JP2007179386A - Method and apparatus for recognizing white line - Google Patents

Abstract

<P>PROBLEM TO BE SOLVED: To significantly enhance the accuracy of recognizing white lines by making it possible to normally recognize white lines even in a variety of situations where the failure to recognize them or recognition errors could occur. <P>SOLUTION: A method for recognizing a white line includes differentiating and binarizing a road image obtained by imaging a road surface A to extract candidate edges of the white line (a), determining the candidate edges to extract edges of the white line, and recognizing the position of the white line. At least two threshold values for the binarization process, for cases when the difference in density between the white line (a) and a road other than the white line (a) is great and when the difference is small, are preset. The threshold values are used alternatively when the candidate edges of the white line are extracted. <P>COPYRIGHT: (C)2007,JPO&INPIT

Description

  The present invention relates to a method and apparatus for recognizing a white line position from a road image obtained by photographing a road surface with an in-vehicle camera.

In recent years, a system for recognizing the position of a white line on a road and controlling the vehicle steering to maintain the lane has been proposed. In this technology, a white line recognition device mounted on a vehicle captures a road image ahead of the vehicle by an imaging means such as a CCD camera, and detects a white line by utilizing the high brightness at the white line position. Specifically, for example, an edge (edge point sequence) is extracted by image differentiation processing and binarization processing, and it is determined whether or not the extracted edge is a white line edge. Furthermore, by performing a Hough transformation known as a straight line detection method for the extracted edge,
Determine the position of the white line.

In such a system, how to accurately detect a white line from a captured road image is an important issue. However, there is a problem that the position of the white line cannot be recognized in the following situation.
(1) When the white line is in the shadow, the brightness difference between the white line and the road is small, so the white line edge cannot be extracted and is unrecognized.
(2) If there are road shoulders, grooves, etc. outside the white line, those edges are judged as white line edges, which results in erroneous recognition.
(3) When the white line is a broken line, the white line does not exist in the image depending on the image acquisition timing, and thus it is not recognized.
(4) When a vehicle, a building, or the like is present in the front, those edges are determined as white line edges, resulting in erroneous recognition.

  Patent Document 1 describes a white line recognition technique that improves white line recognition accuracy in a running lane. In this technique, a captured road image is divided into two areas, a short distance area and a long distance area, and a white line is detected in each area as follows.

  First, in a short distance region, a white line is recognized from the edge in the image and determined. Next, based on the white line determined in the short distance area, an edge tracking area including the area on the extension line is set on the long distance area. Then, the edge is tracked within the set edge tracking region, and when the edge is discontinuous, the white line is recognized and confirmed by jumping over the edge and tracking the previous edge. According to this white line recognition technology, it is possible to reduce erroneous recognition of white lines due to the influence of shadows of vehicles ahead, roadside trees, buildings, and the like, which is a problem in white line recognition in a long-distance region of a captured image.

Japanese Patent Application Laid-Open No. 2004-228561 describes a technique that can prevent white line misrecognition due to blurring of the white line itself or the white line image being unclear and improve white line detection accuracy. In this technique, white line candidate points are extracted from a white line image of a road obtained by photographing the front of the vehicle, a straight line that approximates the white line candidate point sequence is obtained and recognized as a white line, and the white line image is blurred. If there is blurring, the white line candidate point sequence extracted from the current white line image is overlaid with the white line candidate point sequence extracted from the white line image of a certain time ago, and the overlapping is performed. A straight line approximating the white line candidate point sequence is obtained and recognized as a white line.
JP-A-9-73545 JP 2001-236506 A

  In the technique of Patent Document 1, although it is possible to reduce the misrecognition of the white line due to the influence of the shadow of the forward vehicle, the roadside tree, the building, etc. in the long distance area of the photographed image, the white line is removed from the edge in the image in the short distance area. When recognizing, if the white line is in the shadow, if there is a road shoulder, a groove, etc. outside the white line, if the white line is a broken line, etc., there still remains a problem of misrecognition or unrecognition. .

  Although the technique of Patent Document 2 can prevent misrecognition due to blurring of the white line itself or the white line image being unclear and improve white line detection accuracy, as exemplified in the above (1) to (4). If the white line is in the shadow, the shoulders, grooves, etc. exist outside the white line, the white line is a broken line, if there are vehicles, buildings, etc. in front, etc. The problem remains.

