WO2018053836A1

WO2018053836A1 - Paired lane line detection method and device

Info

Publication number: WO2018053836A1
Application number: PCT/CN2016/100102
Authority: WO
Inventors: 黄凯明; 韩永刚
Original assignee: 深圳市锐明技术股份有限公司
Priority date: 2016-09-26
Filing date: 2016-09-26
Publication date: 2018-03-29
Also published as: CN106462755A; CN106462755B

Abstract

A paired lane line detection method and device, said method comprising: obtaining an image of a road surface in front of a vehicle; detecting whether there are two straight lines in the image of the road surface in front of the vehicle, the color of the two straight lines being different to that of the road surface in front of the vehicle in the image; when there are two straight lines in the image of the road surface in front of the vehicle, taking N points on each of the two straight lines, and respectively calculating distances between the N points taken on each of the two straight lines and a center point of the image obtained of the road surface in front of the vehicle, wherein N is an integer, and N is greater than or equal to 2; and determining whether the two straight lines are paired lane lines according to the calculated distances between the N points taken on each of the two straight lines and the center point of the image obtained of the road surface in front of the vehicle as well as a preset classifier. By means of the above method, a detection result of lane lines may be obtained faster.

Description

Paired lane line detection method and device

[0001] Embodiments of the present invention belong to the field of image processing, and in particular, to a method and device for detecting a pair of lane lines.

Background technique

[0002] Lane line detection technology is the basic technology for lane departure, collision warning and automatic driving.

[0003] In the existing lane line detection method, by acquiring a real video stream, detecting a suspect lane line in the real video stream, and finally calculating and determining whether the detected suspect lane line is a pair of lane lines, It takes a considerable amount of resources to calculate and judge whether the detected suspect lane line is a pair of lane lines, so it is difficult to obtain the detection result.

technical problem

Embodiments of the present invention provide a method and an apparatus for detecting a pair of lane lines, which are intended to solve the problem that the existing method is difficult to obtain the detection result of the lane line.

Problem solution

Technical solution

[0005] The embodiment of the present invention is implemented as a method for detecting a pair of lane lines, and the method includes: [0006] acquiring a picture of a road surface in front of the vehicle;

[0007] detecting whether there are two straight lines in the picture of the road surface in front of the vehicle, and the colors of the two lines are different from the colors of the road surface in front of the vehicle in the picture;

[0008] There are two straight lines in the picture of the road surface in front of the vehicle, and N points are taken on the two lines

And respectively calculating the distance between the N points taken on the two straight lines and the center point of the acquired picture of the road surface in front of the vehicle, where N is an integer and N is greater than or equal to 2;

[0009] determining whether the two straight lines are paired lane lines according to the calculated distance between the N points taken on the two straight lines and the center point of the acquired picture of the road surface in front of the vehicle and the preset classifier .

[0010] Another object of the embodiments of the present invention is to provide a pair of lane line detecting devices, where the apparatus includes: [0011] a picture acquiring unit, configured to acquire a picture of a road surface in front of the vehicle;

[0012] a line detecting unit, configured to detect whether there are two straight lines in the picture of the road surface in front of the vehicle, The color of the two lines is different from the color of the road ahead of the vehicle in the picture;

[0013] a distance calculating unit, configured to have two straight lines 图片 in the picture of the road surface in front of the vehicle, take N points on the two straight lines, and calculate respectively on each of the two straight lines The distance between the N points and the center point of the picture of the road surface in front of the acquired vehicle, N is an integer, and N is greater than or equal to 2;

[0014] a pair of lane line determining unit, configured to determine, according to the calculated distance between each of the N points taken on the two straight lines and the center point of the acquired picture of the road surface in front of the vehicle, and a preset classifier Whether the two lines are pairs of lane lines.

Advantageous effects of the invention

Beneficial effect

[0015] In the embodiment of the present invention, whether the two straight lines are paired is determined according to the calculated distance between the N points taken on the two straight lines and the center point of the acquired picture of the road surface in front of the vehicle. The lane line, and the distance between the two points is relatively simple to calculate, so the calculation result can be quickly obtained, and the preset classifier can accurately and quickly obtain whether the two lane lines are paired lane lines according to the obtained calculation result. The result of the judgment.

Brief description of the drawing

DRAWINGS

1 is a flowchart of a method for detecting a pair of lane lines according to a first embodiment of the present invention;

2 is a schematic diagram showing two points on two straight lines and connected to the center of the picture according to the first embodiment of the present invention;

3 is a schematic view showing another method of taking two points on two straight lines and connecting with the center of the picture according to the first embodiment of the present invention;

4 is a structural diagram of a pair of lane line detecting devices according to a second embodiment of the present invention.

