Vehicle low beam light pattern feature point detection method and system based on point symmetry

Abstract

The present invention provides a method and system for detecting characteristic points of a low-beam light type of an automobile based on point symmetry, and relates to the field of image detection. The detection method comprises: performing smoothing and filtering processing on an input initial image I1 to obtain an image I2; performing an open operation on the image I2 to obtain an image I3; performing binarization processing on the image I3 to obtain an image I4; processing the image I4 to obtain a contour point set Q1 of the low-beam light type, and drawing the connecting lines of the contour point set Q1 and its adjacent points in the image I5; comparing the pixel values ​​of the image I5 with the set values ​​column by column from left to right and point by point from top to bottom to obtain a point sequence Q2; starting from the nth point, the point sequence Q2 obtains the symmetry of each point in turn, and stores it in a sequence D1; comparing the values ​​in the sequence D1, the peak value is the characteristic point and corner point of the low-beam light type. Compared with the traditional corner point detection algorithm, the method has strong anti-interference ability and rotation invariance, and can be better applied to the rapid optical detection of automobile lamps.

Description

Technical Field

[0001] The present invention relates to the field of image detection, and in particular to a method and system for detecting feature points of a low-beam light pattern of an automobile based on point symmetry. Background Art

[0002] Car lights play an important role in road lighting, driving status, vehicle appearance and other aspects. Therefore, car light detection is also an important part of car factory inspection. Car light detection mainly detects the light distribution performance of car headlights and the functional status of the whole vehicle lights. Among them, car headlight low beam detection occupies an important position, and its main detection objects are feature points and light and dark cutoff lines. The fast and accurate positioning of feature points and light and dark cutoff lines can provide a good basic condition for the fast detection algorithm of low beam lights.

[0003] At present, most of the proposed low beam detection algorithms first extract the complete cutoff line, and then detect the feature points of the cutoff line. The edge points of the cutoff line are determined according to the maximum gradient of each column of the image, the cutoff line is determined by Hough line detection, and then the feature point positions are extracted according to the geometric position relationship of the cutoff line. From the results, this method needs to judge the rationality of the feature points, exposing the problem of low accuracy of the feature point detection results. The edge points of the cutoff line can be determined by morphological processing methods and background matching methods, and the edge points are divided into horizontal cutoff line edge point sets and inclined cutoff line edge point sets. The fitted straight line of the cutoff line is obtained by straight line fitting, and the cutoff line feature points are obtained by the intersection of the fitted straight lines. In this method, there is a lack of reliable basis for the judgment of the demarcation point, and there is also the problem of low accuracy of the detection result. For the low beam cutoff line with fuzzy characteristics, the improved fuzzy enhancement algorithm has a more positive effect on the feature extraction of the cutoff line. However, the feature point detection results still have the problem of low accuracy, and the algorithm structure is relatively complex, the generalization is poor, and it cannot be well applied to actual production. Summary of the invention

[0004] Purpose of the invention: On the basis of traditional image processing and in combination with the characteristics of the low beam light pattern, a method for detecting feature points of a low beam light pattern of an automobile based on point symmetry is proposed, and further a system for detecting feature points of a low beam light pattern of an automobile for implementing the above detection method is proposed, so as to solve the above problems raised by the prior art.

[0005] Firstly, a method for detecting feature points of a vehicle low beam pattern based on point symmetry is proposed, wherein the feature point refers to the intersection of a left horizontal cut-off line and an inclined cut-off line.

[0006] The method for detecting characteristic points of a low-beam light pattern of an automobile comprises the following steps:

[0007] S1, inputting an initial image I1 of a low beam light pattern, performing a smoothing filter process on the initial image I1 of the low beam light pattern, and obtaining an image I2;

[0008] S2, performing an opening operation on the image I2 to further smooth the image, eliminating burrs on the edge of the image, and obtaining an image I3;

[0009] S3, performing binarization processing on the image I3 to obtain an image I4;

[0010] S4, using the Canny operator to process the image I4 to obtain a contour point set Q1 of the low beam light pattern; drawing the contour point set Q1 of the low beam light pattern and the connecting lines of its adjacent points in a blank image to obtain an image I5;

[0011] S5, comparing the pixel values ​​of the image I5 with the set values ​​from left to right, column by column, and from top to bottom, point by point. When it is detected that the pixel value of a point in a column is equal to the set value, it is determined that the point meets the condition, and all points that meet the comparison condition form a point column Q2;

[0012] S6, starting from the nth point of the point sequence Q2, the symmetry of each point is obtained in sequence and stored in the number sequence D1; the value of n depends on the specific image resolution and the height difference of the feature point;

[0013] S7. Compare the values ​​in the number series D1, and the peak values ​​are the characteristic points and corner points of the low beam light pattern of the vehicle.

