Headlight testing machine, measuring method, and measuring program

Abstract

To reduce the gap between a result obtained on the basis of an image obtained by picking up a screen projection image with a camera and a result obtained by determining the screen projection image by visual observation.SOLUTION: A headlight testing machine includes: a screen on which light from a headlight of a vehicle is projected; a camera that picks up the image projected on the screen; and a processing device that determines a light and dark boundary point group present in a measurement image picked up by the camera, and determines an elbow point on the basis of the determined light and dark boundary point group. The processing device corrects the measurement image with human contrast sensitivity characteristics before determining the elbow point.SELECTED DRAWING: Figure 5

Description

[Technical field]

[0001] The present invention relates to a headlamp tester, a measurement method, and a measurement program. [Background technology]

[0002] In vehicle headlamp testing, elbow points and other adjustments are made based on the image that the headlamp projects onto a screen (hereinafter referred to as the "screen projected image"). The screen projected image indicates the light distribution characteristics of the headlamp.

[0003] Headlights for passing other vehicles (so-called low beams) are required to illuminate the area ahead of the vehicle in a wide and easy-to-see manner, while at the same time preventing dazzling to drivers of oncoming and preceding vehicles. To satisfy these requirements, the light distribution characteristics are managed based on the elbow point.

[0004] The elbow point is defined as the intersection of a horizontal cutoff and a diagonal cutoff in the projected image on the screen. For example, for the headlights of a vehicle driving on the left side of the road, the horizontal part on the right side of the projected image on the screen is the horizontal cutoff, and the part on the left side that slopes upward to the left at a certain angle is the diagonal cutoff.

[0005] More specifically, the elbow point must be adjusted so that in the vertical direction it is slightly below (within 0.57 degrees) the horizontal line on the left side of the screen projection image, and in the horizontal direction it coincides with the direction of travel of the vehicle.

[0006] Regarding the testing of such headlamps and the measurement of light distribution characteristics, the following prior art exists.

[0007] Japanese Patent Application Laid-Open Publication No. 2007-163149 (Patent Document 1) discloses a method for adjusting the optical axis of a downward-facing headlight, in which the optical axis is adjusted depending on whether or not an intersection point (called an elbow point) between a horizontal cut line and a diagonal cut line that constitutes a cut line that is a light-dark boundary point in a light distribution pattern image projected from the downward-facing headlamp falls within a specified inspection pass range.

[0008] Japanese Patent No. 6570816 (Patent Document 2) discloses a color headlight tester that tests headlights by performing color image processing on received light images of the headlights.

[0009] JP 2020-91182 A (Patent Document 3) discloses a color image processing type headlight tester that forms a cutoff line.

[0010] Japanese Patent Application Laid-Open No. 2017-67656 (Patent Document 4) discloses a technique that can detect the elbow point with relatively high accuracy.

[0011] Shinichiro Ito et al., "Detection of the Illumination Direction of Passing Headlights" (Non-Patent Document 1) discloses an improvement in technology that can detect the elbow point with relatively high accuracy. [Prior art documents] [Patent documents]

[0012] [Patent Document 1] JP 2007-163149 A [Patent Document 2] Patent No. 6570816 [Patent Document 3] JP 2020-91182 A [Patent Document 4] JP 2017-67656 A [Non-patent literature]

[0013] [Non-Patent Document 1] Shinichiro Ito et al., "Detection of the direction of headlights for passing other vehicles," Transactions of the Society of Automotive Engineers of Japan, Vol. 48, No. 2, 2017, pp. 587-592 Summary of the Invention [Problem to be solved by the invention]

[0014] One object of the present invention is to reduce the discrepancy between the results obtained based on an image of a screen projected image captured by a camera and the results obtained by visually judging the screen projected image. [Means for solving the problem]

[0015] (Configuration 1) A headlamp tester according to an aspect of the present invention includes a screen onto which light from a vehicle's headlamp is projected, a camera that captures the image projected onto the screen, and a processing device that determines a light-dark boundary point cloud that exists in a measurement image captured by the camera and determines an elbow point based on the determined light-dark boundary point cloud. The processing device corrects the measurement image using human contrast sensitivity characteristics and then determines the elbow point.

[0016] (Configuration 2) In configuration 1, the processing device may generate a corrected image by correcting the measured image by sequentially calculating a convolution integral between the measured image and a filter matrix corresponding to the contrast sensitivity characteristic of a human being.

[0017] (Configuration 3) In configuration 1 or 2, the process of determining the light-dark boundary point group may include a process of generating a light-dark profile indicating a change in light intensity in a second direction of the measurement image at each position along the first direction of the measurement image, and a process of calculating a derivative of the light-dark profile for each light intensity included in the light-dark profile based on a polynomial representing each light intensity and multiple light intensities adjacent to each light intensity.

[0018] (Configuration 4) In configuration 3, the process of determining the light-dark boundary point group may include a process of determining, as the light-dark boundary point, a point at which the amount of change in luminous intensity in the light-dark profile is maximum.

[0019] (Configuration 5) In configuration 3 or 4, the process of calculating the derivative of the brightness profile may include a process of convoluting each luminous intensity included in the brightness profile and a plurality of luminous intensities adjacent to each luminous intensity using a weighting coefficient table.

[0020] (Configuration 6) In any of configurations 3 to 5, the process of determining the light-dark boundary point group may include a process of determining a division included in the light-dark boundary point group by fitting a piecewise linear function to the light-dark boundary point group, a process of determining as a provisional elbow point a bend point having the highest score calculated according to a predetermined criterion among the bend points defining the determined division, and a process of determining the elbow point by fitting a two-piecewise linear function to a range including the provisional elbow point and the bend points adjacent to both sides of the provisional elbow point.

[0021] (Configuration 7) In configuration 6, the process of determining the light / dark boundary point group may further include a process of merging two adjacent sections having similar slopes from among the determined sections, and then repeating the process of determining the sections.

[0022] (Configuration 8) A measurement method according to another aspect of the present invention includes the steps of obtaining a measurement image by capturing an image of light from a vehicle's headlights projected onto a screen, correcting the measurement image using human contrast sensitivity characteristics, determining a group of light-dark boundary points present in the corrected measurement image, and determining an elbow point based on the determined group of light-dark boundary points.