  Therefore, it is an object of the present invention to provide a technique capable of normally recognizing a white line even in an unrecognized or misrecognized situation as described above, thereby significantly improving white line recognition accuracy. And

The present invention employs the following means in order to solve the above problems.
The present invention is a white line recognition method for extracting a white line candidate edge by differential processing and binarization processing of a road image obtained by photographing a road surface, extracting a white line edge by determining the candidate edge, and recognizing a white line position. Thus, at least two threshold values for binarization processing are set, and these threshold values are switched and used.

  According to the white line recognition method of the present invention, at least two threshold values for binarization processing are set, and a method of switching and using these threshold values is employed. Compared to the case of using, for example, when the contrast between the white line and the road is large or small, the white line can be normally recognized even in the unrecognized and erroneous recognition situations that are likely to occur. For example, even when the white line is in the shadow and the contrast is small, the white line can be recognized normally if the threshold is small. Thereby, the white line recognition accuracy can be significantly improved.

  In the present invention, it is desirable to use the two threshold values by alternately switching in accordance with a photographing interval for obtaining a road image. In this way, it is possible to further improve the accuracy of normally recognizing a white line even in a situation where unrecognized and erroneous recognition is likely to occur when the contrast between the white line and the road is large or small.

  In the present invention, in a method of extracting a white line candidate edge by differential processing and binarization processing of a road image obtained by photographing a road surface, extracting a white line edge by candidate edge determination, and recognizing the position of the white line, Of the plurality of road images acquired in series, when the white line position can be recognized from the previously acquired road image, it is desirable to set the scanning range so that the recognized white line position is at the center of the scanning range. In this way, the scanning range can be shortened so that the arithmetic processing is accelerated, and erroneous recognition due to a shoulder, a groove, etc. around the white line can be prevented.

  In the present invention, in a method of extracting a white line candidate edge by differential processing and binarization processing of a road image obtained by photographing a road surface, extracting a white line edge by candidate edge determination, and recognizing the position of the white line, Of the plurality of road images acquired in series, it is desirable to shorten the scanning range when the white line position can be recognized from the previously acquired road image, and to increase the scanning range when the white line position cannot be recognized. In this way, it is possible to prevent the white line from being erroneously recognized or unrecognized by the road shoulder, groove or the like around the white line by keeping the scanning area within a necessary range.

  In the present invention, in the method of extracting the white line candidate edge by the differential processing and binarization processing of the road image obtained by photographing the road surface, extracting the white line edge by determining the candidate edge, and recognizing the position of the white line, It is desirable to employ a step of calculating the white line width from the white line candidate edge and determining the white line edge when the calculated white line width is within a specified range. If this step is adopted, the white line recognition accuracy can be further improved as compared with the recognition method in which the white line width is not detected.

  In the present invention, a white line candidate edge is extracted by differential processing and binarization processing of a road image obtained by photographing a road surface from the vehicle side, a white line edge is extracted by determination of the candidate edge, and a white line position is recognized. Preferably, a step of calculating a lane width from each white line candidate edge of the left white line and the right white line on the basis of the vehicle and determining a white line edge when the calculated lane width is within a specified range is adopted. If this step is adopted, the white line recognition accuracy can be further improved as compared with the method in which the lane width is not detected.

  In the present invention, in a method of extracting a white line candidate edge by differential processing and binarization processing of a road image obtained by photographing a road surface, extracting a white line edge by determining the candidate edge, and recognizing a white line position, 1 When one white line edge is extracted from one scanning line, it is desirable to supplement the scanning range in which the white line edge cannot be extracted with the white line edge acquired in the scanning range before the past specified time. If this step is adopted, for example, even when the white line is a broken line, it is possible to obtain the white line edge by compensating with the past edge.

In the present invention, a white line candidate edge is extracted by differential processing and binarization processing of a road image obtained by photographing a road surface, and a white line edge is extracted by determination of the candidate edge, thereby extracting the white line position. When an approximate straight line is acquired by Hough transformation for the white line edge, it is desirable that a highly reliable white line edge extracted at the lower part of the road image is given a large weight to the vote value of the Hough transformation. In this way, even if noise is slightly extracted in the upper scanning range of the road image, if the correct white line edge is extracted in the lower scanning range, the white line can be recognized successfully. Even if noise is slightly extracted in the lower scanning range, if the correct white line edge is acquired in the upper scanning range, the white line can be recognized successfully. Thereby, the white line recognition accuracy can be further improved.

  The present invention is a white line recognition device that extracts white line candidate edges by differential processing and binarization processing of a road image obtained by photographing a road surface, extracts white line edges by determining candidate edges, and recognizes the position of the white line. A threshold value setting unit capable of setting two threshold values for binarization processing, and a threshold value switching unit which switches the threshold value alternately according to the shooting interval of the road image. Yes.