[0014] In a further embodiment of the first aspect, the smoothing filtering process in step S1 is mean filtering and / or Gaussian filtering.

[0015] In a further embodiment of the first aspect, the smoothing filtering process in step S1 is side window mean filtering and / or side window Gaussian filtering, which further improves the edge preservation capability of the algorithm.

[0016] The image I5 has the same size as the initial image I1.

[0017] In a further embodiment of the first aspect, the pixel value of each point in the contour point set Q1 in step S4 is set to V1, and the pixel value N of each point in the image I5 is expressed as:

[0018]

[0019] Where P ij is the point with coordinates (i, j) in the point set Q1; V1 is an integer between 1 and 255.

[0020] In a further embodiment of the first aspect, the set value in step S5 is equal to V1.

[0021] In a further embodiment of the first aspect, starting from the mth point of the point sequence Q2, the symmetry of each point is obtained in sequence, and the solution formula of the symmetry is as follows:

[0022]

[0023] Where, d m Indicates the symmetry of the mth point; Represents the horizontal coordinate of the point after the ms-th point is rotated about the m-th point; Represents the horizontal coordinate of the m+sth point; Represents the ordinate of the point after the ms-th point is rotated about the m-th point; It represents the horizontal coordinate of the m+sth point; n is the distance between the points, that is, the number of points.

[0024] Taking n as 5 and the 10th point in the point sequence Q2 as an example, the 5th to 9th points in the point sequence are rotated 180° with the 10th point as the center, and the Euclidean distances of the rotated points are calculated with the 11th to 15th points respectively, and finally the sum is taken to get the symmetry of the 10th point.

[0025] In a further embodiment of the first aspect, the solution formula for the symmetry is as follows:

[0026]

[0027] Where, d m Indicates the symmetry of the mth point; Represents the horizontal coordinate of the point after the ms-th point is rotated about the m-th point; Represents the horizontal coordinate of the m+sth point; Represents the ordinate of the point after the ms-th point is rotated about the m-th point; Indicates the horizontal coordinate of the m+sth point; n is the number of points;

[0028] is the weighting coefficient, ;

[0029] in, and They satisfy the following relationships respectively:

[0030]

[0031]

[0032] In a further embodiment of the first aspect, step S7 further comprises:

[0033] S7-1, starting from the L+1th number in the sequence D1, compare the L+1th value with the previous L values ​​and the next L values ​​in sequence, and if and only if the value of the L+1th point is the maximum value among these points, store the L+1th point in the sequence D2;

[0034] S7-2, sequentially perform the operation of step S7-1 on the subsequent numbers of sequence D1 until the L1-Lth number, where L is the calculated window length, and L1 is the length of sequence D1, that is, the number of numbers included in sequence D1;

[0035] S7-3. Determine the type of the detected vehicle low beam pattern:

[0036] If it is a bathtub-type low beam, execute step S7-3a;

[0037] If it is a normal single-feature-point low-beam light pattern, execute step S7-3b.

[0038] In a further embodiment of the first aspect, step S7-3a comprises:

[0039] The four largest numbers in the sequence D2 are obtained in sequence, and these four values ​​and their corresponding serial numbers in the sequence D1 are respectively formed into sequences and stored in the sequence D3; the sequence D3 is expressed as:

[0040]

[0041] Where a1, a2, a3, a4 are the corresponding serial numbers of the four largest values ​​v1, v2, v3, v4 in the sequence D1;

[0042] After obtaining the serial numbers a1, a2, a3, and a4, the a1+nth point, a2+nth point, a3+nth point, and a4+nth point in the point sequence Q2 correspond to the four corner points in the measured low beam light pattern; n is the number of points;

[0043] And according to the size of the four values ​​of a1+n, a2+n, a3+n, and a4+n, the feature points and corner points can be distinguished. That is, the smallest value among the four values ​​is the point corresponding to the Q2 number as the first corner point, the second smallest value is the point corresponding to the Q2 number as the second corner point, the third smallest value is the point corresponding to the Q2 number as the feature point, and the maximum value is the point corresponding to the Q2 number as the third corner point.

