JP2007141167A - Lane detector, lane detection method, and lane detection program - Google Patents

Abstract

<P>PROBLEM TO BE SOLVED: To provide a lane detector, lane detection method, and lane detection program capable of suppressing throughput while maintaining high operating accuracy. <P>SOLUTION: White line edge candidate points are extracted from inside an image photographed with a camera 10. An approximate line to the white line edge is obtained from the extracted candidate points by Hough transformation. Here, in a status 1, the approximate line is obtained with an angular resolution of 9 degrees, for instance, and a wide range of inclination detection (0 to 179 degrees, for instance). When the approximate line is detected ten times, for instance, consecutively in the status 1, a shift is made to a status 2. In the status 2, the approximate line is detected with the angular resolution of 4 degrees, for instance, and a medium range of inclination detection (80 degrees, for instance) centered around the inclination of the approximate line obtained in the status 1. When the approximate line is detected 30 times consecutively in the status 2, a shift is made to a status 3. In the status 3, the approximate line is detected with the angular resolution of 2 degrees, for instance, and a narrow range of inclination detection (40 degrees, for instance) centered around the inclination of the approximate line obtained in the status 2. <P>COPYRIGHT: (C)2007,JPO&INPIT

Description

  The present invention relates to a lane detection device, a lane detection method, a lane detection program, and a straight line detection method for detecting a lane with respect to a photographed vehicle front or rear image.

  Conventionally, there is an algorithm called Hough transform as an algorithm for detecting a straight line in an image. A set of points in the image can be extracted as a line segment by the Hough transform.

As a conventional technique using such a Hough transform, the object position data measured by the measuring means is subjected to a Hough transform to estimate the curvature of the road ahead, and whether the vehicle ahead is in the same lane as the own vehicle by the estimated road curvature. There is a vehicle forward monitoring device that determines whether or not (for example, Patent Document 1 below).
JP 2001-167396 A

  However, in the process of the Hough transform, if a straight line is to be obtained at all angles for an extracted point, a very large amount of processing is required.

  On the other hand, when trying to find straight lines at a certain angle interval instead of all angles, the processing amount is reduced by that amount, but the calculation accuracy is lowered and the straight line accuracy is lowered.

  There is a similar problem in a lane detection device that detects a lane using Hough transform. In particular, in such an apparatus, since various processes are performed in addition to the process based on the Hough transform, it is difficult to spend time only on the process based on the Hough transform.

  The present invention has been made in view of the above problems, and an object of the present invention is to provide a lane detection device, a method thereof, and a lane detection program that suppresses the processing amount and maintains high calculation accuracy.

  In order to achieve the above object, the present invention provides a lane detection device that detects a lane from a candidate point in an extracted image by Hough transform, and is a straight line passing through the candidate point and an angle between adjacent straight lines. Detection in which the resolution is increased while changing the state, the inclination detection range, which is the inclination range of the straight line centered on the candidate point, is reduced while changing the state to obtain an approximate straight line, and the lane is detected by the approximate straight line Means are provided.

  Further, in the lane detection device according to the present invention, when the inclination and intercept of the approximate straight line are within a predetermined range, the detection means increases the angular resolution as compared with the current state, The approximate straight line is obtained by shifting to a state.

  Furthermore, the present invention provides the lane detection device, wherein the detection means reduces the inclination detection range by limiting the inclination detection range in the current state based on the result of the approximate straight line obtained in the previous state. Features.

  Furthermore, the present invention is characterized in that, in the lane detection device, the detection unit weights the candidate points extracted at the lower part of the image larger than the candidate points extracted at the upper part to obtain the approximate straight line. .

  Furthermore, the present invention is characterized in that in the lane detecting device, the detecting means extracts more candidate points at the lower part of the image than the upper candidate points to obtain the approximate line.

