge can be extracted, the left white line obtained as a result of processing the image one frame before is output, and the left white line is extracted for the current frame. Remember what you couldn't do. In this way, when the frames for which the left white line could not be extracted continue for several frames in succession, the white line does not exist and the processing is terminated. When a plurality of lines are found, an edge pair having a longer edge and satisfying the white line width limit is regarded as the edges on both sides of the left white line and is output (matched with the processing result of the image one frame before). If only one can be found, it is compared with the processing result of the image one frame before, and if the edge positions are not extremely separated,
A white line including that edge as one edge is output as a left white line. ◇ Extending the left white line to a distance (23: extension line creating means) Based on the extraction result of the left white line in the vicinity of the own vehicle, linear interpolation is performed up to the point at infinity as shown in FIG. 7 to determine the left white line. To do.

The point at infinity can be calculated in advance by using the above-mentioned camera parameter. That is, the point at infinity is a point that exists on the road surface at infinity (Z → ∞) from the own vehicle, which is expressed as Z → ∞ in the above equation, and y = {(R _y / 2) / tan (Φ _y / 2)} tan Θ ──── It exists on the horizontal line on the image. (This is called the line at infinity) The point at infinity is the intersection of the expression obtained by linearly approximating the left white line near the vehicle extracted earlier.

When the vehicle is running in parallel with the white line, the infinity point of the left white line on the image is located at the center of the screen. In FIG. 7, since the point directly under the camera is set to the origin 0 of the X coordinate, the point at infinity is on the Z axis. However, if the road surface is curved or the vehicle is not running in parallel with the white line, the position of the point at infinity will be displaced in the X-axis direction.

However, the maximum value of the curvature of the curve is specified on the basis of the speed limit, etc. on a road for automobiles, etc., and when the vehicle is traveling normally, the angle between the vehicle and the white line becomes extremely large. is not. Therefore, if the range where the point of infinity can exist on the line of infinity is set and the point of infinity obtained by extending the extracted left white line on the image is not within this range, the extraction of the left white line fails. (Mistaken) Judge as having been made. A method for obtaining knowledge about this range will be described later. ◇ Right white line estimation process (24: sample line estimation means) Finally, the position of the right white line from the point at infinity to the vicinity of the vehicle is estimated on the image using the knowledge about the lane width, and the vehicle lane area is extracted. .

Therefore, by substituting the value of the lane width of the road currently running on the horizon (for example, y = y ₀ ) near the vehicle on the image into the formula,
The difference Δx ₀ between the x coordinate values on the image is calculated. When the equation of the left white line on the image is calculated and the point at infinity is calculated, the x coordinate of the left white line at y = y ₀ is obtained, and this is shifted to the right by Δx _0, so that y =
Estimate the x coordinate (x ₀ ) on y ₀ . The position of the right white line is estimated by connecting (x ₀ , y ₀ ) thus obtained and the point at infinity.

Various methods are conceivable for obtaining the knowledge about the curvature of the curve of the road and the knowledge about the lane width of the road, which are not particularly limited in the present invention, but some examples are given. For example: -Enter settings for each road that the user travels.・ Prepare a database of lane widths for each motorway for each device, and extract and use knowledge according to the roads that pass. -A wireless receiving system is added to the control system including this device, and the data related to the lane width and curvature transmitted from the wireless device installed on the shoulder of the road is received and used. Such an embodiment is conceivable.

The lane width (lane width) and curvature of a road are regulated by the Road Structure Ordinance (Government Ordinance established under the Road Law), and the local government that manages the road decides based on this. It is what Therefore,
Since the number of types of standards is finite, it suffices if the correspondence between passing roads and the standards is clear, and the amount of data will not be enormous even if it is made into a database. In addition, the user may input the designation of the road on which the traffic passes, but (known)
It may also be obtained from the navigation system, again for which various embodiments are conceivable. 3) Rectangle Extracting Means FIG. 3 shows a detailed flowchart of the operation of the rectangle extracting units 3L and 3R. ◇ Horizontal / vertical edge extraction processing (311, 312) First, the horizontal and vertical edges are extracted separately. The edge detection filter shown in FIG. 6A is used for the vertical edge, and the edge detection filter shown in FIG. 6B is used for the horizontal edge. This horizontal edge detection is also a well-known technique like the vertical edge detection described above.