Embodiments of the invention

The present invention will be further described in detail below with reference to the accompanying drawings and embodiments. It is understood that the specific embodiments described herein are merely illustrative of the invention and are not intended to limit the invention. [0021] In the embodiment of the present invention, a picture of the road surface in front of the vehicle is acquired, and two lines are detected in the picture of the road surface in front of the vehicle, and two straight lines are present in the picture of the road surface in front of the vehicle. N points are respectively taken on the straight line, and the distances between the N points taken on the two straight lines and the center point of the acquired picture of the road surface in front of the vehicle are respectively calculated, according to the calculated two straight lines The distance between the N points taken and the center point of the acquired picture of the road surface in front of the vehicle and the preset classifier determine whether the two lines are paired lane lines.

[0022] In order to explain the technical solution described in the present invention, the following description will be made by way of specific embodiments.

[0023] Embodiment 1:

[0024] FIG. 1 is a flow chart showing a method for detecting a pair of lane lines according to a first embodiment of the present invention, which is described in detail as follows:

[0025] Step Sl l, obtaining a picture of the road surface in front of the vehicle.

Specifically, the road surface in front of the vehicle is photographed by a vehicle monitor or the like to acquire a picture corresponding to the road surface in front of the vehicle. Of course, in order to improve the subsequent lane line detection speed, the acquired picture may be preprocessed after acquiring the picture of the road ahead of the vehicle, for example, converting the color picture into a gray picture, etc., to reduce the memory space occupied by the picture itself.

[0027] Step S12: detecting whether there are two straight lines in the picture of the road surface ahead of the vehicle, and the colors of the two lines are different from the colors of the road surface in front of the vehicle in the picture.

[0028] Wherein the two straight lines here have a certain width on the road surface, for example, the width occupied by the common lane line on the road surface. The color of the two lines is different from the color of the road surface to remind the user. Since the color of the road surface is usually gray or black, in order to improve the discrimination, the color of the two lines is usually white.

[0029] Specifically, it is detected whether two or more types of colors of the acquired road surface on the road ahead of the vehicle are present, and if so, whether the area corresponding to the color occupying the road surface of the road ahead of the acquired vehicle is detected to form two Straight line. It should be pointed out that the straight line here is not necessarily a continuous solid line, but also a continuous dotted line, that is, there are multiple line segments in a straight line.

[0030] Step S13, there are two straight lines 图片 in the picture of the road surface in front of the vehicle, N points are taken on the two straight lines, and N points respectively taken on the two straight lines are respectively calculated. The distance from the center point of the picture of the road surface in front of the acquired vehicle, N is an integer, and N is greater than or equal to 2. [0031] wherein there is a certain distance between the points taken on each straight line, and optionally, the intervals between the N points on the two straight lines are correspondingly equal. Specifically, the center point of the picture of the road surface in front of the vehicle is not necessarily the center point between the two lines, as shown in FIG. When the picture on the road surface in front of the vehicle is a rectangle, the center point of the picture on the road surface in front of the vehicle is the center point of the rectangular area, that is, the intersection of the two diagonal lines of the rectangular area.

[0032] Taking FIG. 3 as an example, the 0 point in the figure is the center point of the picture, and two points are taken on two straight lines: A point and B point are taken on the left line, and C is taken on the right line. Point and D point, calculate the distance between the line AO (upper left line), BO (lower left line), CO (upper right line), and DO (lower right line).

[0033] Step S14, determining whether the two straight lines are formed according to the calculated distance between the N points taken on the two straight lines and the center point of the acquired picture of the road surface in front of the vehicle and the preset classifier. Lane line

[0034] Specifically, when the acquired center point of the picture of the road surface in front of the vehicle is also the center point 两条 between the two straight lines, the N points and the obtained points on the two straight lines can be directly obtained through comparison and calculation. Whether the distances of the center points of the pictures of the road ahead of the vehicle are equal, and comparing the distance between any one of the N points and the center point of the picture of the road surface ahead of the vehicle with the preset lane line width, and then according to the comparison result It is judged whether the two straight lines are paired lane lines, but when the center point of the picture of the road surface ahead of the acquired vehicle is not the center point between the two straight lines, the above judgment result is likely to be erroneous.

[0035] Optionally, in order to improve the accuracy of the detection result of the lane line, the step S14 specifically includes:

[0036] Al, dividing the distances of the N points taken on the two straight lines from the center point of the acquired picture of the road surface in front of the vehicle into two groups, and calculating the distance sum of each group separately. For example, to facilitate grouping, you can set N to an even number.

[0037] A2. Calculating the ratio of the distance sums of the two groups, and replacing the ratio of the distance sums of the two groups with the unknown variables in the preset classifier to obtain a calculation result.

[0038] A3. Determine, according to the calculation result, whether the two straight lines are paired lane lines.

[0039] Specifically, when the calculation result is the same as the preset result of identifying the pair of lane lines, it is determined that the two straight lines are paired lane lines, otherwise, it is determined that the two lines are not paired lane lines.

[0040] Optionally, when N=2吋, the two points taken on the two straight lines form a line with the line connecting the center point of the picture of the road surface in front of the vehicle: upper left line, lower left line, right Up and down lines, this line, the A1 Specifically include:

[0041] Al l, dividing the upper left line and the upper right line into upper side point groups, and calculating the distance sum of the upper side point groups.

[0042] A12. Divide the lower left line and the lower right line into lower side point groups, and calculate a distance sum of the lower side point groups.