  Further, in the lane detecting device according to the present invention, the detecting means divides the image into left and right, and an approximate straight line on the other side is obtained with reference to an approximate straight line obtained on the side with a lot of extraction points among the divided left and right images. It is characterized by obtaining.

  Furthermore, the present invention is characterized in that, in the lane detecting device, the detecting means extracts the vicinity of the approximate straight line obtained in the previous state as the candidate point.

  Furthermore, in the lane detection device according to the present invention, when the detection unit continuously detects the approximate straight line a predetermined number of times, the angle resolution is increased from the current state and the inclination detection range is reduced to the next state. The approximate straight line is obtained by shifting.

  Furthermore, in the lane detecting device according to the present invention, when the detecting means cannot detect the approximate straight line in the current state, the state shifts to a state where the angular resolution is the lowest and the inclination detection range is large. Features.

  Furthermore, the present invention provides the lane detection device, wherein the detection means reduces the inclination detection range by setting the inclination detection range within a predetermined range centered on the inclination of the approximate straight line obtained in the previous state. It is characterized by.

  Further, in the lane detection device according to the present invention, the detection means may extract more candidate points to be extracted at the lower part of the image than at the upper part by taking more lines to scan at the lower part of the image than at the upper part. Features.

  In order to achieve the above object, the present invention provides a lane detection method in a lane detection apparatus for detecting a lane from a candidate point in an extracted image by Hough transform, wherein each straight line passes through the candidate point. The angle resolution of adjacent angles is increased while changing the state, and the inclination detection range, which is the inclination range of the straight line centered on the candidate point, is reduced while changing the state to obtain an approximate straight line. The lane is detected.

  Furthermore, in order to achieve the above object, the present invention provides a lane detection program for detecting a lane from a candidate point in an extracted image by a Hough transform, and a straight line passing through the candidate point and an angle adjacent to each straight line. The angle resolution is increased while changing the state, and the inclination detection range, which is the inclination range of the straight line centered on the candidate point, is reduced while changing the state to obtain an approximate line, and the lane is detected by the approximate line. It is characterized by causing a computer to execute the processing to be performed.

  Furthermore, in order to achieve the above object, the present invention provides a straight line detection method in a straight line detection apparatus for detecting a straight line in an image from a candidate point in an extracted image by Hough transform, wherein the straight line passes through the candidate point. While increasing the angle resolution of the adjacent angles of each straight line while making a state transition, an approximate straight line is obtained by reducing the slope detection range, which is the slope range of the straight line centered on the candidate point, while making a state transition, A straight line in the image is detected by an approximate straight line.

  According to the present invention, it is possible to provide a lane detection device that suppresses the processing amount and maintains high calculation accuracy.

  The best mode for carrying out the present invention will be described below with reference to the drawings. FIG. 1 is a diagram illustrating a configuration example of a lane detection device 1 according to the present invention.

  The lane detection device 1 includes a camera 10, a microcomputer 20, and a vehicle body control unit 30.

  The camera 10 is for taking an image in front of the vehicle. The captured image is output to the microcomputer 20.

  For example, as shown in FIG. 2A, the camera 10 is attached in front of the vehicle 2 and in the vicinity of the inner mirror inside the vehicle. When the front of the vehicle 2 is photographed at such a position, an image including two lanes (white lines) 3 and 4 in front of the vehicle 2 is obtained as shown in FIG.

  The microcomputer 20 receives an image from the camera 10, extracts candidate points of the edge portions of the white lines 3 and 4 from the image, performs Hough transform, and detects an approximate straight line of the white lines 3 and 4. After detecting the approximate straight line, the microcomputer 20 outputs data regarding the position and direction. The microcomputer 20 performs various other processes such as detecting a vehicle ahead and detecting an obstacle. The microcomputer 20 serves as detection means for detecting a lane.

  The vehicle body control unit 30 is input with data on the position and direction of the approximate straight line related to the white lines 3 and 4 from the microcomputer 20, and controls the speed of the vehicle 2, for example. The vehicle body control unit 30 is connected to various devices of the vehicle 2. Various data other than data such as the position of the lane are input from the microcomputer 20, and various devices are controlled accordingly.