[0023] (Configuration 9) A measurement program according to yet another aspect of the present invention causes a computer to execute the steps of acquiring a measurement image obtained by capturing an image of light from a vehicle's headlights projected onto a screen, correcting the measurement image using human contrast sensitivity characteristics, determining a group of light-dark boundary points present in the corrected measurement image, and determining an elbow point based on the determined group of light-dark boundary points. Effect of the Invention

[0024] According to an embodiment of the present invention, it is possible to reduce the discrepancy between the results obtained based on an image of a screen projected image captured by a camera and the results obtained by visually judging the screen projected image. [Brief description of the drawings]

[0025] [Figure 1] FIG. 2 is a schematic diagram showing an example of the arrangement of a headlamp tester according to the present embodiment. [Diagram 2] FIG. 2 is a schematic diagram showing a configuration example of a lens unit of a headlamp tester according to the present embodiment. [Diagram 3] FIG. 2 is a schematic diagram showing an example of an electrical configuration of a headlamp tester according to the present embodiment. [Figure 4] 5 is a flowchart showing an example of a processing procedure for a headlamp test using the headlamp tester according to the present embodiment. [Diagram 5] 5 is a flowchart showing a more detailed procedure of the measurement process (step S7) shown in FIG. 4. [Figure 6] 1 is a graph showing an example of a contrast sensitivity function. [Figure 7] FIG. 11 is a diagram for explaining preprocessing using a filter matrix according to the present embodiment. [Figure 8] 6 is a flowchart showing a more detailed procedure of the pre-processing (step S10) shown in FIG. 5. [Figure 9] 10A and 10B are diagrams showing an example of a measurement image captured by a measurement camera of the headlamp tester according to the present embodiment. [Figure 10] 10 is a diagram showing an example of a contour display of the measurement image shown in FIG. 9 and a corrected image generated from the measurement image shown in FIG. 9. FIG. [Figure 11] 5 is a schematic diagram of a measurement image captured by a measurement camera of the headlamp tester according to the present embodiment. FIG. [Figure 12] 12 is a diagram showing an example of a light-dark profile in the Y direction of the measurement image shown in FIG. 11. FIG. [Figure 13] FIG. 13 illustrates an example process for determining the maximum luminous intensity change point using a polynomial fitting method according to an embodiment. [Figure 14] It is a flowchart showing a more detailed processing procedure of the process (step S20) for determining the cutoff shown in FIG. 5. [Figure 15] It is a diagram showing an example of a process for determining an elbow point according to the present embodiment. [Figure 16] It is a flowchart showing a more detailed processing procedure of the process (step S30) for determining the elbow point shown in FIG. 5. [Figure 17] It is a diagram showing an example of a corrected image generated by preprocessing according to the present embodiment. [Figure 18] It is a diagram showing an example of the effect of preprocessing on the process for determining the cutoff according to the present embodiment. [Figure 19] It is a diagram showing an example of the effect on the process for determining the elbow point according to the present embodiment.

Mode for Carrying Out the Invention

[0026] Embodiments of the present invention will be described in detail with reference to the drawings. Note that the same or corresponding parts in the drawings are denoted by the same reference numerals and their descriptions will not be repeated.

[0027] In this specification, "cutoff" means the boundary between the bright part (the part with relatively high light intensity) and the dark part (the part with relatively low light intensity) in the image. In an image that is a set of pixels, "cutoff" may be a group of light and dark boundary points. The line defined by the group of light and dark boundary points can also be referred to as the light and dark boundary line. Therefore, "cutoff" may also mean the light and dark boundary line.

[0028] A part of the cutoff in the image may be determined as a "horizontal cutoff" or an "oblique cutoff".

[0029] <A. Example of Device Configuration> An example of a device configuration for realizing an optical measurement method according to the present embodiment will be described.

[0030] Fig. 1 is a schematic diagram showing an example of the arrangement of a headlamp tester 1 according to the present embodiment. Referring to Fig. 1, the headlamp tester 1 is associated with, for example, a headlamp S of a vehicle V to be tested. Therefore, two headlamp testers 1 may be arranged as a set. Note that a common processing device may control the two headlamp testers 1.

[0031] Each of the headlamp testers 1 includes a lens unit 10, a pillar portion 4 that supports the lens unit 10 so that it can move up and down, and a base portion 6 that supports the pillar portion 4 so that it can move horizontally.

[0032] The lens unit 10 receives light emitted by the headlight S and measures the light distribution characteristics of the headlight S. The lens unit 10 is disposed at a position directly facing the headlight S under test by the pillar portion 4 and the base portion 6. Typically, the lens unit 10 is disposed at a position 1 m forward from the headlight S of the vehicle V under test. However, an image projected onto a screen 13 (see FIG. 2) disposed inside the lens unit 10 (screen projected image) is a similar image equivalent to an image projected 10 m ahead. Therefore, the lens unit 10 can capture the light distribution characteristics 10 m ahead of the headlight S.

[0033] In addition to determining the elbow point, the headlamp tester 1 may be capable of various other inspections and aiming adjustments, etc. The headlamp tester 1 may be capable of testing both passing headlamps (low beam) and driving headlamps (high beam).

[0034] 2 is a schematic diagram showing a configuration example of a lens unit 10 of a headlamp tester 1 according to the present embodiment. Referring to FIG. 2, the lens unit 10 includes a Fresnel lens 12, a screen 13, a measurement camera 14, a facing confirmation camera 15, and a facing adjustment laser 16.

[0035] The Fresnel lens 12 is disposed at the light receiving port of the lens unit 10. The Fresnel lens 12 is a condensing lens for condensing light emitted from the headlight S. The screen 13 is disposed on the optical axis of the Fresnel lens 12. The light from the headlight S of the vehicle V is condensed by the Fresnel lens 12 and then projected onto the screen 13.

[0036] The measurement camera 14 captures the image projected on the screen 13. The image obtained by the measurement camera 14 (hereinafter, also referred to as a "measurement image") includes information on the light distribution characteristic and luminous intensity of the headlight S.

[0037] The facing confirmation camera 15 is disposed on the upper part of the lens unit 10. The facing confirmation camera 15 captures an image (hereinafter also referred to as a "lamp image") that is generated by the headlight S of the vehicle V emitting light. The image captured by the facing confirmation camera 15 may be displayed on the display unit 30 (FIG. 3). The facing confirmation camera 15 is used both when adjusting the facing between the vehicle V and the lens unit 10 (hereinafter also referred to as "vehicle facing") and when adjusting the facing between the headlight S and the lens unit 10 (hereinafter also referred to as "lamp facing").

[0038] The facing adjustment laser 16 is a line laser that irradiates a reference line for positioning the vehicle V. When adjusting the facing between the vehicle V and the lens unit 10 (vehicle facing), the facing adjustment laser 16 irradiates the reference line onto the vehicle V, and when adjusting the facing between the headlight S and the lens unit 10 (lamp facing), the facing adjustment laser 16 irradiates the reference line onto the headlight S.

[0039] 3 is a schematic diagram showing an example of an electrical configuration of the headlamp tester 1 according to the present embodiment. Referring to FIG. 3, the headlamp tester 1 includes a processing device 100 electrically connected to a lens unit 10.