Correspondingly, the A2 specifically includes: calculating a ratio of a distance of the lower side point group and a distance sum of the upper side point group, and replacing the ratio of the distance sum of the two groups with a preset classification. The unknown variable in the device gets a calculation result.

[0044] Optionally, the preset classifier in the embodiment of the present invention is determined by:

[0045] Bl, acquiring a sample image including a pair of lane lines and an unpaired lane line. Wherein, the acquired sample image should include a large number of images of paired lane lines and unpaired lane lines.

[0046] B2, taking N points on each lane line, and calculating the distance between each of the N points taken on each lane line and the center point of the sample image, where N is an integer and N is greater than or equal to 2 . It should be pointed out that

, the spacing of the points on the daily lane line remains equal.

[0047] B3. Divide the distances between the N points taken on each lane line in the same sample image and the center point of the sample image into two groups, and calculate the distance sum of each group separately.

[0048] B4. Calculate a ratio of distance sums of the two groups as a sample ratio. For example, at N = 2 吋, the ratio of the distance of the lower point group to the distance sum with the upper point group can be used as the sample ratio.

[0049] B5. Determine a classifier according to the sample ratio.

[0050] Further, the classifier is determined by an Adaboost iterative algorithm. The core idea of Adaboost is to train a weak classifier (class 2 classifier) under the initial weight data distribution, and then use this weak classifier to judge the accuracy rate. For those wrong judgments (that is, the original label is 1 0, or vice versa, the weight of the sample is increased, and for the sample with the correct classification, the weight is reduced, so that the sample that is misclassified is highlighted, and the sample that is misclassified will be considered more in the next training. So get a new sample distribution (sample weights have been updated). Under the new distribution, training is performed to obtain a weak classifier, and N weak detectors with general detection capabilities are obtained. Then, through a certain algorithm, the general classifiers of these detection capabilities are combined to obtain a strong classifier with strong classification ability.

[0051] Optionally, the B5 specifically includes:

[0052] B501. Set a first classification threshold for determining whether the two lane lines are paired lane lines according to the sample ratio. [0053] B502. Determine a first basic classifier according to the first classification threshold.

[0054] B503, determining, according to the first classification threshold, a preset first weight corresponding to the sample ratio, and whether the two lane lines corresponding to the sample ratio are paired lane lines, determining the first basic classifier Error rate

[0055] B504. Determine a coefficient of the first basic classifier according to an error rate of the first basic classifier, and determine a first weak classifier according to the coefficient of the first basic classifier and the first basic classifier.

[0056] B505. Determine whether the number of misclassification points of the first weak classifier is 0.

[0057] B506, the misclassification point of the first basic classifier is not 0, according to a preset first weight corresponding to the sample ratio, a coefficient of the first weak classifier, and the sample ratio The conclusion of whether the two lane lines are paired lane lines and the first basic classifier determine the first normalization factor.

[0058] B507. Adjust the preset first weight corresponding to the sample ratio according to the first normalization factor, and obtain a second weight corresponding to the sample ratio.

[0059] B508. Set a second classification threshold for determining whether the two lane lines are paired lane lines according to the sample ratio.

[0060] B509. Determine a second basic classifier according to the second classification threshold.

[0061] B510, determining, according to the second classification threshold, the second weight corresponding to the sample ratio, and whether the two lane lines corresponding to the sample ratio are paired lane lines, determining the second basic classifier Error rate.

[0062] B511. Determine a coefficient of the second basic classifier according to an error rate of the second basic classifier, where the second weak classifier is based on a coefficient of the first basic classifier, the first basic classifier The coefficients of the second basic classifier and the second basic classifier are determined.

[0063] B512, determining whether the number of misclassification points of the second weak classifier is 0, if it is 0, using the second weak classifier as a final classifier, otherwise, continuing to determine a new weak classifier, until The number of misclassification points of the determined new weak classifier is zero.

[0064] In order to more clearly describe how to determine the process of the classifier, a specific application example will be described below: [0065] In this example, two points are taken on each lane line, and the resulting lower point is assumed. The ratio of the distance of the group and the distance sum with the upper point group is shown in Table 1:

[0066] Table 1:

[] [Table 1]

[0067]

[0068] (1) First solve the first weak classifier sign[fl(x)] _:

[0069] The preset first weight distribution corresponding to the sample ratio is as follows:

[0070] D1 = (W11, W12, W13, W14, W15)

[0071] Wli = 0.2; i= 1, 2, 3, 4, 5

[0072] According to Table 1, when the first classification threshold for determining whether the two lane lines are paired lane lines is 1.43 吋, the classification error rate is the lowest, and then, the first basic classifier G1 is:

Gl(x)= 1, x> 1.43; Gl(x)=-1, x< 1.43

[0074] As can be seen from Table 1, the first basic classifier G1 has an error classification result (ie, the classification result is wrong for the sequence number 4), and the data weight distribution of the error classification is Gl(x) in the training data set. The error rate above (ie the error rate of the first basic classifier): el = P(Gl(xi) !=yi) = 0.2.