  Next, the Hough transform will be described. For example, it is assumed that the microcomputer 20 extracts candidate points as edges of white lines 3 and 4 from the image from the camera 10. As shown in FIG. 3A, the vertical axis is y, the horizontal axis is x, and the extraction point is (x1, y1). At this time, the straight line passing through the extraction point (x1, y1) is assumed to have an inclination a and an intercept b.

Is represented.

  Now, instead of the slope and intercept, as shown in FIG. 3A, the length of the perpendicular drawn from the origin to the straight line passing through the extraction point (x1, y1) is ρ, and the angle between the perpendicular and the x axis is θ Then, the equation of the straight line is

Is represented. That is, when a certain point (x1, y1) is given, a set of all straight lines passing through this point is represented by a curve (FIG. 3B) on the coordinates of the horizontal axis θ and the vertical axis ρ. In other words, a certain point on this curve corresponds to the straight line shown in FIG.

  Then, as shown in FIG. 4A, when a plurality of extraction points are obtained, a plurality of curves are obtained as shown in FIG. 4B when a curve corresponding to each extraction point is drawn. The point where these curves intersect most often is an approximate straight line (straight line indicated by a dotted line in FIG. 4A) passing through each extraction point. White lines 3 and 4 are detected by this approximate straight line.

  Next, a technique for maintaining high accuracy of calculation with a small amount of processing in processing by Hough transform will be described. FIG. 5 is a diagram conceptually showing such a method.

  First, when an extraction point is extracted from the image in state 1, the angle resolution is 9 °, the inclination detection range is large (in this embodiment, a range of 0 ° to 179 °), and all straight lines passing through the extraction point are calculated. .

  Here, the angular resolution is an angle interval between adjacent straight lines passing through extraction points. Therefore, the angular resolution of 9 ° means that all straight lines passing through the extraction points at intervals of 9 ° are obtained.

  Further, the inclination detection range is a range of inclination of a straight line passing through the extraction point. Therefore, when the inclination detection range is 0 ° to 179 °, a straight line passing through the extraction point is obtained over a range from 0 ° to 179 ° centering on the extraction point.

  In state 1, a combination of the most votes as shown in FIG. 4B described above is obtained from a plurality of extracted points, and an approximate straight line passing through all the extracted points is obtained. In this embodiment, this approximate straight line becomes a white line edge. At this time, the number of times that the approximate straight line has been detected continuously is obtained.

In state 2, an approximate straight line is obtained again from the extracted points, but the angle resolution is 4 °, and the inclination detection range is medium (for example, −40 ° ≦ θ with respect to the inclination θ _{1 of the} approximate straight line obtained in state 1) ₁ ≦ + 40 °, 80 ° range). After detecting the approximate straight line obtained in state 1, the slopes of the white lines 3 and 4 do not change abruptly, so the calculation is performed using the results obtained in the previous detection. In this state 2, the inclination detection range is limited as compared with state 1, and the angular resolution is increased. The resolution is increased to prevent the calculation processing amount from increasing. When the approximate straight line is continuously detected in the state 2, for example, 30 times or more, the state 3 is entered.

In state 3, the angle resolution is 2 °, and the inclination detection range is small (for example, the range of −20 ° ≦ θ ₂ ≦ + 20 ° and 40 ° with respect to the inclination θ _{2 of the} approximate line obtained in state 2). Find an approximate line from a point. As in the case of the state 2, the result obtained in the state 2 is used in consideration of the fact that the slopes of the white lines 3 and 4 do not change rapidly. State 3 further limits the tilt detection range as compared with state 1 and state 2 and increases the angular resolution to improve the accuracy of the approximate straight line. The resolution is increased to prevent the calculation processing amount from increasing. If an approximate straight line can be detected in this state 3, a highly accurate white line edge can be detected.