[0044] In a further embodiment of the first aspect, step S7-3b comprises:

[0045] Obtain the maximum number in sequence D2 by value, and store the maximum number and its corresponding serial number in sequence D1 into a sequence in sequence D3. D3 is expressed as:

[0046]

[0047] In the formula, The maximum value The corresponding serial number in the sequence D1 is obtained Then, the point in Q2 is The points correspond to the characteristic points in the measured low beam light pattern, and n is the number of points.

[0048] A second aspect of the present invention provides a vehicle low beam light pattern feature point detection system, the detection system comprising:

[0049] A reading unit, used for reading an initial image I1 of a low beam light pattern;

[0050] A first processing unit, configured to perform smoothing filtering on the initial image I1 of the low beam light pattern to obtain an image I2;

[0051] A second processing unit is used to perform an opening operation on the image I2 to eliminate burrs on the edge of the image, so as to obtain an image I3;

[0052] A third processing unit, configured to perform binarization processing on the image I3 to obtain an image I4;

[0053] The fourth processing unit processes the image I4 using a Canny operator to obtain a contour point set Q1 of the low beam light pattern; the contour point set Q1 of the low beam light pattern and lines connecting adjacent points thereof are drawn in a blank image to obtain an image I5;

[0054] A fifth processing unit is used to compare the pixel value of the image I5 with the set value from left to right, column by column and from top to bottom, point by point. When it is detected that the pixel value of a point in a column is equal to the set value, it is determined that the point meets the condition, and all points that meet the comparison condition form a point column Q2;

[0055] The sixth processing unit is used to obtain the symmetry of each point in sequence starting from the nth point of the point sequence Q2, and store the symmetry in the number sequence D1; the value of n depends on the specific image resolution and the height difference of the feature point;

[0056] The output unit is used to compare the values ​​in the series D1, and the peak value is the characteristic point and corner point of the low beam light pattern of the automobile.

[0057] According to a third aspect of the present invention, an electronic device is provided, comprising a processor and a memory storing computer program instructions; when the processor executes the computer program instructions, the method for detecting characteristic points of a vehicle low beam pattern as described in the first aspect and its further embodiments is implemented.

[0058] A fourth aspect of the present invention provides a computer-readable storage medium storing at least one executable instruction. When the executable instruction is executed on an electronic device, the electronic device executes the method for detecting characteristic points of a vehicle low beam pattern as described in the first aspect and its further embodiments.

[0059] Beneficial effects: The low beam light pattern feature point detection method based on point symmetry proposed in the present invention can be applied to the detection of feature points, corner points and light and dark cut-off lines of the low beam light pattern. Compared with the traditional corner point detection algorithm, it has the advantages of fewer algorithm adjustment parameters, easy debugging, insensitivity to noise points, strong anti-interference ability, high measurement accuracy, fast measurement speed, etc. It has good practical value in the field of optical detection of automotive lamps. BRIEF DESCRIPTION OF THE DRAWINGS

[0060] Figure 1 Schematic diagram illustrating the difference in symmetry between non-corner points and corner points.

[0061] Figure 2 This is a flow chart of the low beam light pattern feature point detection method proposed in the present invention.

[0062] Figure 3 This is a schematic diagram of the cut-off line of the low beam light type; Figure 3 (a) is a schematic diagram of a bathtub-type low beam cutoff line; (b) is a schematic diagram of a common single-characteristic point low beam cutoff line.

[0063] Figure 4 It is a "bathtub-type" low beam pattern and its various processing diagrams; Figure 4 (a) is a "bathtub-shaped" low beam pattern; (b) is a "bathtub-shaped" low beam pattern with noise; (c) is a schematic diagram of the "bathtub-shaped" low beam pattern after rotating 30 degrees.