[0075] The coefficient of Gl(x) is determined by the following equation: al = l/21n((l-el)/el) = 1η2 = 0.6931.

[0076] Then the first weak classifier is: sign[fl(x)] = ln2Gl(x).

[0077] (2) Solving the second weak classifier sign[f2(x)] _:

[0078] Since the first basic classifier G1(x) has an error classification result, according to the formula of the first weak classifier, the first weak classifier is in the training data set (ie, the data set corresponding to Table 1). There is also a misclassification point. Therefore, the first normalization factor needs to be determined by:

[0079] The first normalization factor Zl = Wll* exp(-al * yl * Gl(xl)) + W12* exp(-al * y2 * Gl(x2))

+ W13* exp(-al * y3 * Gl(x3)) + W14* exp(-al * y4 * Gl(x4)) + W15* exp(-al * y5

* Gl(x5)) = 0.8.

[0080] adjusting the first weight according to the first normalization factor, and the obtained second weight distribution is as follows:

[0081] D2 = (W21, W22, W23, W24, W25)

W2i = wli/Zl * exp(-al*yi*Gl(Xi)), i= l, 2, 3, 4, 5 D2 = (0.125, 0.125, 0.125, 0. 5, 0.125)

[0084] In combination with the data of Table 1, a second classification threshold for determining whether two lane lines are paired lane lines is set

When the second classification threshold is 1.55吋, the classification error rate is the lowest, and the second basic classifier G2 is:

G2(x)= 1 , x > 1.55; G2(x)= -1 , x < 1.55

[0086] The second basic classifier G2(x) has two misclassification results (ie, the results corresponding to the sequence numbers 2 and 3 of Table 1), and the data weight distributions of the two misclassifications are G2(x) in the training data. The error rate on the set: e2 = P(Gl(xi) != yi) = 0.125 + 0.125 = 0.25.

[0087] The coefficient of G2(x): a2 = l/21n((l-e2)/e2) =1η3 ^1/3 = 0.5493.

[0088] Then the second weak classifier sign[f2(x)]= al * Gl(x) + a2 * G2(x) = ln2 * Gl(x) + In * G2(x) =

0.6931 * Gl(x) + 0.5493 * G2(x).

[0089] According to the formula of the second weak classifier sign[f2(x)], sign[f2(x)] has 0 misclassification points on the training data set (when "sign[f2( _X )] "The calculation result is greater than -1 and less than 0吋, and the "sign[f2( _X )]" is rounded to -1. Similarly, when the calculation result of "sign[f2(x)]" is greater than 0 and less than 1吋, the "sign[f2(x)]" is rounded to 1).

[0090] Of course, if it is determined that the number of misclassification points of the second weak classifier is not 0, it is necessary to determine a new weak classifier (third classifier, fourth classifier, etc.) until the determined new The number of misclassification points of the weak classifier is zero.

[0091] Before determining the new weak classifier (third classifier), it is necessary to determine the second normalization factor Z2 =

W21* exp(-al * yl * G2(xl)) + W22* exp(-al * y2 * G2(x2)) + W23* exp(-al * y3 * G2(x3)) + W24* exp(- Al * y4 * G2(x4)) + W25* exp(-al * y5 * G2(x5)) = 1/2 * ln3 ^1/3

[0092] The weight distribution of the training data updated according to the second normalization factor is as follows:

[0093] D3 = (W31, W32, W33, W34, W35)

W3i = w2i/Z2 * exp(-a2*yi*G2(Xi)) , i = l, 2, 3, 4, 5

[0095] D3 = (l/12, 1/4, 1/4, 1/3, 01/12)

[0096] In step S14, according to the preset classifier, it can quickly determine whether the two lines are paired lane lines, for example, assuming that the preset classifier is the second weak classifier sign[f2(x)] = al * Gl(x) + a2 * G2(x) = 1η2 * Gl(x) + In * G2(x) = 0.6931 * Gl(x) + 0.5493 * G2(x) , when the lower point group is judged The distance between the distance and the distance from the upper point group is 1.4403吋, which is substituted into sign[f2(x)],

Sign[0.697 * Gl( 1.4403) + 0.5493 * G2( 1.4403)] = sign[0.6931 * 1 + 0.5493 * (-1)]

= 1, get the judgment that "two lines are paired lane lines".

[0098] When it is determined that the ratio of the distance of the lower side point group and the distance sum of the upper side point group is 1.1666 吋, substituting s ign[f2(x)],

[0099] G(x) = sign[0.6931 * Gl(1.1666) + 0.5493 * G2(1.1666)] = sign[0.6931 * (-1) + 0.5493 * (-1)] = -1, get "two lines The judgment of the unpaired lane line.