[0040] The processing device 100 is, for example, a general-purpose computer and includes one or more processors 102, memory 104, and storage 106.

[0041] The processor 102 reads one or more programs stored in the storage 106 into the memory 104 and executes them. The processor 102 is composed of, for example, a central processing unit (CPU) or a graphics processing unit (GPU). The processor 102 may include a hardwired logic circuit such as a field-programmable gate array (FPGA) or an application specific integrated circuit (ASIC).

[0042] The memory 104 functions as a working memory for executing programs by the processor 102. The memory 104 is, for example, a dynamic random access memory (DRAM) or a static random access memory (SRAM).

[0043] The storage 106 stores programs (computer-readable instructions) executed by the processor 102 and data necessary for processing. The storage 106 is composed of a non-transitory computer-readable medium such as a hard disk or a flash memory. The storage 106 stores, for example, a system program 108 and a measurement program 110. The system program 108 includes computer-readable instructions such as an operating system. The measurement program 110 includes computer-readable instructions for implementing measurement processing (including image processing) described below.

[0044] The processing device 100 is connected to an operation unit 20, a display unit 30, and a drive device 40. The processing device 100 may include an IO (Input Output) circuit and a relay circuit as interfaces.

[0045] The operation unit 20 accepts user operations. The operation unit 20 may include physical buttons and levers in addition to a keyboard, a mouse, etc. At least a part of the operation unit 20 may be disposed in a box attached to the column 4. The processing device 100 starts a measurement process according to the user operations accepted by the operation unit 20. The processing device 100 also issues commands to the drive device 40 according to the user operations accepted by the operation unit 20.

[0046] The display unit 30 displays the captured image and the processing results by the processing device 100. The display unit 30 is, for example, a liquid crystal display.

[0047] The display unit 30 may display information necessary for the headlamp tester 1, such as the measured value of luminous intensity, the measured value of irradiation direction, the measurement result, pass / fail of the test, the current test mode, tester settings, etc. The display unit 30 may display images for visually checking the vehicle V when the vehicle is facing directly, the headlamp S when the lamp is facing directly, the light distribution characteristics of the headlamp S, etc. The display unit 30 may display a user interface screen for the user to make adjustments.

[0048] The driving device 40 includes an actuator such as a motor, and is disposed inside the pillar portion 4, etc. The driving device 40 changes the position of the lens unit 10 in the up and down direction by operating according to commands from the operation unit 20. In this case, the driving device 40 may include an elevation motor for raising and lowering the lens unit 10. The driving device 40 changes the positions of the pillar portion 4 and the lens unit 10 in the width direction of the vehicle V by operating according to commands from the operation unit 20. In this case, the driving device 40 may include a movement motor for horizontally moving the lens unit 10.

[0049] The processing device 100 performs image processing, as described below, using the measurement image obtained by capturing an image with the measurement camera 14, and displays the processing results (including the measurement results) on the display unit 30. The user can understand the measurement results by referring to the display unit 30.

[0050] The processing device 100 may be connected to a printer for printing out processing results (including measurement results).

[0051] <B. Example of Processing Procedure for Headlamp Test> Next, an example of the processing procedure for a headlamp test using the headlamp tester 1 according to the present embodiment will be described.

[0052] FIG. 4 is a flowchart showing an example of the processing procedure for a headlamp test using the headlamp tester 1 according to the present embodiment. Referring to FIG. 4, the user operates the operation unit 20 to select a test mode (for example, a headlamp for driving or a headlamp for passing) (step S1).

[0053] The user operates the operation unit 20 to start the irradiation of the alignment adjustment laser 16 in order to adjust the vehicle alignment (step S2). Then, while the user checks the image captured by the alignment confirmation camera 15 on the display unit 30, the vehicle V and the lens unit 10 are aligned (vehicle alignment) (step S3).

[0054] Subsequently, in order to adjust the lamp alignment, while the user checks the image captured by the alignment confirmation camera 15 on the display unit 30, the user adjusts the position of the lens unit 10 (step S4). More specifically, the user operates the operation unit 20 to raise and lower the lens unit 10 and / or move the headlamp tester 1 in the left - right direction so that the lamp image displayed on the display unit 30 is at a predetermined position.

[0055] When the adjustment of the lamp alignment is completed, the user operates the operation unit 20 to instruct the start of measurement (step S5).

[0056] The processing device 100 acquires a measurement image captured by the measurement camera 14 (step S6). That is, the processing device 100 acquires a measurement image obtained by imaging an image (screen projection image) in which the light from the headlamp S of the vehicle V is projected onto the screen 13.

[0057] The processing device 100 executes a measurement process on the acquired measurement image (step S7). The measurement image generated by the measurement camera 14 may be resampled and processed to a predetermined image size (for example, 256×256 pixels).

[0058] The processing device 100 outputs the result of the measurement process to the display unit 30 (step S8), and stores the result of the measurement process in the storage 106 or the like (step S9). Then, the headlamp test ends.

[0059] Fig. 5 is a flowchart showing a more detailed processing procedure of the measurement process (step S7) shown in Fig. 4. Each step shown in Fig. 5 may be realized by one or more processors 102 of the processing device 100 executing the measurement program 110.

[0060] 5, the processing device 100 performs pre-processing on the measurement image (step S10). As will be described in detail later, in the pre-processing, the processing device 100 corrects the measurement image using human contrast sensitivity characteristics. The processed image obtained by the pre-processing is also referred to as a "corrected image" below.

[0061] The pre-processing may be omitted. Therefore, except for the following description of the pre-processing, the term "measurement image" may be read as "corrected image."

[0062] Next, the processing device 100 determines a cutoff present in the measurement image (or the corrected image) (step S20). As described above, the cutoff is the boundary between light and dark areas in an image. As will be described in detail later, in the process of determining the cutoff (step S20), the processing device 100 determines a set of light-dark boundary points (light-dark boundary point group) present in the measurement image (or the corrected measurement image (corrected image)).

[0063] Subsequently, the processing device 100 determines the elbow points based on the determined light and dark boundary point group (step S30). Note that the processing device 100 may determine whether the position of the determined elbow points exists within a predetermined pass region. Then, the processing from step S8 and below in FIG. 3 is executed.

[0064] In this way, the processing device 100 determines the light and dark boundary point group (cutoff) existing in the measurement image captured by the measurement camera 14, and determines the elbow points based on the determined light and dark boundary point group. At this time, the processing device 100 may correct the measurement image with the contrast sensitivity characteristics of a human and then determine the elbow points.

[0065] Hereinafter, more detailed content of the measurement process shown in FIG. 5 will be described.