  If detection of the approximate straight line fails even once in each state, the state shifts to state 1. Normally, the white lines 3 and 4 of the road are located obliquely with respect to the image from the camera 10. For example, an approximate straight line is detected in the vertical direction, or the width of the white lines 3 and 4 is larger than a certain interval. Furthermore, it is assumed that the detection of the approximate straight line has failed when the approximate straight line obtained last time and the approximate straight line obtained this time are matched and greatly changed. Actually, in each case, it is assumed that the detection of the approximate line fails when a certain threshold is set and the approximate line is detected exceeding the threshold.

  Next, how the state of the curve on the ρ-θ coordinate changes in each state will be described. FIG. 6 shows an example. The value of the angular resolution in each state is the same as in the example of FIG.

As shown in FIG. 6A, in the state 1, an approximate straight line is obtained from the extraction points in the entire inclination detection range (0 ° ≦ θ ≦ 179 °). However, since the angular resolution is 9 °, the horizontal axis of the curve is plotted at 9 ° intervals. A plurality of curves are obtained from the plurality of extracted points, and the highest voting position (θ ₁ and ρ ₁ ), that is, an approximate line is obtained from these curves. If this is successful 10 times, for example, the state 2 is entered.

As shown in FIG. 6B, in state 2, the entire straight line is obtained from the extracted points in the range of ± 40 ° with the inclination detection range centering on the inclination θ ₁ of the approximate straight line detected in state 1. Since the angular resolution in this case is 4 °, the horizontal axis is plotted at intervals of 4 °. In this state 2, the highest voting position is obtained from a plurality of curves. For example, when the detection of the approximate straight line is successful 30 times in succession, the state shifts to state 3.

As shown in FIG. 6C, in state 3, the inclination detection range is obtained for all straight lines in a range of ± 20 ° with respect to the inclination θ _{2 of the} approximate straight line detected in state 2. In this case, since the angular resolution is 2 °, the horizontal axis is plotted at intervals of 2 °.

  As shown in FIG. 6A to FIG. 6C, in the state 1, the linear inclination range is detected in the entire range, and the range is narrowed as the state shifts from the state 1 to the state 3. On the other hand, the angular resolution decreases as the state shifts from state 1 to state 3.

  That is, the curve on the [rho]-[theta] coordinate becomes smoother as the detection range becomes narrower as the state shifts from state 1 to state 3. Therefore, although the accuracy of the approximate straight line gradually increases, the number of calculations does not increase because the detection range is narrowed.

  Note that the numerical values shown in the examples of FIGS. 5 and 6 are examples, and if the values are set so that the angular resolution increases and the inclination detection range becomes narrower as the state 1 is shifted to the state 3, the example described above. Has the same effect as.

  Next, based on the above points, the operation for detecting the approximate straight line of the lane detection device 1 will be described.

  FIG. 7 is an example of a flowchart showing the operation. First, when starting the operation of this processing, the microcomputer 20 sets an extraction range of white line edge candidate points (S11).

  For example, as shown in FIG. 9, the microcomputer 20 scans the image obtained from the camera 10 in the horizontal direction and extracts candidate points. At this time, the extraction range is set in the vicinity of the previously calculated approximate straight line (or a straight line stored in the memory in the microcomputer 20 in advance). For example, a range of 100 pixels before and after the previously calculated approximate straight line scanning direction is set. As a result, the processing amount can be reduced as compared with the extraction in the range of all pixels.

  Since the microcomputer 20 detects the left white line 3 and the right white line 4 when extracting candidate points by scanning, the microcomputer 20 halves the left white line candidate point and the right white line candidate point.

  After setting the extraction range, the microcomputer 20 extracts white line edge candidate points within the range (S12). At this time, the microcomputer 20 increases the number of lines extracted from the image in the lower part of the image than in the upper part of the image.