[0100] In the first embodiment of the present invention, a picture of the road surface in front of the vehicle is acquired, and two lines are detected in the picture on the road surface in front of the vehicle, and two straight lines are present in the picture on the road surface in front of the vehicle. N points are taken on each of the two straight lines, and the distances between the N points taken on the two straight lines and the center point of the acquired picture of the road surface in front of the vehicle are respectively calculated, according to the calculated two straight lines The distance between each of the N points taken and the center point of the acquired picture of the road surface in front of the vehicle and the preset classifier determine whether the two lines are paired lane lines. It is determined whether the two straight lines are paired lane lines and the distance between the two points is determined according to the calculated distance between the N points taken on the two straight lines and the center point of the acquired picture of the road surface in front of the vehicle. The calculation is relatively simple, so that the calculation result can be obtained quickly, and the preset classifier can accurately and quickly obtain the determination result of whether the two lane lines are paired lane lines according to the obtained calculation result.

It should be understood that, in the embodiment of the present invention, the size of the sequence numbers of the foregoing processes does not mean the order of execution sequence, and the execution order of each process should be determined by its function and internal logic, and should not be taken to the embodiment of the present invention. The implementation process constitutes any limitation.

Embodiment 2:

4 is a structural diagram of a pair of lane line detecting devices according to a second embodiment of the present invention. The paired lane line detecting device can be used in various smart terminals, including a mobile phone and a vehicle. Equipment, etc. For the convenience of description, only parts related to the embodiment of the present invention are shown.

The paired lane line detecting device includes: a picture acquiring unit 41, a line detecting unit 42, a distance calculating unit 43, and a pair of lane line determining unit 44. among them:

[0105] The picture obtaining unit 41 is configured to acquire a picture of the road surface in front of the vehicle.

[0106] Specifically, the road surface in front of the vehicle is photographed by a vehicle monitor or the like to obtain a map corresponding to the road surface in front of the vehicle. Film. Of course, in order to improve the subsequent lane line detection speed, the acquired picture may be preprocessed after acquiring the picture of the road ahead of the vehicle, for example, converting the color picture into a gray picture, etc., to reduce the memory space occupied by the picture itself.

[0107] The line detecting unit 42 is configured to detect whether there are two straight lines in the picture of the road surface in front of the vehicle, and the colors of the two lines are different from the colors of the road surface in front of the vehicle in the picture.

[0108] Wherein the two straight lines here have a certain width on the road surface, for example, the width occupied by the common lane line on the road surface. The color of the two lines is different from the color of the road surface to remind the user. Since the color of the road surface is usually gray or black, in order to improve the discrimination, the color of the two lines is usually white.

Specifically, it is detected whether two or more types of colors of the acquired road surface on the road ahead of the vehicle are present, and if so, whether the area corresponding to the color occupying the road surface of the road ahead of the acquired vehicle is detected is formed separately. Straight line. It should be pointed out that the straight line here is not necessarily a continuous solid line, but also a continuous dotted line, that is, there are multiple line segments in a straight line.

[0110] The distance calculating unit 43 is configured to have two straight lines in the picture of the road surface in front of the vehicle, and take N points on the two straight lines, and calculate respectively on the two straight lines. The distance between the N points and the center point of the picture of the road surface in front of the acquired vehicle, N is an integer, and N is greater than or equal to 2.

[0111] Optionally, the center point of the picture of the road surface ahead of the vehicle is not necessarily the center point between the two lines.

[0112] The paired lane line determining unit 44 is configured to determine, according to the calculated distance between the N points taken on the two straight lines and the center point of the acquired picture of the road surface in front of the vehicle, and a preset classifier Whether the two lines are paired lane lines.

[0113] Optionally, the pair of lane line determining unit 44 includes:

[0114] a distance and calculation module, configured to divide the distances of the N points taken on the two straight lines from the center point of the acquired picture of the road surface in front of the vehicle into two groups, and calculate the distance and the distance of each group respectively. .

[0115] The classifier calculation result determining module is configured to calculate a ratio of the distance sums of the two groups, and replace the ratio of the distances of the two groups with the unknown variable in the preset classifier to obtain a calculation result.

[0116] The classifier calculation result comparison module is configured to determine, according to the calculation result, whether the two straight lines are paired lane lines.

[0117] Optionally, when N=2吋, two points taken on two straight lines and a picture of the road surface in front of the acquired vehicle The line connecting the center points forms 4 lines: the upper left line, the lower left line, the upper right line, and the lower right line. Here, the distance and calculation module includes:

[0118] The distance and calculation module of the upper side point group is configured to divide the upper left line and the upper right line into upper side point groups, and calculate the distance sum of the upper side point groups.

[0119] The distance and calculation module of the lower point group is configured to divide the lower left line and the lower right line into lower side point groups, and calculate the distance sum of the lower side point groups.

Correspondingly, the classifier calculation result determining module is specifically configured to calculate a ratio of a distance between the lower side point group and a distance sum of the upper side point group, and a ratio of the distance sums of the two groups Replace the unknown variable in the preset classifier to get a calculation result.

[0121] Optionally, the paired lane line detecting device determines a preset classifier by using the following unit:

[0122] The sample image obtaining unit is configured to acquire a sample image including a pair of lane lines and an unpaired lane line.

[0123] The distance data acquiring unit is configured to take N points on each lane line, and calculate a distance between each of the N points taken on each lane line and the center point of the sample image, where N is an integer. N is greater than or equal to 2.