[0066] <C. Pretreatment for Measurement Image (Step S10)> The measurement process executed by the processing device 100 according to the present embodiment may include pretreatment for the measurement image (step S10 in FIG. 5). The pretreatment may take into account the contrast sensitivity characteristics of a human. Hereinafter, the pretreatment considering the contrast sensitivity characteristics of a human will be described.

[0067] When determining the elbow points of the passing headlamp, it is necessary to determine the cutoffs (horizontal cutoff and diagonal cutoff) in the screen projection image. The cutoffs exist in the gradient part of light and dark, that is, the part where light and dark are ambiguous. Therefore, if a contrast difference occurs between the image projected onto the screen 13 (screen projection image) and the measurement image obtained by imaging with the measurement camera 14, measurement errors of the cutoffs and the elbow points may occur.

[0068] As an index indicating how faithfully an imaging device such as a camera (including an optical system such as a lens) reproduces and images the image projected onto the screen, the modulation transfer function (MTF) is known. The modulation transfer function means the contrast sensitivity characteristics with respect to the spatial frequency.

[0069] In general, the contrast sensitivity characteristic of an imaging device decreases as the spatial frequency increases, but since the contrast sensitivity characteristic of an imaging device is approximately linear with respect to the spatial frequency, the contrast sensitivity is maintained over a relatively wide range of spatial frequencies.

[0070] In addition to being tested using the headlamp tester 1, headlamp testing may also be performed by a qualified inspector visually observing an image projected on a screen. Ideally, the same results should be obtained in either method. The inventors of the present application have found a new solution to prevent differences in results between the results obtained using the headlamp tester 1 and the results obtained by visual observation by an inspector.

[0071] That is, the measurement process according to this embodiment may employ pre-processing that takes into consideration the difference in contrast sensitivity characteristics that exists between an imaging device such as the measurement camera 14 and human perception sensitivity. By employing such pre-processing, it is possible to bring closer the result obtained by measuring and processing the measurement image obtained by the measurement camera 14 capturing the image projected on the screen 13 of the headlamp tester 1 (screen projection image) to the result obtained by visually judging the screen projection image.

[0072] The human eye also has a characteristic of contrast sensitivity (visual sensitivity) with respect to spatial frequency, similar to the modulation transfer function of an imaging device. Such a characteristic is called the contrast sensitivity function (CSF). Various mathematical models have been proposed for the contrast sensitivity function, and any mathematical model may be adopted.

[0073] For example, for a spatial frequency u, a simple approximation of the contrast sensitivity function as shown in the following equation (1) is known.

[0074]

number

[0075] Fig. 6 is a graph showing an example of a contrast sensitivity function, which shows the change in contrast sensitivity with respect to the spatial frequency u (cycle / degree) calculated according to equation (1).

[0076] 6, the contrast sensitivity function shows that the contrast sensitivity drops sharply in the low frequency range and monotonically decreases in the high frequency range. As shown in FIG. 6, the contrast sensitivity function shows that the contrast sensitivity changes significantly with respect to the spatial frequency. The applicants of the present application have found a new finding that such a feature of the contrast sensitivity function may be the cause of a difference between the inspector's perception of the light and dark gradients present in the screen projected image and the measurement processing result of the measurement image obtained by capturing the screen projected image with the measurement camera 14.

[0077] As a specific processing procedure, an image (corrected image) showing the screen projected image perceived by the human eye is generated by applying a filter of the contrast sensitivity function to the measurement image obtained by imaging with the measurement camera 14. Note that the influence of the modulation transfer function of the measurement camera 14 on the contrast sensitivity function can be ignored.

[0078] Since the contrast sensitivity function is a characteristic of contrast with respect to spatial frequency, the measurement image also needs to be transformed (Fourier transformed) into the spatial frequency domain. The measurement image g(x,y) defined by the two-dimensional coordinates (x,y) is Fourier transformed according to equation (2).

[0079]

number

[0080] where u,v are spatial frequencies and G(u,v) represents the measured image in the spatial frequency domain.

[0081] To apply the contrast sensitivity function filter to the measured image G(u,v), a two-dimensional extension of the contrast sensitivity function shown in the above equation (1) is prepared as C(u,v). The image domain corrected image gc(x,y) is obtained by inverse Fourier transforming the product of the measured image G(u,v) and the contrast sensitivity function C(u,v) according to equation (3).

[0082]

number

[0083] The corrected image gc(x, y) refers to the measured image after the filter of the contrast sensitivity function has been applied, that is, an image in which the spatial frequency characteristics of the contrast sensitivity of the human eye are reflected.

[0084] Note that the Fourier transform and inverse Fourier transform of two-dimensional data require a lot of processing resources. Also, the larger the image size (data size), the more calculation time is required. Therefore, the following method may be adopted.

[0085] Specifically, the contrast sensitivity function c(x,y) in the image domain is prepared by inverse Fourier transforming the contrast sensitivity function C(u,v), which is a two-dimensional expansion of the contrast sensitivity function. Specifically, the inverse Fourier transform is performed according to equation (4).

[0086]

number

[0087] When the measured image g(x,y) is convoluted with the contrast sensitivity function c(x,y), the result is given by equation (5), where "*" indicates the convolution integral.

[0088]

number

[0089] Equation (5) can be transformed into the following equation (6).

[0090]

number

[0091] That is, convolution of the contrast sensitivity function c(x,y) with the measured image g(x,y) produces the same results as the series of calculations including the Fourier transform described above. Therefore, it means that a filter matrix (e.g., about 3 rows × 3 columns) that mimics the contrast sensitivity function c(x,y) is prepared in advance, and the measured image g(x,y) is sequentially filtered with the matrix to obtain the corrected image gc(x,y).

[0092] By using such a relatively small filter matrix, the calculation time can be reduced.

[0093] Fig. 7 is a diagram for explaining pre-processing using a filter matrix according to this embodiment. Referring to Fig. 7, a filter matrix 60 having an image size smaller than that of a measurement image g(x, y) is prepared. The filter matrix 60 is sequentially applied to the measurement image g(x, y). The relative positions of the measurement image g(x, y) and the filter matrix 60 are shifted by a predetermined number of pixels (for example, one pixel) and convolution integral is performed, thereby sequentially calculating pixel values ​​of a corrected image gc(x, y).

[0094] When the filter matrix 60 is shifted by one pixel in both the X and Y directions, a corrected image gc(x, y) can be generated by performing x×y convolution integrals.

[0095] The contrast sensitivity function c(x,y) can be approximated by a difference of Gaussian (DoG), which is the difference between two Gaussian functions. Therefore, a filter matrix 60 for the measurement image g(x,y) may be generated from the DoG.

[0096] In this way, the processing device 100 generates a corrected image gc(x,y) by correcting the measured image g(x,y) by sequentially calculating the convolution integral of the measured image g(x,y) and the filter matrix 60 corresponding to the human contrast sensitivity characteristic. By using the filter matrix 60 with a small image size, the calculation time can be significantly reduced.