[0064] Figure 5 This is an effect diagram of image I2 obtained by performing smoothing filtering on the initial image I1.

[0065] Figure 6 This is the effect diagram of image I3 obtained by performing an opening operation on image I2.

[0066] Figure 7 This is the effect diagram of image I4 obtained by binarizing image I3.

[0067] Figure 8 The low beam light pattern profile calculated for the image I4 is shown in FIG.

[0068] Fig. 9 It is the changing curve of the sequence D2.

[0069] Fig.10 The following is a diagram showing the effect of detecting feature points and corner points using the method of the present invention.

[0070] Fig.11 This is a cutoff line detection effect diagram of the method of the present invention.

[0071] Fig.12 This is a diagram showing the detection effect of the method proposed in the present invention on a single feature point low beam light pattern; Fig.12 (a) is a common single feature point low beam light pattern diagram; (b) is a feature point detection effect diagram of the method of the present invention; (c) is a cutoff line detection effect diagram of the method of the present invention.

[0072] Fig.13 This is a comparison chart of the detection effects of the algorithm proposed by the present invention and the traditional algorithm; Fig.13 (a) is the detection effect of the algorithm of the present invention; (b) is the cutoff line detection effect of the algorithm of the present invention; (c) is the detection effect of the Harris detection algorithm; (d) is the detection effect of the SHI-Tomasi detection algorithm; (e) is the detection effect of the fast detection algorithm; (f) is the detection effect of the Moravec detection algorithm.

[0073] Fig.14 This is a comparison diagram of the effects of the algorithm proposed by the present invention and the traditional algorithm under the influence of noise; Fig.14 (a) is the detection effect of the algorithm of the present invention; (b) is the cutoff line detection effect of the algorithm of the present invention; (c) is the detection effect of the Harris detection algorithm; (d) is the detection effect of the SHI-Tomasi detection algorithm; (e) is the detection effect of the fast detection algorithm; (f) is the detection effect of the Moravec detection algorithm.

[0074] Fig.15 This is a comparison diagram of the detection effect of the algorithm proposed in the present invention on the rotated image; Fig.15 (a) is the algorithm of the present invention for Figure 4 (c) Detection effect; (b) The algorithm of the present invention is used to Figure 4 (c) Cut-off line detection effect. DETAILED DESCRIPTION

[0075] In the following description, a large number of specific details are provided to provide a more thorough understanding of the present invention. However, it is apparent to those skilled in the art that the present invention can be implemented without one or more of these details. In other examples, in order to avoid confusion with the present invention, some technical features known in the art are not described.

[0076] The low beam characteristic point mentioned in the present invention is the intersection of the horizontal part and the inclined part of the low beam cut-off line, that is, Figure 3 (a) and Figure 3 Point C in (b) is a characteristic point. The first corner point from left to right of the bathtub-type light pattern is defined as the first corner point in the present invention. Figure 3 (a) Point A; the second corner point from left to right of the bathtub-type light pattern is defined as the second corner point in the present invention, such as Figure 3 (a) point B; the third corner point from left to right of the bathtub-type light pattern is defined as the third corner point in the present invention, such as Figure 3 (a) Point D.

[0077] Reference Figure 1-15 The algorithm proposed in the present invention and its detection effect are described in detail. Here, the actual low-beam light pattern image taken by an industrial camera is used as the low-beam image input to illustrate the effectiveness of the algorithm.

[0078] Figure 1 The difference in symmetry between non-corner points and corner points is shown, where the upper part of the figure is a schematic diagram of a straight line containing 21 points, O1 is its middle point, and it is not a corner point. It can be seen that the ten points on the left and right are symmetrical about point O1, that is, the ten points on the left after rotation about O1 coincide with the ten points on the right, and the sum of the distances between their corresponding points is 0; the lower part of the figure is a schematic diagram of a broken line containing 21 points, O2 is its middle point, and it is a corner point. It can be seen that contrary to the above description, the ten points on the left after rotation about O1 not only coincide with the ten points on the right, but also the sum of the distances between the corresponding points is large, and the points on the left and right sides have a low degree of symmetry. It can be seen that at the corner point position, the symmetry of points within a certain length on the left and right is generally lower than the symmetry within the same length on the left and right at the non-corner point position. More specifically, at the corner point position, some points of the same number and same distance are taken on the left and right sides. For example, after all the points on the left side are rotated about the corner point, the sum of their distances from the corresponding points on the right is generally large, which is much larger than the value obtained by performing the same calculation operation at other non-corner points. The low beam feature point detection algorithm of the present invention is proposed based on this principle.