[0124] different sets of distance and calculation units are used to divide the distances of N points taken on each lane line in the same sample image from the center point of the sample image into two groups, and calculate each group separately. Distance and.

[0125] A sample ratio calculating unit is configured to calculate a ratio of distance sums of the two groups as a sample ratio.

[0126] A classifier determining unit, configured to determine a classifier according to the sample ratio.

[0127] When the classifier 确定 is determined by an Adaboost iterative algorithm, the classifier determining unit includes:

[0128] The first weak classifier determining module is configured to set, according to the sample ratio, a first classification threshold for determining whether the two lane lines are paired lane lines. Determining a first basic classifier based on the first classification threshold. And determining, according to the first classification threshold, the preset first weight corresponding to the sample ratio, and whether the two lane lines corresponding to the sample ratio are paired lane lines, determining an error rate of the first basic classifier. A coefficient of the first basic classifier is determined according to an error rate of the first basic classifier, and a first weak classifier is determined according to a coefficient of the first basic classifier and the first basic classifier.

[0129] The misclassification point number judging module of the first weak classifier is configured to determine whether the number of misclassification points of the first weak classifier is 0.

[0130] The first normalization factor determining module is configured to: when the misclassification point of the first basic classifier is not 0, according to the preset first weight corresponding to the sample ratio, the coefficient of the first weak classifier Sample ratio The conclusion of whether the corresponding two lane lines are paired lane lines and the first basic classifier determine the first normalization factor.

[0131] The second weak classifier determining module is configured to adjust the preset first weight corresponding to the sample ratio according to the first normalization factor to obtain a second weight corresponding to the sample ratio. A second classification threshold for determining whether the two lane lines are paired lane lines is set based on the sample ratio. A second basic classifier is determined based on the second classification threshold. And determining, according to the second classification threshold, the second weight corresponding to the sample ratio, and whether the two lane lines corresponding to the sample ratio are paired lane lines, determining an error rate of the second basic classifier. Determining coefficients of the second basic classifier according to an error rate of the second basic classifier, according to coefficients of the first basic classifier, coefficients of the first basic classifier, the second basic classifier, and The second basic classifier determines the second weak classifier.

[0132] The misclassification point number judging module of the second weak classifier is configured to determine whether the number of misclassification points of the second weak classifier is 0, and if it is 0, the second weak classifier is used as a final The classifier, otherwise, continues to determine the new weak classifier until the determined number of misclassification points for the new weak classifier is zero.

[0133] In the second embodiment of the present invention, it is determined whether the two straight lines are formed due to the calculated distance between the N points taken on the two straight lines and the center point of the acquired picture of the road surface in front of the vehicle. For the lane line, the distance between the two points is relatively simple to calculate, so the calculation result can be quickly obtained, and the preset classifier can accurately and quickly obtain whether the two lane lines are paired lanes according to the obtained calculation result. The result of the line determination.

[0134] Those of ordinary skill in the art will appreciate that the elements and algorithm steps of the various examples described in connection with the embodiments disclosed herein can be implemented in electronic hardware, or a combination of computer software and electronic hardware. Whether these functions are performed in hardware or software depends on the specific application and design constraints of the solution. A person skilled in the art can use different methods for implementing the described functions for each particular application, but such implementation should not be considered to be beyond the scope of the present invention.

[0135] It is to be understood by those skilled in the art that, for the convenience and brevity of the description, the specific working process of the system, the device and the unit described above may refer to the corresponding process in the foregoing method embodiments, and details are not described herein again. .

[0136] In the several embodiments provided by the present application, it should be understood that the disclosed system, apparatus, and method may be implemented in other manners. For example, the device embodiments described above are merely illustrative, For example, the division of the unit is only a logical function division, and the actual implementation may have another division manner, for example, multiple units or components may be combined or may be integrated into another system, or some features may be ignored, or not. carried out. In addition, the mutual coupling or direct coupling or communication connection shown or discussed may be an indirect coupling or communication connection through some interface, device or unit, and may be in an electrical, mechanical or other form.

[0137] The unit described as a separate component may or may not be physically distributed, and the component displayed as a unit may or may not be a physical unit, that is, may be located in one place, or may be distributed to multiple On the network unit. Some or all of the units may be selected according to actual needs to achieve the purpose of the solution of the embodiment.

[0138] In addition, each functional unit in each embodiment of the present invention may be integrated into one processing unit, or each unit may exist physically separately, or two or more units may be integrated into one unit.

[0139] The functions, if implemented in the form of software functional units and sold or used as separate products, may be stored in a computer readable storage medium. Based on such understanding, the technical solution of the present invention, which is essential or contributes to the prior art, or a part of the technical solution, may be embodied in the form of a software product, which is stored in a storage medium, including The instructions are used to cause a computer device (which may be a personal computer, server, or network device, etc.) to perform all or part of the steps of the methods described in various embodiments of the present invention. The foregoing storage medium includes: a USB flash drive, a removable hard disk, a read only memory (ROM, Read-Only)

Memory, random access memory (RAM), disk or optical disk, and other media that can store program code.