[0097] Please note that it is meaningless for humans to see the corrected image gc(x,y) generated by the above preprocessing. This is because when a human sees the corrected image gc(x,y), the filter of the contrast sensitivity function of the human eye is applied again and the image is perceived. Therefore, the difference between the corrected image and the measured image should be evaluated using a contour display or a one-way light-dark profile.

[0098] Fig. 8 is a flowchart showing a more detailed processing procedure of the pre-processing (step S10) shown in Fig. 5. Fig. 8(A) shows an example of the processing procedure of the pre-processing using Fourier transform and inverse Fourier transform. Fig. 8(B) shows an example of the processing procedure of the pre-processing using a filter matrix.

[0099] 8(A), the processing device 100 performs a Fourier transform on the acquired measurement image to generate a measurement image G(u,v) in the spatial frequency domain (step S101). The processing device 100 calculates the product of the measurement image G(u,v) and a contrast sensitivity function C(u,v) to generate a corrected image in the spatial frequency domain (step S102). The processing device 100 performs an inverse Fourier transform on the corrected image in the spatial frequency domain to generate a corrected image gc(x,y) in the image domain (step S103). Then, the processing returns.

[0100] 8(B), the processing device 100 sets the filter matrix at the initial position of the acquired measurement image (step S111), and performs convolution integral between the filter matrix and a corresponding region of the measurement image to calculate pixel values ​​(step S112). The calculated pixel values ​​are stored in association with the coordinates of the filter matrix.

[0101] The processing device 100 determines whether filter matrices have been set at all positions (step S113). If there are remaining positions where the filter matrices are not set (NO in step S113), the processing device 100 shifts the filter matrices (step S114) and repeats the processing from step S112 and below.

[0102] If the filter matrices have been set at all positions (YES in step S113), the processing device 100 constructs a corrected image from the pixel values calculated in step S112 (step S115). Then, the process returns.

[0103] FIG. 9 is a diagram showing an example of a measurement image captured by the measurement camera 14 of the headlight tester 1 according to the present embodiment. The measurement image shown in FIG. 9 shows a screen projection image of the oncoming vehicle headlight. The measurement image shown in FIG. 9 has an image size adjusted to 256 pixels × 256 pixels as an example.

[0104] FIG. 10 is a diagram showing an example of a contour display of the measurement image shown in FIG. 9 and a corrected image generated from the measurement image shown in FIG. 9. FIG. 10(A) shows an example of a contour display of the pixel values (luminance) of the measurement image shown in FIG. 9. FIG. 10(B) shows an example of a contour display of the pixel values of the corrected image generated from the measurement image shown in FIG. 9.

[0105] When actually confirmed for a plurality of people, it was found that the cut-off perceived from the measurement image shown in FIG. 9 is closer to the contour display shown in FIG. 10(B) than in FIG. 10(A). That is, it is shown that there is a possibility that a result closer to the result of the test performed by an inspector visually observing the screen projection image can be output by performing the preprocessing.

[0106] <D. Cut-off determination process (step S20)> Next, a process for determining the cutoff (light-dark boundary point group) present in the measurement image (or corrected image) will be described. The following process can be applied to the corrected image obtained by the pre-processing considering the above-mentioned human contrast sensitivity characteristics, and can also be applied to the measurement image. For convenience of explanation, an example of processing the measurement image will be described, but the corrected image may also be processed.

[0107] Fig. 11 is a schematic diagram of a measurement image captured by the measurement camera 14 of the headlamp tester 1 according to the present embodiment. The measurement image shown in Fig. 11 shows an example of the light distribution characteristics of a passing headlamp. In Fig. 11, the X and Y directions correspond to the imaging plane of the measurement camera 14, and the Z direction corresponds to the pixel value of each pixel.

[0108] 11, a horizontal cutoff and a diagonal cutoff are determined based on the cutoffs present in the measurement image. The intersection of the horizontal cutoff and the diagonal cutoff is determined as the elbow point.

[0109] In recent years, headlights for passing vehicles with light distribution characteristics in which part of the cutoff is blurred have been put into practical use to prevent dazzling of oncoming vehicles even when the vehicle moves up and down significantly. In the measurement image of such a headlight for passing vehicles, as shown in Fig. 11, part of the cutoff shows a curved shape (for example, a concave shape shown in Fig. 11) rather than a straight line.

[0110] According to the prior art, a recess in a measurement image may be erroneously recognized as an elbow point. In addition, an elbow point obtained by calculation including unclear parts may be significantly different from an elbow point determined by visual inspection. Furthermore, when the cutoff itself is unclear or there is a lot of noise near the cutoff, the cutoff may not be determined accurately. The cutoff determination process described below can solve such problems.

[0111] 12 is a diagram showing an example of a light-dark profile in the Y direction of the measurement image shown in FIG. 11. The light-dark profile shown in FIG. 12 is a diagram showing an example of a light-dark profile in the Y direction of the measurement image shown in FIG. i ,y0) to coordinate (x i ,y n ) shows the change in luminosity along the

[0112] Referring to FIG. 12, at position x i In the Y-direction brightness profile at x, the point where the change in pixel value (luminance) is maximum is determined to be the cutoff point (brightness boundary point). i After the light-dark boundary points are determined for , a portion of the set of determined light-dark boundary points is determined as horizontal cutoffs and another portion is determined as diagonal cutoffs.

[0113] For example, a method is known in which the luminous intensity is converted into a logarithm according to the Weber-Fechner law, and the point at which the change in the logarithm is maximum (hereinafter also referred to as the "point of maximum luminous intensity change") is determined as the light-dark boundary point.

[0114] In an actual measurement image, noise may occur in a near-cutoff region 70 (a range that may include the light-dark boundary point) of the light-dark profile shown in Fig. 12. Generally, when an arithmetic operation such as differentiation is performed to calculate the amount of change in luminous intensity, the change due to noise is emphasized, which may make analysis difficult. For this reason, smoothing such as moving average is also used.

[0115] Although the moving average reduces the change due to noise, the profile shape may be distorted. As a result, it may not be possible to determine the correct light-dark boundary point.

[0116] In addition, a method is also known in which the points where the luminous intensity is minimum and maximum in the light-dark profile are determined, and then the intermediate point between the two is determined as the light-dark boundary point. Although this method can be realized with a simple calculation algorithm, there is a possibility that the correct light-dark boundary point cannot be determined.

[0117] In the measurement process according to this embodiment, the maximum luminous intensity change point is determined by the following numerical process.