As the vertical edge image, the one obtained by the vehicle lane recognition units 2L and 2R may be used. Projection calculation process (321, 322: projection calculation means) Projection in the horizontal direction (the number of pixels in the horizontal direction) for the detected horizontal edge, and projection in the vertical direction for the vertical edge (the number of pixels in the vertical direction) ) Is calculated separately. -Horizontal / vertical line mask calculation (331, 332) The following processing is performed using the projection values in the horizontal and vertical directions. The horizontal processing will be described below with reference to FIG. 8, but the same applies to the vertical processing.

A line having the maximum projection value among the projection values in the horizontal direction and one line before and after the line are set as a horizontal line extraction mask as a line having the strongest edge in the original image. Here, several lines before and after the 3 lines for which the horizontal line mask is set (for example, 3 lines) are lines before and after a large number of strong edges exist, and the projection value is set to 0 because it is unlikely that there are edges, and the mask is set. Not set as. Then, remove the extracted mask line,
The same process is repeated for all horizontal lines of the image. Extraction of horizontal / vertical line candidates (314, 324) Next, the extracted horizontal / vertical line mask is scanned on the edge image, and if an edge exists in the mask region, the center of the mask region (3 lines) The pixels of the line are set to 1 and horizontal / vertical line candidates are extracted. ◇ Rectangle Search Process (315, 325) In this process, a process of extracting a rectangle having a large area is performed in order to extract a candidate for a vehicle ahead.

Therefore, for each line of the horizontal lines extracted by the horizontal / vertical line candidate extraction unit, the line having the longest length from the intersecting point is searched for and extracted. For example, as shown in FIG. 9, for a certain horizontal line A, a vertical line B that intersects the horizontal line A from the right end is searched, and then a horizontal line C that intersects the vertical line B from the lower end to the vertical line B is searched for. A rectangle ABCD is extracted by executing a process of searching a vertical line D which intersects the horizontal line C from the left end with respect to the horizontal line C and also intersects with the original vertical line A. Further, in order to extract the rectangle having the largest area including the horizontal line A, another vertical line (for example, the vertical line B ′) inside (left side of) the vertical line B is searched, and the same process as above is performed. Another rectangle (for example, rectangle AB'C'D) is extracted by. By comparing the areas of the plurality of rectangles thus obtained, the rectangle having the largest area is extracted and adopted as a candidate for the front vehicle. 4) Rectangle collation and distance calculation unit (4) The rectangle collation unit is located at the same position with respect to the rectangular regions extracted in the left and right images (for example), and has the same graphical features (area, rectangle configuration). Rectangles having vertical and horizontal line lengths).

A method of this correspondence will be described with reference to FIG. In FIG. 10, D: distance between left and right cameras L: distance between camera and target object Ob: position of target object C ₁ : position of left camera (center of lens) C ₂ : position of right camera (center of lens) O ₁ : Intersection between the lens optical axis of the left camera and the film plane O ₂ : Intersection of the lens optical axis of the right camera and the film plane f: Focal length of the camera P ₁ : Point where the object appears on the film plane of the left camera. Coordinates (x ₁ , y ₁ )
Has O ₁ as the origin. P _2: object objects appear point to the film surface of the right camera. Coordinates (x ₂ , y ₂ )
Has O ₂ as the origin. Is.

As is clear from FIG. 10, the closer the distance to the object is, the larger the parallax on the image is. Therefore, the maximum parallax is previously obtained from the parameters of the left and right cameras, and the upper left point ((x ₃ ,
y ₃ )) and the lower right point (denoted (x ₄ , y ₄ )) on the right image from (x ₃ , y ₃ ), (x ₄ , y ₄ ), to the left in the maximum parallax range. Up to, and the degree of agreement is high (eg ± α (α = 1 or
If there is a rectangle (matching with the pixel shift in 2)), calculate the difference between the positions of both rectangles and use it as the parallax.