The above description is only a specific embodiment of the present invention, but the scope of the present invention is not limited thereto, and any person skilled in the art can easily think of changes or within the technical scope disclosed by the present invention. Alternatives are intended to be covered by the scope of the present invention. Therefore, the scope of the invention should be determined by the scope of the claims.

Claims

Claim

[Claim 1] A method for detecting a pair of lane lines, the method comprising:

Obtain a picture of the road surface in front of the vehicle;

Detecting whether there are two straight lines in the picture of the road surface in front of the vehicle, and the colors of the two lines are different from the colors of the road surface in front of the vehicle in the picture;

There are two straight lines in the picture of the road surface in front of the vehicle, N points are taken on the two straight lines, and N points taken on the two straight lines and the obtained road surface in front of the vehicle are respectively calculated. The distance from the center point of the picture, N is an integer, N is greater than or equal to 2;

According to the calculated distance between the N points taken on the two straight lines and the center point of the acquired picture of the road surface in front of the vehicle and the preset classifier, it is judged whether the two straight lines are paired lane lines.

[Claim 2] The method according to claim 1, wherein the calculated distances between the N points taken on the two straight lines and the center point of the acquired picture of the road surface in front of the vehicle and The preset classifier determines whether the two straight lines are paired lane lines, and specifically includes: dividing the distance between the N points taken on the two straight lines and the center point of the acquired picture of the road surface in front of the vehicle into two Group, and calculate the distance sum of each group separately;

Calculating the ratio of the distances of the two groups, and replacing the ratio of the distances of the two groups with the unknown variables in the preset classifier to obtain a calculation result;

Whether the two straight lines are paired lane lines is determined according to the calculation result.

[Claim 3] The method according to claim 2, wherein, when N = 2 吋, the two points taken on the two straight lines are connected with the line connecting the center points of the pictures of the road surface in front of the vehicle. 4 lines: the upper left line, the lower left line, the upper right line, and the lower right line, where the distance between the N points taken on the two straight lines and the center point of the acquired picture of the road surface in front of the vehicle is divided into Two groups, and calculate the distance and the distance of each group separately, including:

The upper left line and the upper right line are divided into an upper point group, and the distance between the upper side point group is calculated; the lower left line and the lower right line are divided into lower side point groups, and the distance between the lower side point group is calculated; And calculating the ratio of the sum of the distances of the two groups, and replacing the ratio of the distances of the two groups with the unknown variables in the preset classifier to obtain a calculation result, which specifically includes: Calculating a ratio of the distance of the lower side point group and the distance sum of the upper side point group, and replacing the ratio of the distance sums of the two groups with the unknown variable in the preset classifier to obtain a calculation result.

[Claim 4] The method according to claim 2 or 3, wherein the preset classifier is determined by:

Obtaining a sample image including a pair of lane lines and an unpaired lane line; taking N points on each lane line, and respectively calculating N points taken on each lane line and a center point of the sample image Distance, N is an integer, N is greater than or equal to 2;

The distance between the N points taken on each lane line in the same sample image and the center point of the sample image is divided into two groups, and the distance sum of each group is calculated respectively; the ratio of the distance sum of the two groups is calculated as a sample Ratio; determining the classifier based on the sample ratio.

[Claim 5] The method according to claim 4, wherein the determining the classifier according to the sample ratio comprises:

Setting a first classification threshold for determining whether the two lane lines are paired lane lines according to the sample ratio;

Determining, according to the first classification threshold, a first basic classifier;

Determining an error rate of the first basic classifier according to the first classification threshold, the preset first weight corresponding to the sample ratio, and whether the two lane lines corresponding to the sample ratio are paired lane lines;

Determining a coefficient of the first basic classifier according to an error rate of the first basic classifier, determining a first weak classifier according to the coefficient of the first basic classifier and the first basic classifier;

Determining whether the number of misclassification points of the first weak classifier is 0;

If the misclassification point of the first basic classifier is not 0, according to the preset first weight corresponding to the sample ratio, the coefficient of the first weak classifier, and the two lane lines corresponding to the sample ratio Determining a first normalization factor for the conclusion of the pair of lane lines and the first basic classifier; Adjusting the preset first weight corresponding to the sample ratio according to the first normalization factor, to obtain a second weight corresponding to the sample ratio;

Setting a second classification threshold for determining whether the two lane lines are paired lane lines according to the sample ratio;

Determining a second basic classifier according to the second classification threshold;

Determining an error rate of the second basic classifier according to the second classification threshold, the second weight corresponding to the sample ratio, and whether the two lane lines corresponding to the sample ratio are paired lane lines;

Determining coefficients of the second basic classifier according to an error rate of the second basic classifier, according to coefficients of the first basic classifier, coefficients of the first basic classifier, the second basic classifier, and Describe a second basic classifier to determine a second weak classifier;

Determining whether the number of misclassification points of the second weak classifier is 0. If it is 0, the second weak classifier is used as the final classifier. Otherwise, the new weak classifier is continuously determined until the determined new weak class is determined. The number of misclassification points of the classifier is zero.