[0118] As shown in Figure 12, when the luminous intensity Z for each position in the Y direction is plotted, j Let us assume a range of (2m+1) points centered on . The position y k Luminous intensity z k c is expressed as a quadratic polynomial as shown in the following equation (7).

[0119]

number

[0120] At this time, the actual measured luminosity z k t and luminosity z k c The sum of squared residuals (evaluation function) with can be expressed as the following equation (8).

[0121]

number

[0122] Using the least squares method, the values ​​of the coefficients a, b, and c can be determined so that the sum of squared residuals shown in equation (8) is minimized. In this case, the position y j The value of the quadratic polynomial at luminosity z j c =c.

[0123] In other words, determining the values ​​of the coefficients a, b, and c that minimize the sum of squared residuals shown in equation (8) means smoothing using a polynomial fitting method (quadratic polynomial). Smoothing using this type of polynomial fitting method can reduce the effects of noise.

[0124] The first derivative of the above equation (7) is given by equation (9).

[0125]

number

[0126] Center point position y j At luminosity z j c The first derivative of is equal to the coefficient b. In other words, once the coefficients b and c are determined, the luminous intensity z j c The first derivative of is also determined at the same time.

[0127] Generally, in a method of calculating the amount of change in luminance from the difference in luminance between adjacent pixels, the change due to noise is emphasized, which may make analysis difficult. In contrast, the numerical processing described above calculates the differential while smoothing, and can determine the point at which the amount of change in luminance is maximum while reducing the influence of noise. In addition, although smoothing may distort the profile shape, the distortion of the profile shape can be suppressed by using a polynomial fitting method with a higher degree. As a result, the light / dark boundary point can be determined more accurately.

[0128] As a specific processing procedure, it is not necessary to calculate the sum of squared residuals shown in equation (8) at each point of the light-dark profile so that it is minimized. j t The luminosity z is calculated by convolving the (2m+1) points centered on j c For example, we can calculate the derivative of luminosity z j c The first derivative of can be calculated as shown in equation (10).

[0129]

number

[0130] luminous intensity z j c The second derivative of can be calculated in a similar manner.

[0131] Fig. 13 shows an example of a process for determining the maximum luminous intensity change point using a polynomial fitting method according to the present embodiment. Fig. 13(A) shows an example of a brightness profile. Fig. 13(B) shows an example of a first-order differential result of the brightness profile shown in Fig. 13(A). Fig. 13(C) shows an example of a second-order differential result of the brightness profile shown in Fig. 13(A).

[0132] As shown in Fig. 13(A), a case is assumed where a maximum luminous intensity change point is determined in a brightness profile. As shown in Fig. 13(B), in the first derivative of the brightness profile, the maximum luminous intensity change point corresponds to a local maximum value (peak). According to the polynomial fitting method using the quadratic polynomial shown in equation (7), the first derivative of the brightness profile is determined at the position y j The set of values ​​of coefficient b in

[0133] As shown in FIG. 13C, in the second derivative of the brightness profile, the maximum brightness change point corresponds to the zero crossing point where the sign changes from positive to negative. According to the polynomial fitting method using the quadratic polynomial shown in Equation (7), the second derivative of the brightness profile is j It is the set of values ​​of coefficient a×2 in

[0134] If the actually measured brightness profile is directly subjected to second-order differentiation, the change due to noise is overemphasized and the second-order differentiation of the brightness profile cannot be obtained correctly. In contrast, by using the polynomial fitting method as in the present embodiment, the second-order differentiation of the brightness profile can be obtained correctly.

[0135] As mentioned above, we use the polynomial fitting method to find the position x i The light / dark boundary point is determined for each pixel.

[0136] FIG. 14 is a flowchart showing a more detailed procedure of the process (step S20) of determining the cutoff shown in FIG.

[0137] 14, the processing device 100 repeats the processes of steps S202 to S207 at each position along the X direction (corresponding to the first direction) of the measurement image (or the corrected image).

[0138] More specifically, the processing device 100 detects one point in the X direction (position x i ) (step S201), and for the selected point, a brightness profile in the Y direction of the measurement image (or the corrected image) is generated (step S202). That is, the processing device 100 generates a brightness profile indicating a change in luminance in the Y direction (corresponding to the second direction) of the measurement image (or the corrected image).

[0139] The processing device 100 then calculates, for each light intensity included in the light-dark profile, a derivative of the light-dark profile based on a polynomial that represents each light intensity and a plurality of light intensities adjacent to each light intensity.

[0140] More specifically, the processing device 100 detects a point in the light-dark profile (position y j ) (step S203), and the luminosity z j t and the luminosity of the 2m neighbors (luminosity z j-m t ~z j-1 t and luminosity z j+1 t ~z j+m t ) is generated (step S204), and a set of luminosities is generated by convolution integration using the weighting coefficient table W. j Luminous intensity z at j The differential of is calculated (step S205).

[0141] The processing device 100 judges whether or not all points in the target light-dark profile have been selected (step S206). If any unselected points remain in the target light-dark profile (NO in step S206), the processing from step S203 onwards is repeated.

[0142] In this way, the processing device 100 calculates the differentiation of the light and dark profile by performing a convolution integral on each light intensity included in the light and dark profile and a plurality of light intensities adjacent to each light intensity using the weight coefficient table W.

[0143] If all points of the target light and dark profile have been selected (YES in step S206), the processing device 100 determines the point at which the change amount of the light intensity of the light and dark profile becomes the maximum value as the light and dark boundary point. More specifically, the processing device 100 selects the maximum light intensity change point (position y b ) in the generated first-order differentiation and / or second-order differentiation as the light and dark boundary point of the position x i being selected (step S207).

[0144] When determining the light and dark boundary point, the maximum light intensity change point may be determined after logarithmizing the light and dark profile, the first-order differentiation of the light and dark profile, and / or the second-order differentiation of the light and dark profile.

[0145] The processing device 100 determines whether all points in the X direction of the target measurement image have been selected (step S208). If there are remaining unselected points in the X direction of the target measurement image (NO in step S208), the processing below step S201 is repeated.

[0146] If all points in the X direction of the target measurement image have been selected (YES in step S208), the processing device 100 outputs a light and dark boundary point group composed of a set of the position x i and the light and dark boundary point (position y b ) (step S209). Then, the process returns.

[0147] <E. Determination process of elbow point (step S30)> Next, a process of determining an elbow point based on the set of light and dark boundary points (light and dark boundary point group: cut-off) determined for each position x i will be described.

[0148] After the light-dark boundary point group is determined, point groups corresponding to the horizontal cutoff and the diagonal cutoff are determined from the light-dark boundary point group, and an intersection point between the horizontal cutoff and the diagonal cutoff is determined as an elbow point.