[0079] Embodiment 1:

[0080] The flowchart of the low beam feature point detection algorithm proposed by the present invention is as follows: Figure 2 As shown, specifically:

[0081] S1: This embodiment takes the "bathtub type" low beam light type as an example, inputs the initial image I1, such as Figure 4 As shown in (a), the resolution is 1600*1200. The initial image I1 is smoothed by using a uniform filter operator to obtain image I2, as shown in Figure 5 As shown, uniform filtering helps to improve the smoothness of the image edge. In this embodiment, the filter kernel size used for uniform smoothing is 11*11.

[0082] S2: Perform an opening operation on image I2 to obtain image I3, such as Figure 6As shown. The opening operation can further smooth the edge contour line and eliminate edge burrs. In this embodiment, the structure kernel size of the opening operation is 19*19;

[0083] S3: Binarize the image I3 to obtain a binary image I4, such as Figure 7 As shown, in this embodiment, the binarization threshold is set to 50;

[0084] S4: Use the Canny operator to process the image I4 to obtain the contour point set Q1 of the low beam light type, and then draw the contour point set Q1 and the connecting lines of its adjacent points in the blank image I5, as shown in Figure 8 As shown;

[0085] The point set Q1 and the lines connecting its adjacent points form a new point set Q1'; the pixel value of each point in the point set Q1' is set to V1, and V1 is set to 255 in this embodiment.

[0086] S5. Compare the pixel values ​​of the image I5 with the set values ​​from left to right, column by column, and from top to bottom, point by point. When a point pixel value is detected to be equal to the set value in a column, the point is determined to meet the condition. All points that meet the comparison condition form a point column Q2.

[0087] S6. Starting from the nth point, the symmetry of each point in the point sequence Q2 is calculated in turn and stored in the number sequence D1; the value of n depends on the specific image resolution and the height difference of the feature points;

[0088] In this embodiment, the symmetry degree is solved by using a weighted coefficient The formula has the advantage of reducing the sensitivity to noise points and having stronger anti-noise ability, thus making the algorithm more robust. The specific formula is:

[0089]

[0090] Where, d m Indicates the symmetry of the mth point; Represents the horizontal coordinate of the point after the ms-th point is rotated about the m-th point; Represents the horizontal coordinate of the m+sth point; Represents the ordinate of the point after the ms-th point is rotated about the m-th point; Indicates the horizontal coordinate of the m+sth point; n is the distance between points, that is, the number of points; is the weighting coefficient, and its specific calculation formula is:

[0091]

[0092] in, and Respectively meet:

[0093]

[0094]

[0095] In this embodiment, n is set to 20;

[0096] S7. Compare the values ​​in the sequence D1, and the peak values ​​obtained are the characteristic points and corner points of the low beam light pattern.

[0097] Starting from the L+1th number in the sequence D1, the L+1th value is compared with the previous L values ​​and the next L values ​​in sequence. If and only if the value of the L+1th point is the maximum value among these points, the L+1th point is stored in the sequence D2; then the above operation is performed on the L+2th number, L+3th number, etc. of the sequence D1 in sequence until the L1-Lth number; where L is the calculated window length, which is 15 in this embodiment, and L1 is the length of D1, which is 1290 in this embodiment, that is, the number of numbers it contains; Fig. 9 is the variation curve of D2, where the four black marked points correspond to the first corner point, the second focus, the feature point and the fourth corner point from left to right;

[0098] Then, the four largest numbers in D2 are obtained in sequence, and these four largest values ​​and their corresponding serial numbers in D1 are respectively formed into a sequence and stored in the sequence D3. D3 can be expressed as:

[0099]