When there are a plurality of rectangles having the same graphic features in the maximum search range on the right image, the one having the larger parallax is associated. This is because when there are a plurality of rectangles formed by images of a plurality of vehicles, the one in front of the preceding vehicle can be recognized by adopting one with a larger parallax, that is, one with a smaller distance. The rectangle calculated in this way is regarded as a front vehicle, and the distance is calculated. In Fig. 10, △ ObC ₁ C ₂ and △
ObP ₁ P ₂ are similar, and the following relationships hold. (Here, in a normal camera, the film surface is on the side opposite to Ob with C ₁ as the center, but in this model, the film surface is in the same direction as the object for the sake of simplicity.) (Lf ): (D−x ₁ + x ₂ ) = L: D ・ ・ ・ By transforming this expression, L = Df / (x ₁ -x ₂ ) ・ ・ ・ ・ where (x ₁ -x ₂ ) is the film surface (that is, the image surface) of the left and right cameras, which is the difference in position when the same object is imaged, and is the parallax. Therefore, the distance from the camera to the object can be calculated from the formula based on the parallax.

If there is no matching rectangle, the length of the horizontal line (vertical line if the cameras are installed above and below) of the rectangle on the image where the rectangle is not extracted (right image). Find a pair of two vertical edges (vertical edges if the camera is installed above and below) that exist at intervals corresponding to, calculate the parallax by associating these vertical edge pairs, and calculate the distance by the above method. calculate.

The rectangles thus extracted and associated with each other are checked by checking whether they exist within the range between the left and right marking lines obtained by the vehicle lane recognition units 2L and 2R. In this case, if the vehicle is recognized as a forward vehicle traveling in the own vehicle lane, the vehicle that requires the most attention can be extracted from a plurality of forward vehicles in autonomous driving.

[0042]

As described above, according to the road environment recognition apparatus of the present invention, a shadow, a sign character on the road, and a forward vehicle are present from an image captured by a TV camera mounted on a self-driving vehicle. However, it is possible to easily and accurately recognize the position of the marker line, and at the same time, it is possible to separately recognize even if two or more forward cars overlap each other on the image.

As described above, since it is possible to obtain accurate information about the marking line and the vehicle ahead, by using the road environment recognition device of the present invention, a self-sustained travel control device for steering the vehicle so as not to collide with the vehicle ahead. Can be configured.

[Brief description of drawings]

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

FIG. 2 is a flowchart showing a processing operation of a vehicle lane recognition unit 2 of FIG.

FIG. 3 is a flowchart showing a processing operation of a rectangle extraction unit 3 in FIG.

FIG. 4 is a diagram showing how the imaging means (camera) is mounted on a vehicle.

FIG. 5 is a diagram showing a correspondence relationship between a coordinate system of an imaging screen and a coordinate system of the ground.

FIG. 6 is a diagram showing an example of a vertical edge detection filter and a horizontal edge detection filter used in an embodiment of the present invention.

FIG. 7 is a diagram showing an example of a forward road image obtained by the imaging means 1.

FIG. 8 is a diagram illustrating an example of a horizontal line extraction mask area creation process.

FIG. 9 is a diagram illustrating an example of a rectangle search process.

FIG. 10 is an explanatory diagram of distance measurement.

[Explanation of symbols]

 DESCRIPTION OF SYMBOLS 1 ... Imaging means 2 ... Own vehicle lane recognition part 3 ... Rectangle extraction part 4 ... Rectangle matching / distance calculation part

─────────────────────────────────────────────────── ─── Continuation of the front page (51) Int.Cl. ⁶ Identification code Internal reference number FI technical display location G09B 29/10 A

Claims (11)