  For example, as shown in FIG. 10A, the lower part of the image is near the vehicle 2 and the white lines 3 and 4 are thick, and the upper part of the image is far from the vehicle 2, so the white lines 3 and 4 are thin. Since the white lines 3 and 4 are thick in the lower part of the image, if more candidate points can be extracted in this range, a more probable approximate straight line can be obtained.

  That is, by increasing the number of extraction lines at the bottom of the image, as shown in FIG. 10A, the two extraction points A1 and A2 at the bottom of the image can be extracted, and the extraction point A3 at the top of the image cannot be extracted. When each curve is obtained from the extracted points at the bottom of the image, each curve has a similar shape as shown in FIG. The intersection is highly likely to intersect at one point, and therefore a more probable approximate straight line can be obtained.

  In order to increase the number of extracted lines, for example, a target area of an image is set as shown in FIG. 10 (A), the total number of lines is scanned in the target area, and every 10 lines are scanned outside the target area. To be equal. Of course, any number of lines may be set as long as the number of lines scanned in the target area is larger than the number of lines scanned in the other areas.

  Returning to FIG. 7, after extracting the candidate points, the microcomputer 20 determines whether or not the number of white line edge candidate points extracted is equal to or greater than a predetermined number (S13). This is because the reliability of a straight line obtained by obtaining an approximate straight line by Hough transform after securing a certain number of extractions is high. The predetermined number is, for example, a value such as “5” or “10”. Of course, other values may be used.

  When the number of extractions is less than the predetermined number (NO in S13), the microcomputer 20 clears the number of continuous extractions (S15). Up to now, when the approximate lines of the white lines 3 and 4 have been extracted continuously, the number of extraction is cleared because the approximate line cannot be detected. That is, the state 1 is transferred. And a process transfers to S11 again and repeats the above-mentioned process.

  On the other hand, when the number of extractions is equal to or greater than the predetermined number (YES in S13), the microcomputer 20 sets the angular resolution used in the Hough transform (S14). Depending on the state, the angular resolution is set to a predetermined value, for example, 9 °, 4 °, or 2 °.

  Next, the microcomputer 20 sets the range of the tilt detection angle necessary for the Hough transform (S17). As described above, the detection angle range is set according to the current state. In the case of state 2 or state 3, the range is limited based on the tilt detection angle obtained in the previous state.

  Next, the microcomputer 20 performs voting for the approximate straight line candidate (S18). At each extraction point, a curve is obtained on the ρ-θ coordinate based on the set angle resolution and inclination detection range, and an intersection where each curve intersects most often is obtained.

  In this case, if the candidate points are weighted when voting for the Hough transform, the intersection of the weighted curves can be selected and obtained when there are a plurality of intersections. For example, as shown in FIGS. 11A and 11B, three candidate points are extracted, the curve of the candidate point A1 closer to the vehicle 2 is weighted by +3, and the curve of the farthest candidate point A3 is used. A weight of +1 and a weight of +2 are applied to the curve of the candidate point A2 in the middle. Then, when obtaining the intersection of each curve, for example, as shown in FIG. 11B, when there are two intersections X1 and X2, it is not the intersection X2 by the extraction points A1 and A3 but the extraction points A1 and A2. The direction of the intersection point X1 is selected. This is because the candidate point closer to the vehicle 2 has higher reliability, and therefore, it is easier to obtain a correct approximate straight line approximating the edges of the actual white lines 3 and 4.

  Next, the microcomputer 20 detects an approximate straight line (S19 in FIG. 8). Since ρ and θ are obtained from the combination of the most votes, an approximate straight line is obtained using these values. That is, from Equation 2,

  And an approximate straight line is obtained by substituting the obtained ρ and θ.

  As described above, candidate points are divided into a left side and a right side with respect to one image, and approximate lines are obtained respectively. When detecting the approximate line, the left and right white line candidate points are used as an index for detecting the approximate line on the other side with reference to the side with more extracted points.