[0100] a1, a2, a3, a4 (434, 476, 705, 753 in this embodiment) are respectively the corresponding serial numbers of the four largest values ​​v1, v2, v3, v4 (85.7, 118.4, 187.2, 231.8 in this embodiment) in the D1 sequence. After a1, a2, a3, a4 are obtained, the a1+nth point, a2+nth point, a3+nth point and a4+nth point in the point sequence Q2 correspond to the four corner points in the measured low beam light pattern, and the feature points and the corner points can be distinguished according to the sizes of the four values ​​a1+n, a2+n, a3+n, a4+n, that is, the smallest value among the four values ​​is the point corresponding to the Q2 serial number as the first corner point, the second smallest value is the point corresponding to the Q2 serial number as the second corner point, the third smallest value is the point corresponding to the Q2 serial number as the feature point, and the maximum value is the point corresponding to the Q2 serial number as the third corner point; Fig.10 In this embodiment, the coordinates of the first corner point are (454, 356), the coordinates of the second corner point are (496, 369), the coordinates of the feature point are (725, 359), and the coordinates of the fourth corner point are (773, 326).

[0101] At this point, the feature points and all corner points of the low beam light pattern have been detected. We only need to take the first point and the last point in the contour point set and connect them with the measured feature points and corner points in sequence to get the light and dark cutoff line, such as Fig.11 as shown in .

[0102] Embodiment 2:

[0103] Combination Fig.12 To illustrate the detection effect of the algorithm proposed in the present invention on the ordinary single feature point low beam light pattern.

[0104] In the detection of this embodiment, the real shot picture is cropped into a size of 400*280 for detection, such as Fig.12 (a). The detection steps of this embodiment are similar to those of the first embodiment, the only difference being that the low beam pattern detected in this embodiment only includes one feature point and no other corner points, so the last step needs to be changed to find the maximum value in D2 by value, and this number and its corresponding serial number in D1 are respectively formed into a sequence and stored in the sequence D3. D3 can be expressed as:

[0105]

[0106] in, (235 in this example) is the maximum value (10.1 in this example) The corresponding serial number in the D1 sequence is obtained Then, the point in Q2 is (255 in this embodiment) points correspond to the feature points in the measured low beam light pattern, n is the point length for calculating the symmetry, and in this embodiment it is 20. The feature point coordinates in this embodiment are (255, 131), and the feature point detection effects and the bright and dark cut-off line detection effects are as follows: Fig.12 (b) and Fig.12 As shown in (c).

[0107] Embodiment three:

[0108] Combination Figure 4 and Fig.13 To further illustrate the comparison between the algorithm proposed in the present invention and other corner detection algorithms.

[0109] Figure 4 (a) is a "bathtub-type" light image, which is the mainstream low-beam light type in the industry. It is used as an input image, and feature point detection is performed using the Harris corner detection method, the SHI-Tomasi corner detection method, the Fast corner detection method, and the Moravec corner detection algorithm, and the detection results are compared with those of the algorithm proposed in the present invention.

[0110] Fig.13(a) is the result of the algorithm of the present invention detecting the characteristic points and corner points of the low beam pattern. It can be seen that the algorithm proposed by the present invention can accurately detect the characteristic points of the low beam and other corner points (the gray cross points in the figure). At the same time, the low beam cutoff line can be obtained. Fig.13 (b). In contrast Fig.13 From the results of (c) to (f), the Harris corner detection method, the SHI-Tomasi corner detection method, and the fast corner detection method cannot accurately detect the feature points of the low beam, and the algorithm has too many parameters, which is not easy to debug. Fig.13 From the results of (f), although the Moravec corner detection algorithm detects the low-beam feature points, it also detects many interference points, which makes it difficult to accurately locate the feature points actually required, and its operating efficiency is low.

[0111] Embodiment 4:

[0112] With the help of Figure 4 and Fig.14 To verify the anti-noise capability of the algorithm proposed in the present invention, the parameters used in the algorithm are consistent with those in the first embodiment.

[0113] against Figure 4 The low beam pattern in (a) is added with strong Gaussian random noise with a mean of 0 and a variance of 0.01, as shown in Figure 4 (b) shows that the algorithm proposed in the present invention is used for detection and compared with the traditional corner detection algorithm. The results are shown in Fig.14 .Depend on Fig.14 From (a) to (f), we can see that the operator proposed in the present invention can still accurately identify the low beam feature points even in the case of strong noise. However, the Harris corner detection method, the SHI-Tomasi corner detection method, and the Fast corner detection method are seriously affected by noise, and many noise points are identified as corner points, and the feature points cannot be accurately detected; for the Moravec corner detection algorithm, it is less affected by noise, but there is still the problem of interference points, and the real corner points cannot be accurately located, and the operation efficiency is low.