  According to the present invention, a white line can be normally recognized even in various unrecognized and misrecognized situations, whereby the white line recognition accuracy can be significantly improved.

Embodiments of a white line recognition method and apparatus according to the present invention will be described below with reference to the drawings. In addition, although the white line recognition technique of a present Example demonstrates the example applied to a passenger car, it can be applied also to other various vehicles.
(Hardware configuration)
1 and 2 are block diagrams of the white line recognition apparatus according to the present embodiment. The white line recognition apparatus includes a camera unit 1 as an image input unit and an image processing unit 2 for an image input from the camera unit 1. The camera unit 1 is composed of a CCD camera having an optical filter 11, a lens 12, an image sensor 13, and the like, and a road image in front of the vehicle photographed by the CCD camera is input to the image processing unit 2.

The image processing unit 2 is an ASIC (Application Specific Integrated Circuit) 20 as an integrated circuit for image processing, and a CPU (Central Processing) including a microprocessor that controls each device including the ASIC 20 and calculates and processes data. Unit) 3
0, a connector 40 as an external device interface (I / F) such as each control system and each sensor 50 is provided.

  The ASIC 20 includes an image acquisition unit 21 that acquires a road image input from the camera unit 1, a differentiation processing unit 22 that performs preprocessing necessary when extracting white line candidate points from the acquired road image, and a binarization process Part 23. Furthermore, the ASIC 20 extracts two fixed threshold values (set for a case where the luminance difference (density difference) between the white line and the road is large and small when the white line candidate edge is extracted by the differential processing and binarization processing of the road image. A threshold selection unit (see FIG. 2) for switching between a high threshold and a low threshold) alternately or selectively is used.

The CPU 30 reads out and executes an image processing program stored in advance in a memory (not shown), thereby extracting white line candidate points from the binarized road image output from the ASIC 20 (white line candidate point extracting unit). 31) and a function (white line recognition unit 32) for obtaining a straight line approximating the extracted white line candidate point sequence by Hough transform and recognizing it as a white line.

  FIG. 2 is a functional block diagram illustrating the functions of the respective units included in the image processing unit 2 more specifically. As shown in FIG. 2, the road image input from the image input unit 1 to the white line candidate point extraction unit 31 is first processed by the preprocessing unit (ASIC) 20. The preprocessing unit 20 has a threshold selection function, a differentiation processing function, and a binarization processing function.

A white line candidate point extraction unit 31 has a scanning area determination function and a white line edge candidate acquisition function for determining a scanning position and a scanning width of a scanning line for extracting a white line candidate point from a binarized input road image. , White line edge determination function, past white line edge complementation function, and the like. The white line recognition unit 32 has a well-known straight line approximation processing function based on the Hough transform. However, as will be described later, a highly reliable white line edge acquired at the lower part of the image can be given a large weight to the vote value of the Hough transform. Includes functionality.

Next, the white line recognition technology will be described more specifically together with the simulation results and driving experiment examples with actual vehicles.
(Problem countermeasures and new algorithms)
In order to improve the unrecognized and misrecognized problems that have been a problem in the conventional algorithm, a new algorithm, which will be described in the following, was developed. Details of the algorithm will be described later.

(Reduction of θ range in Hough transform)
In the conventional algorithm, the range of the θ value handled in the Hough transform is set to θ = 0 ° to 180 °. Accordingly, in order to detect straight lines having all inclinations appearing in the image, for example, a solid object such as a building or a vehicle in which θ = 0 ° (straight line) is recognized. Here, a solid object such as a building or a vehicle having θ = 0 ° (straight line) is, for example, a left or right straight part of a building that extends vertically, or a left end part or a right end part of a truck bed when viewed from the rear. It also has a linear portion extending in the vertical direction on the image.

<New algorithm 1>
In the new algorithm 1, the range of the θ value handled by the Hough transform is set, and θ = 30 ° to 70 °
It was. The basis for this is shown below.
(1) Definition of requirements for white line recognition The requirements for white line recognition are defined as follows.
1) A situation where the vehicle deviates from the white line after 1 second in a situation where the vehicle is traveling in its own lane.
2) The corresponding vehicle speed is 50km to 120km.
3) The state where the vehicle deviates from the white line of the own lane is a state where a part of the vehicle protrudes from the white line of the own lane.