  For example, as shown in FIG. 12A, consider the case where the left image has five extraction points and the right image has three extraction points. When a curve is obtained from the extracted points on the left side, for example, there is one concentrated intersection as shown in FIG. On the other hand, when a curve is obtained from the extracted points on the right side, for example, as shown in FIG. 12C, it is difficult to obtain an intersection where the extracted points are concentrated by a small amount compared to the right side. Accordingly, when the right approximate line is detected, the approximate straight line detected on the left side is used as an index, and for example, an optimal approximate line is selected from the left approximate line. Thereby, the accuracy of the approximate straight line can be improved.

  Next, the microcomputer 20 detects the number of approximate straight line detections (S20). As described above, the number of consecutive detections is an index for determining the likelihood of the approximate straight line, and state transition is performed based on this. Note that if the slope and intercept of the obtained approximate straight line (S19) are within a certain range, it may be determined that the approximate straight line is continuously detected. In this case, instead of detecting the number of times of continuous detection, the certainty is determined from the approximate straight line obtained in S19 by its inclination or the like.

  Then, it is determined whether a white line edge is detected (S21). For example, whether or not a white line edge is detected may be determined based on whether or not an approximate straight line has been obtained in the state 3.

  When a white line edge is detected (YES), a series of processing ends (S22). When it cannot be detected (NO in S21), the process proceeds to S11 again and the above-described process is repeated.

  In the above-described example, more candidate points at the lower part of the image are extracted from the upper part (S12, FIGS. 10A and 10B), or the candidate points extracted at the lower part of the image are weighted (S18, FIG. 11A). ) And (B)), and the side with a large number of extractions at the candidate points in the left and right images is used as an index for detecting the approximate straight line on the other side (S19, FIGS. 12A and 12B). As a result, the calculation accuracy of the approximate straight line was improved. In the above-described example, the candidate point search range is set in the vicinity of the previously calculated approximate straight line (S11, FIG. 9) to reduce the candidate point extraction time and suppress the processing amount. All of these examples do not need to be performed in this processing, and any one or two of them may be performed. This is because even in such a case, the approximate straight line is detected while changing the state, so that the processing accuracy can be suppressed and the calculation accuracy can be sufficiently maintained.

  In the example described above, an approximate straight line is obtained by transitioning three states. However, for example, two states may be transited, and an approximate straight line may be obtained while transitioning four or more states. May be. As in the case of the three states, when the number of continuous straight line detections (or the slope or intercept of the approximate straight line obtained in a certain state) reaches a predetermined number (or within a predetermined range), the next state is entered. Thus, it is only necessary to increase the angle resolution and reduce the tilt detection range. In either case, the same operational effects as the example of the three states are achieved.

  Furthermore, in the above-described example, the lane is taken as an example of the target to be detected. In addition, a straight line in an image such as a line indicating a parking range in a parking lot or a road guide line in a facility may be detected. By operating with the same configuration as the above-described example, the same operational effects can be obtained.

It is a figure which shows the structural example of the lane detection apparatus which concerns on this invention. 2A shows an example of a vehicle to which a camera is attached, and FIG. 2B shows an example of an image photographed by the camera. 3A shows an example of a straight line passing through the candidate points, and FIG. 3B shows an example of a curve after the Hough transform. FIG. 4A shows an example of a straight line passing through a plurality of candidate points, and FIG. 4B shows a curve after Hough transform for a plurality of candidate points. It is a figure which shows the state transition in the case of Hough conversion. It is a figure which shows the state transition in the case of Hough conversion. It is an example of the flowchart for detecting an approximate straight line. It is an example of the flowchart for detecting an approximate straight line. It is a figure which shows the extraction range of a candidate point. FIG. 10A shows an example of candidate points, and FIG. 10B shows a curve after the Hough transform of candidate points. FIG. 11A shows an example of candidate points, and FIG. 11B shows a curve after the Hough transform of candidate points. FIG. 12A shows an example of candidate points, and FIGS. 12B and 12C show curves of candidate points after Hough transform.