[0114] Embodiment five:

[0115] With the help of Figure 4 and Fig.15 To further verify the rotation invariance of the algorithm of the present invention, the parameters used in the algorithm are consistent with those in Example 1.

[0116] use Figure 4 (c) As input image, Figure 4 (c) Yes Figure 4 (a) is a rotation image, given by Figure 4 (a) Rotated 30° clockwise.

[0117] Fig.15 (a) is the algorithm of the present invention Figure 4 (c) Results of low beam light pattern corner point detection. It can be seen from the figure that after the image is rotated at a large angle, the algorithm of the present invention can still accurately detect the low beam feature points and other included corner points (gray cross points in the figure); Fig.15 (b) is the low beam cutoff line and its feature points and corner points detected by the algorithm of the present invention.

[0118] Through this embodiment, it can be found that the algorithm of the present invention can also accurately detect the feature points and corner points of the rotating low beam light pattern, and accurately draw the low beam cutoff line. This positively illustrates that the algorithm of the present invention is also applicable to the situation where the low beam light pattern is tilted at a large angle, and also verifies that the algorithm of the present invention is rotationally invariant.

[0119] As described above, although the present invention has been shown and described with reference to specific preferred embodiments, it should not be construed as limiting the present invention itself. Various changes may be made to it in form and detail without departing from the spirit and scope of the present invention as defined in the appended claims.

Claims

1. A method for detecting characteristic points of a vehicle low beam pattern based on point symmetry, characterized in that: The steps include: S1, inputting an initial image I1 of a low beam light pattern, performing a smoothing filter process on the initial image I1 of the low beam light pattern, and obtaining an image I2; S2, performing an opening operation on the image I2 to eliminate burrs on the edge of the image to obtain an image I3; S3, performing binarization processing on the image I3 to obtain an image I4; S4, using the Canny operator to process the image I4 to obtain a contour point set Q1 of the low beam light pattern; drawing the contour point set Q1 of the low beam light pattern and the connecting lines of its adjacent points in a blank image to obtain an image I5; S5, comparing the pixel values ​​of the image I5 with the set values ​​from left to right, column by column, and from top to bottom, point by point. When the pixel value of a point in a column is detected to be equal to the set value, it is determined that the current point satisfies the condition, and all the points that meet the comparison condition form a point column Q2; S6. Starting from the nth point of the point sequence Q2, the symmetry of each point is obtained in turn and stored in the number sequence D1; the value of n depends on the specific image resolution and the height difference of the feature points; the solution formula of the symmetry is as follows: ； Where, d m Indicates the symmetry of the mth point; Represents the horizontal coordinate of the point after the ms-th point is rotated about the m-th point; Represents the horizontal coordinate of the m+sth point; Represents the ordinate of the point after the ms-th point is rotated about the m-th point; Indicates the horizontal coordinate of the m+sth point; n is the number of points; is the weighting coefficient, ; in, and They satisfy the following relationships respectively: ； ； In the formula, s represents the distance between the points; Indicates the distance between the msth point and the m+sth point after rotation about the mth point; It represents the average distance between the msth point and the m+sth point after rotating about the mth point; S7, comparing the values ​​in the sequence D1, the peak values ​​are the characteristic points and corner points of the low beam light pattern of the vehicle, specifically including: S7-1, starting from the L+1th number in the sequence D1, compare the L+1th value with the previous L values ​​and the next L values ​​in sequence, and if and only if the value of the L+1th point is the maximum value among these points, store the L+1th point in the sequence D2; S7-2, sequentially perform the operation of step S7-1 on the subsequent numbers of sequence D1 until the L1-Lth number, where L is the calculated window length, and L1 is the length of sequence D1, that is, the number of numbers included in sequence D1; S7-3. Determine the type of the detected vehicle low beam pattern: If it is a bathtub-type low beam, execute step S7-3a; If it is a normal single-feature-point low-beam light pattern, execute step S7-3b; Wherein, step S7-3a includes: The four largest numbers in the sequence D2 are obtained in sequence, and these four values ​​and their corresponding serial numbers in the sequence D1 are respectively formed into sequences and stored in the sequence D3; the sequence D3 is expressed as: ； Where a1, a2, a3, a4 are the corresponding serial numbers of the four largest values ​​v1, v2, v3, v4 in the sequence D1; After obtaining the serial numbers a1, a2, a3, and a4, the a1+nth point, a2+nth point, a3+nth point, and a4+nth point in the point sequence Q2 correspond to the four corner points in the measured low beam light pattern; n is the number of points; And according to the size of the four values ​​of a1+n, a2+n, a3+n, and a4+n, the feature points and corner points can be distinguished. That is, the smallest value among the four values ​​is the point corresponding to the Q2 number as the first corner point, the second smallest value is the point corresponding to the Q2 number as the second corner point, the third smallest value is the point corresponding to the Q2 number as the feature point, and the maximum value is the point corresponding to the Q2 number as the third corner point; Step S7-3b includes: Calculate the maximum number in sequence D2 by value, and form a sequence with the maximum number and its corresponding serial number in sequence D1 and store them in sequence D3. D3 is expressed as: ； In the formula, The maximum value The corresponding serial number in the sequence D1 is obtained Then, the point in Q2 is The points correspond to the characteristic points in the measured low beam light pattern, and n is the number of points.