(2) Environmental definition Since white line recognition on the expressway is assumed, the width of the lane is set to 3.5 m. The size of the vehicle was the same as that of the experimental vehicle, and the passenger car had a total length of 4900 mm and a total width of 1795 mm.

(3) Determination of the lower limit value of θ.
As shown in FIG. 3A, when considering a situation in which the vehicle 3 departs from the white line (left white line) a1 of its own lane after 1 second, the value of θ shown in FIG. This is a case where the vehicle is tilted to the left side even if it is in a state adjacent to the left white line a1 (reference position). That is, this is a case where the traveling direction of the vehicle 3 is not parallel to the left white line a1 but is slightly in the direction of the left white line a1.

  FIG. 3B shows the relationship between the left white line a1 and θ by the camera image. Here, since the left white line a1 is displayed based on a perspective real image in the case of a camera image, the left white line a1 at a position farther from the vehicle is displayed with an inclination closer to the center of the image. Therefore, θ corresponds to an angle formed by the perpendicular line 4 with respect to the left white line a1 and the upper limit line (a line corresponding to the traveling direction of the vehicle not parallel to the white line) 5 of the image.

  A value necessary for obtaining the lower limit value of θ in FIG. 3B was calculated (see FIG. 4). At that time, in order to facilitate the calculation, the direction of the vehicle was set to be parallel to the left white line a1 as shown in FIG. Further, as a condition for calculating the lower limit value of θ, the distance L0 = 13888 mm that the vehicle 3 travels at 50 km / h per second, the distance L1 = 6944.0 mm that is 1/2 of L0, the center of the vehicle (the position where the vehicle is mounted) The distance L2 from the left side white line a1 was set to 937.21 mm.

  FIG. 5 shows an image coordinate system of the in-vehicle camera in FIG. From the coordinates of the point A (−66.778, 72.974) and the coordinates of the point B (−133.55, 32.377), the lower limit value of θ in FIG. 5 was 32.367 °. Considering the mounting error, the angle is set to 30 °.

(4) Determination of the upper limit value of θ When considering a situation where the vehicle 3 departs from the white line a1 on the left side of its own lane after 1 second, the value of θ is maximized as shown in FIGS. 6 (a) and 6 (b). This is a case where the vehicle 3 is adjacent to the right white line a2 of its own lane R (reference position) and departs from the left white line a1 after 1 second. In FIG. 6B, a value necessary for obtaining the upper limit value of θ was calculated (see FIG. 7). In FIG. 7, the position of the camera, not the vehicle 3, is defined so as to be adjacent to the white line for easy calculation.

  Accordingly, in FIG. 7, the vehicle 3 travels at 50 km / h at a distance of 1 km L0 = 13888 mm, the distance from the left side of the vehicle to the intersection A with the white line a1 on the left side L1 = 10442 mm, and from the right side of the vehicle to the left side line 1 The distance L2 to the intersection B is 17567 mm, and the distance L2 from the center of the vehicle (on-vehicle camera mounting position) to the left white line a1 is 897.5 mm.

  FIG. 8 shows an image coordinate system of the in-vehicle camera at the upper limit value of θ in FIG. From the coordinates of point A (88.816, 37.788) and the coordinates of point B (52.793, 95.521), the lower limit value of θ in FIG. 8 was 67.819 °, but the camera mounting error Was set to 70 °.

(Upper scanning line width reduction)
In the conventional algorithm, as shown in FIG. 9, the width (scanning range) of the scanning line b that scans the white line a (hereinafter collectively referred to as the white line a including the left white line, the right white line, and the center line) is constant. (For example, 150 pixels). However, since the road on the image becomes narrower toward infinity, the conventional algorithm has extended the scanning range to a portion protruding from the road (an area inappropriate as a white line determination range).

<New algorithm 2>
In the new algorithm 2, as shown in FIG. 10, the scanning range of the scanning line b in the upper half is prevented from protruding outside the road by making the search range of the scanning line b of the upper half half (for example, 75 pixels). . This number of pixels (75 pixels) is an optimum value obtained by simulation. In addition, when a white line position can be recognized from a previously acquired road image among a plurality of road images acquired in time series, the scanning range is set so that the recognized white line position becomes the center of the scanning range. I made it. As a result, the scanning range can be shortened so as to speed up the arithmetic processing, and erroneous recognition due to the shoulders, grooves, etc. around the white line can be prevented.

  Furthermore, when the white line position can be recognized from the previously acquired road image, the scanning range can be shortened, and when the white line position cannot be recognized, the scanning range can be lengthened. As a result, the scanning area can be kept within a necessary range, and erroneous recognition or unrecognition of the white line due to a shoulder, a groove or the like around the white line can be prevented.