Explanation of symbols

1: Lane detection device, 2: Vehicle, 3: White line on the left side, 4: White line on the right side, 10: Camera, 20: Microcomputer, 30: Body control unit, θ: Straight line inclination, ρ: Vertical line drawn from the origin to a straight line Length, A1-A3: Candidate points

Claims (14)

In a lane detection device for detecting a lane by Hough transform from candidate points in the extracted image,
While increasing the angle resolution of the straight lines passing through the candidate points and adjacent angles of the respective lines while changing the state, changing the state of the inclination detection range that is the inclination range of the straight line centered on the candidate point Detecting means for obtaining an approximate straight line by reducing the lane by the approximate straight line;
A lane detection device comprising:   When the inclination and intercept of the approximate straight line are within a predetermined range, the detection means increases the angular resolution from the current state, and shifts to the next state where the tilt detection range is small to obtain the approximate straight line. The lane detection device according to claim 1.   2. The lane detection device according to claim 1, wherein the detection means reduces the inclination detection range by limiting the inclination detection range in the current state based on the result of the approximate straight line obtained in the previous state. .   2. The lane detection device according to claim 1, wherein the detection unit weights the candidate points extracted at the lower part of the image larger than the candidate points extracted at the upper part to obtain the approximate straight line.   2. The lane detection device according to claim 1, wherein the detection means obtains the approximate straight line by extracting more candidate points in the lower part of the image than in the upper candidate point.   The detection means divides the image into left and right, and obtains an approximate straight line on the other side with reference to an approximate straight line obtained on the side with a large number of extraction points among the divided left and right images. Item 1. A lane detection device according to item 1.   The lane detection device according to claim 1, wherein the detection unit extracts a vicinity of the approximate straight line obtained in a previous state as the candidate point.   The detecting means, when detecting the approximate straight line a predetermined number of times, obtains the approximate straight line by moving to the next state in which the angle resolution is increased from the current state and the inclination detection range is reduced. The lane detection device according to claim 1.   9. The lane according to claim 1 or 8, wherein the detection means shifts to a state where the angular resolution is the lowest in the state and the inclination detection range is large when the approximate line cannot be detected in the current state. Detection device.   4. The lane detection according to claim 3, wherein the detection means reduces the inclination detection range by setting the inclination detection range within a predetermined range centering on the inclination of the approximate straight line obtained in the previous state. apparatus.   2. The lane detection device according to claim 1, wherein the detection means extracts more candidate points to be extracted at the lower part of the image than at the upper part by taking more lines to scan at the lower part of the image than at the upper part. . In the lane detection method in the lane detection device for detecting the lane by Hough transform from the candidate points in the extracted image,
While increasing the angle resolution of the straight lines passing through the candidate points and adjacent angles of the respective lines while changing the state, changing the state of the inclination detection range that is the inclination range of the straight line centered on the candidate point A lane detection method characterized by obtaining an approximate line by reducing the lane and detecting the lane by the approximate line. In a lane detection program for detecting a lane by Hough transform from candidate points in the extracted image,
While increasing the angle resolution of the straight lines passing through the candidate points and adjacent angles of the respective lines while changing the state, changing the state of the inclination detection range that is the inclination range of the straight line centered on the candidate point A process of obtaining an approximate straight line and detecting the lane by the approximate straight line;
A lane detection program that causes a computer to execute the above. In the straight line detection method in the straight line detection device for detecting a straight line in the image by Hough transform from the candidate points in the extracted image,
While increasing the angle resolution of the straight lines passing through the candidate points and adjacent angles of the respective lines while changing the state, changing the state of the inclination detection range that is the inclination range of the straight line centered on the candidate point A straight line detection method, characterized in that an approximate straight line is obtained by reducing the size and a straight line in the image is detected by the approximate straight line.

Images