2. The method for detecting characteristic points of automobile low beam light pattern based on point symmetry according to claim 1, characterized in that: The smoothing filtering process in step S1 is side window mean filtering and / or side window Gaussian filtering; The image I5 has the same size as the initial image I1.

3. The method for detecting characteristic points of automobile low beam light pattern based on point symmetry according to claim 1, characterized in that: In step S4, the pixel value of each point in the contour point set Q1 is set to V1, and the pixel value N of each point in the image I5 is expressed as: ； Where P ij is the point with coordinates (i, j) in the point set Q1; V1 is an integer between 1 and 255.

4. The method for detecting characteristic points of automobile low beam light pattern based on point symmetry according to claim 3, characterized in that: The set value in step S5 is equal to V1.

5. The method for detecting characteristic points of automobile low beam light pattern based on point symmetry according to claim 1, characterized in that: Starting from the mth point of the point sequence Q2, the symmetry of each point is obtained in turn. The solution formula of the symmetry is as follows: ； Where, d m Indicates the symmetry of the mth point; Represents the horizontal coordinate of the point after the ms-th point is rotated about the m-th point; Represents the horizontal coordinate of the m+sth point; Represents the ordinate of the point after the ms-th point is rotated about the m-th point; It represents the horizontal coordinate of the m+sth point; n is the distance between the points, that is, the number of points.

6. A vehicle low beam light pattern feature point detection system, the system is used to implement the vehicle low beam light pattern feature point detection method based on point symmetry as described in any one of claims 1 to 5, characterized in that: include: A reading unit, used for reading an initial image I1 of a low beam light pattern; A first processing unit, configured to perform smoothing filtering on the initial image I1 of the low beam light pattern to obtain an image I2; A second processing unit is used to perform an opening operation on the image I2 to eliminate burrs on the edge of the image, so as to obtain an image I3; A third processing unit, configured to perform binarization processing on the image I3 to obtain an image I4; The fourth processing unit processes the image I4 using a Canny operator to obtain a contour point set Q1 of the low beam light pattern; the contour point set Q1 of the low beam light pattern and lines connecting adjacent points thereof are drawn in a blank image to obtain an image I5; A fifth processing unit is used to compare the pixel value of the image I5 with the set value from left to right, column by column and from top to bottom, point by point. When it is detected that the pixel value of a point in a column is equal to the set value, it is determined that the current point satisfies the condition, and all points that meet the comparison condition form a point column Q2; The sixth processing unit is used to obtain the symmetry of each point in sequence starting from the nth point of the point sequence Q2, and store the symmetry in the number sequence D1; the value of n depends on the specific image resolution and the height difference of the feature point; The output unit is used to compare the values ​​in the series D1, and the peak value is the characteristic point and corner point of the low beam light pattern of the automobile.

Images