(Optimization of scanning range start / end position)
In the conventional algorithm, as shown in FIG. 11, the scanning start position of the scanning line b is at a position that is too high and the end position is at a position that is too low. For this reason, the scanning line b above the image scans the sky and landscape, and the scanning line b below scans the hood of the vehicle, and the scanning line extends to an inappropriate area as a white line determination range.

<New algorithm 3>
In the new algorithm 3, as shown in FIG. 12, the start position in the vertical direction of the scanning range by the scanning line b is set at infinity of the image, and the end position is set in front of the hood of the vehicle. As a result, the number of scanning lines can be reduced so as to speed up the arithmetic processing, and erroneous recognition due to a shoulder, a groove, etc. around the white line can be prevented.

(Weighing votes during Hough conversion)
In the conventional algorithm, the vote values at the time of Hough conversion are all “1”. In other words, in the white line edge extraction, the lower part of the image with high reliability (near the vehicle) and the upper part of the image with poor reliability are set to vote with the same value during the Hough transform.

<New algorithm 4>
In the new algorithm 4, in the white line edge extraction, a high weight is given to the vote value at the time of the Hough conversion for the lower part of the image (near the vehicle) with high reliability. Specifically, the vote value of the edge extracted by the lower half scanning line is “3”, and the vote value of the edge extracted by the upper half scanning line is “2”. Vote values “3” and “2” are optimum values obtained by simulation. However, the weight of the voting value needs to be adjusted again when the white line edge extraction method is changed (because the reliability ratio between the upper side and the lower side in white line edge extraction changes).

  In this way, even if noise is slightly extracted in the upper scanning range of the road image, if the correct white line edge is extracted in the lower scanning range, the white line can be recognized successfully. Even if noise is slightly extracted in the lower scanning range, if the correct white line edge is acquired in the upper scanning range, the white line can be recognized successfully. Thereby, the white line recognition accuracy can be further improved.

(Extraction of both ends of white line edge and judgment of distance between edges of both ends)
In the conventional algorithm, white line edges are extracted by the following procedures (1) and (2).
(1) In order to extract the inner edge, the search line is searched from the inner side to the outer side, and a High pixel (inner edge) having a width of 3 pixels or more is extracted.
(2) In order to extract 5 consecutive neighborhoods, it is confirmed that there are 5 consecutive high pixels having a width of 3 pixels or more in the vertical direction of the edge extracted in (1).

  Examples of white line (right side) edge extraction are shown in FIGS. An image subjected to the differential filter + binarization process is an image shown in FIG. In the figure, d indicates noise. FIG. 14 shows the edge extraction state by enlarging the inside of the thick frame c in the image of FIG. In the example of FIG. 14, the white line edge can be extracted from the scanning line b1. However, the scanning lines b2 and b3 are misrecognized by extracting noise inside the white line edge.

<New algorithm 5>
In the new algorithm 5, in order to discriminate a white line, the following steps (1) to (5) are used to extract edges at both ends (inner side and outer side) of the white line and perform determination based on the distance between the edges at both ends. A schematic diagram of the extraction procedure is shown in FIG.

(1) In order to extract the inner edge, the search line is searched from the inner side to the outer side, and a High pixel (inner edge) having a width of 3 pixels or more is extracted.
(2) In order to extract 5 consecutive neighborhoods, it is confirmed that there are 5 consecutive high pixels having a width of 3 pixels or more in the vertical direction of the edge extracted in (1).
(3) In order to extract a low pixel, the outside of the edge extracted in (1) is searched, and it is confirmed that a low pixel having a width of 3 pixels or more exists.
(4) In order to extract the outer edge, the outside of the low pixel in (3) is searched, and a high pixel (outer edge) having a width of 3 pixels or more is extracted.
(5) In order to extract five consecutive neighborhoods, it is confirmed that there are five consecutive high pixels having a width of 3 pixels or more in the vertical direction of the edge extracted in (4).
(6) To check the white line width, confirm that the interval between the inner edge extracted in (1) and the outer edge extracted in (4) is 40 pixels or less (within the specified range). If there is, it is recognized as a white line. The reason why the number of pixels is 40 pixels is that the maximum value of the width of the white line edge in the image used in the simulation is 30 pixels, and a margin of 10 pixels is added to the width.