[0149] In one example of the prior art, a tentative elbow point is calculated, and then several points on either side of the tentative elbow point are extracted, and a horizontal cutoff and a diagonal cutoff are determined by a linear regression equation. The intersection of the determined horizontal cutoff and the diagonal cutoff is determined as the elbow point.

[0150] However, according to such prior art, the linear regression equations representing the horizontal cutoff and the diagonal cutoff may vary depending on the method of selecting the points on either side of the tentative elbow point. In particular, when the light-dark boundary point group (cutoff) represents a curved shape with curvature, there is a high degree of arbitrariness, making it difficult to accurately determine the linear regression equation.

[0151] In the measurement process according to this embodiment, a two-piece linear function consisting of two piece lines centered on the elbow point is adopted, based on the premise that the intersection of two clear lines (horizontal cutoff and diagonal cutoff) is the elbow point.

[0152] If the coordinates of the intersection indicating the elbow point are (x1, y1), the coordinates of the left end of the left line are (x0, y0), and the coordinates of the right end of the right line are (x2, y2), then equation (11) is established.

[0153]

number

[0154] However, the light-dark boundary point group (cutoff) is not expressed by two simple straight lines, but may have complex shapes such as convex shapes, concave shapes, wavy shapes, etc. Therefore, as shown in formula (12), an n-piecewise linear function consisting of n-piece straight lines may be adopted.

[0155]

number

[0156] where (x0, y0) is the coordinate of the left end of the first section, and (x n ,y n ) is the coordinate of the right end of the nth division.

[0157] Using the nonlinear least squares method, equation (11) or (12) is fitted to the light-dark boundary point group. In other words, the parameters of equation (11) or (12) are determined. The parameters determined using the nonlinear least squares method are calculated by machine learning so as to minimize the sum of squared residuals (evaluation function), and are statistical values ​​with little arbitrariness.

[0158] The nonlinear least squares method requires initial parameters, since iterative calculations are performed starting from the initial parameters. x0 and x n The initial parameters of x1 to x2 may be the x coordinates of the left and right ends of the light-dark boundary point group. n-1 The initial parameters of y0 to y1 are set to the x coordinates of the light-dark boundary points divided into n parts. n-1 The initial parameters can be calculated by using the central value of the measurement image as the initial parameters.

[0159] The Levenberg-Marquardt algorithm may be used as the nonlinear least squares calculation algorithm. By setting optimal damping parameters and applying a singularity resolution algorithm, a robust and fast convergence calculation can be achieved.

[0160] Fig. 15 is a diagram showing an example of the process of determining an elbow point according to the present embodiment. Fig. 15 shows the result of determining parameters of a seven-piecewise linear function (i.e., the number of pieces n=7) by the nonlinear least squares method for a light-dark boundary point group calculated from a measurement image in which part of the horizontal cutoff is unclear.

[0161] 15, both ends of the light-dark boundary point group are determined to be coordinates (x0, y0) and (x7, y7), respectively. Six points from the light-dark boundary point group excluding both ends are determined as bending points (coordinates (x1, y1) to (x6, y6)).

[0162] Of the determined bending points, a score is calculated based on, for example, the following points:

[0163] (a) An inflection point where the slope of the line connecting the adjacent inflection points is negative (b) A bending point with a large change in y-coordinate between adjacent bending points In the example shown in Figure 15, the bend point at coordinate (x2, y2) has the highest score and is determined to be the tentative elbow point. At this time, attention is paid not to the tentative elbow point itself, but to the coordinates (x1, y1) and (x3, y3) adjacent to the tentative elbow point. In other words, it can be determined that a horizontal cutoff and a diagonal cutoff exist between the two bend points adjacent to the tentative elbow point.

[0164] Therefore, the elbow point is determined by fitting a two-piecewise linear function (that is, the number of pieces n=2) to a piece defined by two bending points adjacent to the tentative elbow point.

[0165] Note that the light-dark boundary points existing between coordinates (x4, y4) and (x7, y7) do not affect the determination of the elbow point. Therefore, the horizontal cutoff and diagonal cutoff are determined from the parameters of the two-piecewise linear function determined by the nonlinear least squares method, and there is no problem even if the sum of squared residuals between the determined horizontal cutoff or diagonal cutoff is large.

[0166] In this manner, in the process of determining an elbow point according to the present embodiment, the range of the light-dark boundary point group used to determine the horizontal cutoff and the diagonal cutoff for the elbow point is determined first. The horizontal cutoff and the diagonal cutoff are determined using the light-dark boundary point group included in the determined range.

[0167] In addition, a nonlinear least squares method is used to fit an n-piecewise linear function to the light-dark boundary point group to determine both ends (coordinates (x0, y0) and (x7, y7)) and six bending points (coordinates (x1, y1) to (x6, y6)). Then, the slopes of the lines calculated for adjacent sections (section i-1 and section i) according to equation (13) are calculated as follows: i-1 and S i may be calculated.

[0168]

number

[0169] Here, the slope S i-1 ≒ Tilt S i If so, it can be determined that the adjacent sections have the same inclination. Therefore, the two target sections can be combined into one section and the calculation can be performed again. In this way, the number of sections can be reduced one by one to determine the provisional elbow.

[0170] FIG. 16 is a flowchart showing a more detailed procedure of the process (step S30) of determining the elbow point shown in FIG.

[0171] 16, the processing device 100 sets the number of divisions to an initial value (step S301). For example, the number of divisions n may be set to 7 as the initial value. The processing device 100 calculates initial parameters from the light-dark boundary point cloud according to the set number of divisions (step S302).

[0172] The processing device 100 starts parameter fitting by the nonlinear least squares method (step S303). The processing device 100 determines whether the sum of squared residuals (evaluation function) calculated for the current parameters satisfies a convergence condition (step S304).

[0173] If the sum of squared residuals does not satisfy the convergence condition (NO in step S304), the processing device 100 updates the current parameters (step S305) and repeats the processes in and after step S304.

[0174] If the number of iterative calculations exceeds a predetermined threshold, that is, if convergence does not occur within the predetermined threshold, the number of sections may be increased.

[0175] If the sum of squared residuals satisfies the convergence condition (YES in step S304), the processing device 100 determines both ends and inflection points using the current parameters (step S306).

[0176] In this manner, the processing device 100 determines the segments included in the light-dark boundary point group by fitting a piecewise linear function to the light-dark boundary point group.

[0177] Next, the processing device 100 calculates the slope of each segment (step S307). The processing device 100 judges whether any adjacent segments have a similar slope (step S308). If any adjacent segments have a similar slope (YES in step S308), the processing device 100 reduces the number of segments (step S309) and reduces the parameters corresponding to the segments having a similar slope (step S310). Then, the processing from step S303 onwards is repeated.

[0178] The processing device 100 merges two adjacent sections having similar slopes among the determined sections, and then repeats the process of determining the section (steps S303 to S310).