[Claims] 1. An image pickup means (1) for picking up an image in front of a traveling direction of a vehicle, a means (21) for creating an edge image in which only a contour line of an object is extracted from an image obtained by the image pickup means, and the edge. In the part corresponding to the road surface near your vehicle in the image,
In addition, an edge portion that is present in a portion corresponding to the road surface on the left side or the right side of the host vehicle and that is continuous for a predetermined length or longer is extracted, and the extracted edge portion is recognized as a marker line on the left side or the right side of the host vehicle. A road environment recognition device comprising a marking line recognition means (22). 2. An infinity point, which is a point where an edge recognized as a marker line by the marker line recognizing means (22) intersects an infinite distance line on an image determined by a parameter relating to the image pickup means (1). The road environment recognition device according to claim 1, further comprising extension line creating means (23). 3. The extension line creating means (23) calculates a possible range of an infinity point of a vehicle lane on the basis of knowledge information about a road on which the vehicle is running and parameters related to the imaging means (1). Every time, it is verified whether the created extension line passes within the range of the point at infinity. If it passes, it is determined that the left or right marker line is correctly recognized, and if it does not pass, the recognition line is erroneously recognized. The road environment recognition device according to claim 2, wherein the road environment recognition device is determined to have been performed. 4. The position of the extension line created by the extension line creating means (23), the parameter relating to the image pickup means (1), and the lane width information of the road on which the vehicle is running are used. Marking line estimating means (24) for estimating the position on the image of the marking lines on the opposite side of the marking lines recognized by the marking line recognition means among the marking lines forming the vehicle lane
The road environment recognition device according to claim 2, further comprising: 5. A plurality of image pickup means (1L, 1R) for picking up an image in front of the traveling direction of the vehicle, and a rectangle extraction means (3L.3R) for extracting a rectangular portion from each image obtained by the plurality of image pickup means. Rectangle matching means (4) for collating each piece of rectangle information extracted by the rectangle extracting means (3L, 3R) to extract rectangles having the same graphical feature, and the extracted rectangles. A road environment recognition device, which recognizes a vehicle as a front vehicle. 6. When the rectangle matching means (4) extracts a plurality of combinations of rectangles having the same graphical feature, the parallaxes of the respective combinations are compared and the rectangle having the largest parallax is compared. The road environment recognition device according to claim 5, wherein the combination is recognized as the closest forward vehicle. 7. With respect to a rectangle on an image extracted by the rectangle matching means (4), a shift in the position of the rectangle on each of the images and a parallax based on the difference in the image pickup positions of the plurality of image pickup means. The road environment recognition device according to claim 5, further comprising a distance calculation unit that calculates a distance to a vehicle in front of the vehicle. 8. The rectangle extracting means (3L, 3R) has projection calculating means (321, 322) for calculating a projection value of only a horizontal edge and a projection value of only a vertical edge, and a projection value of a predetermined value or more. Horizontal line / vertical line candidate extraction means (34) for extracting a line as a line in which a rectangle may exist.
1, 342), the road environment recognition device according to claim 5. 9. The horizontal line / vertical line candidate extracting means (34)
1, 342) is set as a mask area over the entire image from the horizontal or vertical direction with respect to the line with many projection values and the lines with a predetermined width before and after the line, and the portion where the mask and the horizontal / vertical edge overlap. 9. The road environment recognition device according to claim 8, wherein is extracted as a candidate for horizontal and vertical lines forming a rectangle. 10. The horizontal / vertical line candidate extraction unit (341,
342) extracting a plurality of vertical lines intersecting with the horizontal line (71) extracted in 342), first extracting a horizontal line intersecting with the vertical line for each of the extracted vertical lines, and then extracting the horizontal line, Rectangle search means (35) for extracting a vertical line that intersects both the first horizontal line and a plurality of rectangular regions, calculates the area of each rectangular region, and extracts the largest rectangular region. The road environment recognition device according to claim 8, which has. 11. A plurality of image pickup means (1L, 1R) for picking up an image in front of a traveling direction of a vehicle, and means (21) for creating an edge image in which only the contour line of an object is extracted from an image obtained by the image pickup means. And in the portion corresponding to the road surface near the own vehicle in the edge image,
In addition, an edge portion that is present in a portion corresponding to the road surface on the left side or the right side of the host vehicle and that is continuous for a predetermined length or longer is extracted, and the extracted edge portion is recognized as a marker line on the left side or the right side of the host vehicle. The marker line recognition means (22) and an edge recognized as a marker line by the marker line recognition means (22) intersects an infinity line on an image determined by a parameter relating to the imaging means (1). An extension line creating means (23) for extending to an infinite point, a position of the extension line created by the extension line creating means (23), and a parameter relating to the imaging means (1),
Using the information on the lane width of the road on which the vehicle is traveling, the position on the image is estimated for the marker line on the opposite side of the marker line recognized by the marker line recognition means among the marker lines forming the own vehicle lane. Marker line estimating means (24), and the marker line recognizing means (22) and the marker line estimating means (24) in a rectangular portion from each image obtained by the plurality of imaging means (1L, 1R). A road environment recognition device characterized in that the road environment recognition device comprises a rectangular portion extracting means (3L, 3R) for extracting rectangular information existing in a range between the two marker lines obtained by the above.