[Claim 6] A paired lane line detecting device, wherein the device comprises:

a picture obtaining unit, configured to obtain a picture of a road surface in front of the vehicle; a line detecting unit, configured to detect whether there are two straight lines in the picture of the road surface in front of the vehicle, and the colors of the two lines are different from the color of the road surface in front of the vehicle in the picture a distance calculating unit, configured to have two straight lines 图片 in the picture of the road surface in front of the vehicle, take N points on the two straight lines, and calculate N pieces respectively on the two straight lines The distance between the point and the center point of the picture of the road surface in front of the vehicle, where N is an integer and N is greater than or equal to 2;

a pair of lane line determining unit, configured to determine the two straight lines according to the calculated distance between the N points taken on the two straight lines and the center point of the acquired picture of the road surface in front of the vehicle, and a preset classifier Whether it is a pair of lane lines.

[Claim 7] The device according to claim 6, wherein the pair of lane line determining units comprises:

a distance and calculation module for N points taken on the two straight lines and the acquired vehicle The distance between the center points of the pictures of the road ahead is divided into two groups, and the distance sum of each group is calculated separately;

a classifier calculation result determining module, configured to calculate a ratio of the distance sums of the two groups, and replacing the ratio of the distances of the two groups with the unknown variables in the preset classifier, to obtain a calculation result classifier calculation result comparison module, And determining whether the two straight lines are paired lane lines according to the calculation result.

[Claim 8] The apparatus according to claim 7, wherein when N = 2 吋, the two points taken on the two straight lines are connected with the line connecting the center points of the pictures of the road surface in front of the vehicle. 4 lines: the upper left line, the lower left line, the upper right line, and the lower right line. Here, the distance and calculation module includes:

a distance and calculation module of the upper side point group, configured to divide the upper left line and the upper right line into an upper side point group, and calculate a distance sum of the upper side point group;

a distance and calculation module of the lower point group, configured to divide the lower left line and the lower right line into lower side point groups, and calculate a distance between the lower side point groups;

Correspondingly, the classifier calculation result determining module is specifically configured to calculate a ratio of a distance between the lower side point group and a distance sum of the upper side point group, and replace the ratio of the distance sum of the two groups to a preset The unknown variable in the classifier gets a calculation result.

[Claim 9] The device according to claim 7 or 8, wherein the device comprises:

a sample image obtaining unit, configured to acquire a sample image including a pair of lane lines and an unpaired lane line;

The distance data acquisition unit is configured to take N points on each lane line, and calculate the distance between each of the N points taken on each lane line and the center point of the sample image, where N is an integer and N is greater than or Equal to 2;

Different sets of distances and calculation units are used to divide the distances of N points taken on each lane line in the same sample image from the center point of the sample image into two groups, and calculate the distance sum of each group separately;

a sample ratio calculating unit, configured to calculate a ratio of distance sums of the two groups as a sample ratio; A classifier determining unit, configured to determine a classifier according to the sample ratio.

[Claim 10] The device according to claim 9, wherein the classifier determining unit includes a first weak classifier determining module, configured to determine, according to the sample ratio, whether two lane lines are a first classification threshold of the pair of lane lines; determining, according to the first classification threshold, a first basic classifier; corresponding to the first classification threshold, the preset first weight corresponding to the sample ratio, and the sample ratio Determining an error rate of the first basic classifier based on whether the two lane lines are pairs of lane lines; determining coefficients of the first basic classifier according to an error rate of the first basic classifier, according to the first basic classification The coefficient of the device and the first basic classifier determine the first weak classifier;

a misclassification point number judging module of the first weak classifier, configured to determine whether the number of misclassification points of the first weak classifier is 0;

a first normalization factor determining module, configured to: when the misclassification point of the first basic classifier is not 0, according to a preset first weight corresponding to the sample ratio, a coefficient of the first weak classifier, Whether the two lane lines corresponding to the sample ratio are the pair of lane lines and the first basic classifier determines the first normalization factor;

a second weak classifier determining module, configured to adjust the preset first weight corresponding to the sample ratio according to the first normalization factor, to obtain a second weight corresponding to the sample ratio; and set according to the sample ratio a second classification threshold for determining whether the two lane lines are a pair of lane lines; determining a second basic classifier according to the second classification threshold; and determining a second weight according to the second classification threshold and the sample ratio Determining, by a conclusion of whether the two lane lines corresponding to the sample ratio are paired lane lines, determining an error rate of the second basic classifier; determining a coefficient of the second basic classifier according to an error rate of the second basic classifier Determining, according to the coefficients of the first basic classifier, the first basic classifier, the coefficients of the second basic classifier, and the second basic classifier, a second weak classifier;

a misclassification point number judging module of the second weak classifier, configured to determine whether the number of misclassification points of the second weak classifier is 0, and if 0, use the second weak classifier as a final classifier, Otherwise, continue to determine the new weak classifier until the identified new weak classifier misclassification The number of points is 0.