(White line edge and lane width judgment)
Further, with respect to the white line edge recognized in this way, as shown in FIG. 16, the white line edge (the edge of the left white line a1 and the right white line a2) can be held on both the left and right sides of the paired scanning lines. Only the coordinates are extracted, the interval between the left and right white line edges (lane width R) is confirmed, and if the lane width R is within the specified range, it is determined as a white line edge. That is, the lane width R is calculated from the white line candidate edges of the left white line a1 and the right white line a2 with reference to the vehicle, and when the calculated lane width R is within a specified range, it is determined as a white line edge.

(Performance evaluation by simulation)
A new algorithm was simulated on a PC to evaluate its performance. In the simulation, an image (BMP file) collected in an actual running experiment was used.

(Evaluation results)
As a result of judging the recognition success, unrecognition, and misrecognition for the image superimposed on the white line recognition result, it was confirmed that the recognition rate was improved compared to the previous driving experiment. Therefore, the white line recognition accuracy can be improved as compared with the white line recognition method that does not detect the white line width. Furthermore, the white line recognition accuracy can be further improved as compared with the white line recognition method that does not detect the lane width.

<New algorithm 6>
(Use edge information of past frames)
When the white line is a broken line, the white line may not be recognized because the number of extracted edges is small. For this reason, the edge information of the past frame is held and used to compensate for edge information that is insufficient with the current frame alone, and the white line position can be recognized. In the case of FIG. 17, only three white line edges can be extracted, and the white line position cannot be recognized. However, the edge extracted in the previous frame is retained as shown in FIG. 18, and the position of the white line is specified by supplementing the edge with respect to the scanning line in which the edge could not be extracted in the current frame as shown in FIG. can do. In FIG. 19, a portion surrounded by a broken line e in a rectangular shape is a portion supplemented by an edge acquired in a past frame. As a result of simulation and examination in this algorithm, edge information is retained and used up to the previous four frames.

<New algorithm 7>
(Use two thresholds during binarization)
If the road surface suddenly becomes bright at the exit of the tunnel, the white line is in shadow, the white line is faint, or the like, the white line cannot be extracted from the image when the contrast between the white line and the road is small. However, it is possible to extract features by setting a threshold value for binarization processing to be equal to or less than the density difference between the white line in the shadow and the road (appropriate threshold value). For example, when the density value of the white line is 70 and the density value of the road is 40, the threshold value may be set to 30. An appropriate threshold value in the binarization process varies depending on the weather, time zone, and road conditions, but it is difficult to dynamically set the threshold value to an appropriate value depending on the situation. If the threshold is fixed, there are advantages and disadvantages when the threshold is high and when the threshold is low.

  That is, at a low threshold, there is a merit that a white line can be extracted even when the contrast between the white line and the road is small, but there is a demerit that an edge such as a shadow of a preceding vehicle or a building is extracted. The reverse is true for high thresholds. Therefore, in the new algorithm 7, two fixed threshold values (a high threshold value and a low threshold value) having a low threshold value of 30 and a high threshold value of 50 are alternately used every 100 ms. This makes it possible to absorb the demerits by taking advantage of the advantages and disadvantages of the high and low thresholds.

  The binarized images at the respective threshold values are shown in FIGS. In FIG. 20 (threshold 30), the white line edge is clearly visible, but the edge blurs and the noise d is large. In FIG. 21 (threshold value 50), the edge of the white line on the left side is faint, but the edge blurs and the noise is small.

  Therefore, the white line a can be normally recognized even in an unrecognized or erroneous recognition occurrence situation that is likely to occur when, for example, the contrast between the white line a and the road is large or small. For example, even when the white line a is in the shadow and the contrast is small, the white line a can be recognized normally if the threshold is small. Thereby, the white line recognition accuracy can be significantly improved.

<New algorithm 8>
(Refine white line scanning area)
In order to prevent misrecognition of extracting edges other than the white line due to the scanning area extending outside the white line, the range of the scanning area is narrowed down. Simply narrowing the scanning area leads to unrecognition of the white line, so the width of the lower scanning line is reduced depending on the state of the calculation table.

  When the calculation table state is, for example, less than 10 consecutive white line edge extraction processing successes, state 1 is assumed, and when the calculation table is in state 2 (10 to less than 30 consecutive white line edge extraction processing successes), the correct white line There is a high possibility that the position (or its vicinity) has been acquired. Since the images are acquired in time series (every 0.1 second), the next white line position is considered to be near the previous white line position. Further, when the calculation table is in state 3 (the white line edge extraction process has been successfully performed 30 times or more), the reliability of correctly acquiring the white line position and the next white line position in state 2 or higher is high. Therefore, in the case of state 2, the width of the lower scanning line is narrowed, and in the case of state 3, it is further narrowed.

  Therefore, the scanning range can be shortened when the white line position can be recognized from the previously acquired road image, and the scanning range can be lengthened when the white line position cannot be recognized. As a result, the scanning area can be kept within a necessary range, and erroneous recognition or unrecognition of the white line due to a shoulder, a groove or the like around the white line can be prevented.

  Next, an example of a white line recognition procedure in such a white line recognition technique will be described with reference to the flowchart of FIG. This process is performed by the image processing unit 2, but will be described with reference to the configuration shown in FIGS. This process is repeatedly executed during the operation of the white line recognition processing apparatus.

  A road image photographed by the camera unit 1 as an image input unit is input (step S1). Then, in the preprocessing unit 20 of the image processing unit 2, when the density difference between the white line and the road other than the white line is large, one of two threshold values for binarization processing set in advance for a small case (For example, the low threshold 30 shown in Table 11) is selected (step S2), and a road image is acquired by the selected threshold. The acquired road image is subjected to differentiation processing (step S3) and binarization processing (step S4) in order, and then input to the white line candidate point extraction unit 31. Regarding the selection of the two threshold values, in the embodiment, the two threshold values are alternately selected in accordance with the photographing interval by the camera unit 1. Therefore, two threshold values are used alternately.

  When a road image is input to the white line candidate point extraction unit 31, a scanning area is determined. That is, as shown in FIG. 12, the scanning line position for determining the scanning line start position and end position for the road image is determined (step S5), and the scanning line width which is the length of the scanning line is further determined (step S6). For example, as shown in FIG. 10, the scanning line width is set so that the upper scanning line width and the lower scanning line width of the white line are different. Next, scanning for white line edge candidate extraction by the scanning line is performed, white line edge candidates are acquired, and white line edges are determined.

  In the white line candidate point extraction unit 31, when the white line edge on the scanning line is acquired (step S7), it is determined whether the white line inner and outer edges have been acquired (step S8). If YES, the process proceeds to step S9. If NO, the process proceeds to step S12. In step S9, it is determined whether or not the white line width is appropriate by detecting the edge interval shown in FIG. 15. If YES, the process proceeds to step S10, and if NO, the process proceeds to step S12.

  In step S10, it is determined whether or not the lane width corresponding to the left and right white line interval of the vehicle is appropriate. If YES, the extracted edge coordinates are held (step S11), and the process proceeds to step S13. If NO, the process proceeds to step S12.

  On the other hand, if NO is determined in steps S8, S9, and S10, the process proceeds to step S12. Here, after the past edge coordinates are held for each of steps S8, S9, and S10, the process proceeds to step S13.

  Next, in step S13, it is determined whether or not all scanning lines have been examined. If YES, the process proceeds to step S14, and if NO, the process returns to step S7. Then, a loop process similar to the above is performed until YES is determined in step S13 (until all scanning lines are checked).

If YES is determined in step S13, the white line recognition unit 32 performs straight line approximation processing by Hough transform, and the approximate straight line is recognized as a white line.

  In the above embodiment, the technology for recognizing the white line from the road image obtained by photographing the road surface in front of the vehicle has been described. Applicable. In such a case, it is possible to cope with such a situation by installing a rear in-vehicle camera (imaging means) that can photograph a road surface, a parking lot, or the like behind the vehicle.

It is a block diagram of the white line recognition apparatus which concerns on the Example of this invention. It is a functional block diagram of the white line recognition apparatus which concerns on the Example of this invention. It is explanatory drawing of the white line recognition technique which concerns on the Example of this invention. It is explanatory drawing of the white line recognition technique which concerns on the Example of this invention. It is explanatory drawing of the white line recognition technique which concerns on the Example of this invention. It is explanatory drawing of the white line recognition technique which concerns on the Example of this invention. It is explanatory drawing of the white line recognition technique which concerns on the Example of this invention. It is explanatory drawing of the white line recognition technique which concerns on the Example of this invention. It is explanatory drawing which shows the scanning line width of the conventional white line recognition technique. It is explanatory drawing of the road image which shows the scanning line width of the white line recognition technique which concerns on the Example of this invention. It is explanatory drawing of the road image which shows the scanning line start position and end position of the conventional white line recognition technique. It is explanatory drawing of the road image which shows the scanning line start position and end position of the white line recognition technique which concerns on the Example of this invention. It is explanatory drawing which shows the example of a road image after the binarization process in the conventional white line recognition technique. It is the elements on larger scale of FIG. It is explanatory drawing which shows the white line edge extraction example of the white line recognition technique which concerns on the Example of this invention. It is a figure which shows the example of lane width extraction of the white line recognition technique which concerns on the Example of this invention. It is explanatory drawing which shows the example of a road image in case the left side white line in the white line recognition technique which concerns on the Example of this invention is a broken line. It is explanatory drawing which shows the road image example of the 1st front frame in the white line recognition technique which concerns on the Example of this invention. It is explanatory drawing which shows the road image example of the present flame | frame in the white line recognition technique which concerns on the Example of this invention. It is explanatory drawing which shows the example of a binarization process image of the threshold value 30 which concerns on the Example of this invention. It is explanatory drawing which shows the example of a binarization process image of the threshold value 50 which concerns on the Example of this invention. It is a processing flowchart in the white line recognition technique which concerns on the Example of this invention.

Explanation of symbols

1 Camera unit (image input unit)
DESCRIPTION OF SYMBOLS 11 Optical filter 12 Lens 13 Image sensor 2 Image processing part 20 ASIC
21 Image acquisition unit 22 Differentiation processing unit 23 Binary processing unit 30 CPU
31 White line candidate point extraction unit 32 White line recognition unit 40 Connector A Road surface a White line a1 Left white line a2 Right white line b Scanning line

Claims (9)

A white line recognition method for extracting a white line candidate edge by differential processing and binarization processing of a road image obtained by photographing a road surface, extracting a white line edge by candidate edge determination, and recognizing a white line position,
A white line recognition method, characterized in that at least two threshold values for binarization processing are set, and the threshold values are switched and used.   2. The white line recognition method according to claim 1, wherein the two threshold values are alternately used in accordance with a shooting interval for obtaining the road image. A white line recognition method for extracting a white line candidate edge by differential processing and binarization processing of a road image obtained by photographing a road surface, extracting a white line edge by candidate edge determination, and recognizing a white line position,
Of the plurality of road images acquired in time series, when the white line position can be recognized from the previously acquired road image, the scanning range is set so that the recognized white line position is at the center of the scanning range. A white line recognition method. A white line recognition method for extracting a white line candidate edge by differential processing and binarization processing of a road image obtained by photographing a road surface, extracting a white line edge by candidate edge determination, and recognizing a white line position,
Among a plurality of road images acquired in time series, the scanning range is shortened when the white line position can be recognized from the previously acquired road image, and the scanning range is lengthened when the white line position cannot be recognized. A white line recognition method. A white line recognition method for extracting a white line candidate edge by differential processing and binarization processing of a road image obtained by photographing a road surface, extracting a white line edge by candidate edge determination, and recognizing a white line position,
A white line recognition method, wherein a white line width is calculated from the white line candidate edge, and a white line edge is determined when the calculated white line width is within a specified range. This is a white line recognition method for extracting a white line candidate edge by differential processing and binarization processing of a road image obtained by photographing a road surface from the vehicle side, extracting a white line edge by candidate edge determination, and recognizing a white line position. And
A white line recognition method, wherein a lane width is calculated from each white line candidate edge of a left white line and a right white line with reference to the vehicle, and a white line edge is determined when the calculated lane width is within a specified range. A white line recognition method for extracting a white line candidate edge by differential processing and binarization processing of a road image obtained by photographing a road surface, extracting a white line edge by candidate edge determination, and recognizing a white line position,
When one white line edge is extracted for one scanning line, the scanning range in which the white line edge cannot be extracted is complemented with the white line edge acquired in the scanning range before the past specified time. A white line recognition method. A white line recognition method for extracting a white line candidate edge by differential processing and binarization processing of a road image obtained by photographing a road surface, extracting a white line edge by candidate edge determination, and recognizing a white line position,
A white line characterized by giving a high weight to the vote value of the Hough transform to the reliable white line edge extracted at the bottom of the road image when obtaining an approximate straight line by Hough transform for the extracted white line edge Recognition method. A white line recognition device that extracts a white line candidate edge by differential processing and binarization processing of a road image obtained by photographing a road surface, extracts a white line edge by candidate edge determination, and recognizes a white line position,
Threshold setting means capable of setting two threshold values for binarization processing;
A white line recognition apparatus comprising: threshold value switching means that alternately uses the threshold value in accordance with a shooting interval of the road image when extracting the candidate edge of the white line.

Images