[0179] If none of the adjacent sections have the same slope (NO in step S308), the processing device 100 calculates a score for each determined bend point (step S311) and determines the bend point with the highest score as the provisional elbow point (step S312). In this way, the processing device 100 determines the bend point with the highest score calculated according to a predetermined criterion among the bend points defining the determined section as the provisional elbow point.

[0180] The processing device 100 calculates coordinates (bending points) adjacent to both sides of the determined temporary elbow point, and determines the range of the light and dark boundary point group used to determine the horizontal cutoff and the diagonal cutoff (step S313).

[0181] The processing device 100 sets the number of divisions to "2" (step S314). The processing device 100 sets both ends of the determined light and dark boundary point group and the temporary elbow point as initial parameters (step S315).

[0182] The processing device 100 starts parameter fitting by the non-linear least squares method (step S316). The processing device 100 determines whether the sum of squared residuals (evaluation function) calculated with the current parameters satisfies the convergence condition (step S317).

[0183] If the sum of squared residuals does not satisfy the convergence condition (NO in step S317), the processing device 100 updates the current parameters (step S318) and repeats the processing from step S317 and below.

[0184] If the sum of squared residuals satisfies the convergence condition (YES in step S317), the processing device 100 determines the elbow point using the current parameters (step S319). Then, the process returns.

[0185] In this way, the processing device 100 determines the elbow point by fitting a two-segment linear function to the range including the temporary elbow point and the bending points adjacent to both sides of the temporary elbow point.

[0186] The processing device 100 may determine whether the position of the determined elbow point exists within a predetermined pass region.

[0187] <F. Effect> Next, an example of the effect of the measurement process according to the present embodiment will be described.

[0188] Fig. 17 is a diagram showing an example of a corrected image generated by pre-processing according to the present embodiment. Fig. 17 shows a corrected image obtained by pre-processing a measurement image showing a screen projected image of a passing headlamp.

[0189] The coordinates of the corrected image shown in FIG. i ,y0) and coordinates (x i ,y n ) the cutoff determination process according to the present embodiment was applied to the light / dark profile between the

[0190] 18A and 18B are diagrams showing an example of the effect of pre-processing on the cutoff determination process according to the present embodiment. Fig. 18A shows a profile obtained by logarithmizing the luminosity of the light-dark profile of the measurement image. Fig. 18B shows a profile obtained by logarithmizing the luminosity of the light-dark profile of the correction image.

[0191] Compared to the Y-logZ profile (FIG. 18(A)) obtained by logarithmizing the luminance of the light-dark profile of the measured image, the Y-logZ profile (FIG. 18(B)) obtained by logarithmizing the luminance of the light-dark profile of the corrected image shows a large drop in luminance just before approaching the light-dark boundary point. In other words, the profile shown in FIG. 18(B) includes a concave portion. This characteristic corresponds to a phenomenon called Mach bands. Mach bands are a type of optical illusion in which, when a light area and a dark area with different luminance are adjacent to each other, the human eye perceives the area near the boundary on the dark side as darker and the area near the boundary on the light side as brighter. This optical illusion is caused by the contrast sensitivity function of the human eye. The contrast sensitivity function can be said to be a kind of edge enhancement filter that is inherently provided in the human eye to enable the human eye to perceive areas with ambiguous light and dark more clearly.

[0192] That is, the profile shown in FIG. 18(B) takes into consideration the contrast sensitivity function of the human eye.

[0193] Even when determining the light-dark boundary point, it is considered that determining the maximum light intensity change point in a state closer to that of the human eye can determine a cut-off closer to human perception.

[0194] FIG. 19 is a diagram showing an example of the effect on the determination process of the elbow point according to the present embodiment. FIG. 19 shows a cut-off (logZ) determined from a profile obtained by logarithmizing the light intensity of the light-dark profile of the measurement image, and a cut-off (logZc) determined from a profile obtained by logarithmizing the light intensity of the light-dark profile of the corrected image.

[0195] When actually checking a plurality of people, as a result, it was found that the cut-off perceived from the image shown in FIG. 19 is closer to logZc than logZ. That is, it is shown that there is a possibility that a result closer to the result of the test performed by the inspector visually observing the screen projection image can be output by performing the preprocessing.

[0196] <G. Advantages> According to the present embodiment, the deviation between the result obtained based on the image of the screen projection image captured by the camera and the result obtained by visually judging the screen projection image can be reduced.

[0197] According to the present embodiment, when determining the light-dark boundary point, since the differential is calculated while smoothing the light-dark profile, the point where the amount of change in light intensity is maximum can be determined while reducing the influence of noise. Also, by using the polynomial fitting method with an increased degree, the distortion of the profile shape can be suppressed. As a result, the light-dark boundary point can be determined more accurately.

[0198] According to the present embodiment, by fitting a piecewise linear function to the light-dark boundary point group, the sections included in the light-dark boundary point group are determined. Based on the determined sections, after determining the provisional elbow point, by fitting a two-section linear function, the elbow point can be determined more accurately.

[0199] The embodiments disclosed herein should be considered to be illustrative and not restrictive in all respects. The scope of the present invention is defined by the claims, not the above description, and is intended to include all modifications within the scope and meaning equivalent to the claims. [Explanation of symbols]

[0200] 1 Headlamp tester, 4 Pillar section, 6 Base section, 10 Lens unit, 12 Fresnel lens, 13 Screen, 14 Measurement camera, 15 Orientation confirmation camera, 16 Orientation adjustment laser, 20 Operation section, 30 Display section, 40 Drive device, 60 Filter matrix, 70 Cutoff vicinity area, 100 Processing device, 102 Processor, 104 Memory, 106 Storage, 108 System program, 110 Measurement program, S Headlamp, V Vehicle.

Claims

1. a screen onto which light from the vehicle's headlights is projected; a camera for capturing an image projected onto the screen; a processing device that determines a light-dark boundary point cloud that exists in the measurement image captured by the camera, and determines an elbow point based on the determined light-dark boundary point cloud; The processing device corrects the measurement image based on human contrast sensitivity characteristics and then determines the elbow point.

2. 2. The headlamp tester according to claim 1, wherein the processing device generates a corrected image by correcting the measured image by sequentially calculating a convolution integral of the measured image and a filter matrix corresponding to the contrast sensitivity characteristic of the human.

3. The process of determining the light-dark boundary point group includes: At each position along the first direction of the measurement image, generating a light-dark profile indicative of a change in light intensity in a second direction of the measurement image; and calculating, for each light intensity included in the light-dark profile, a derivative of the light-dark profile based on a polynomial that represents each light intensity and a plurality of light intensities adjacent to each light intensity.

4. The headlamp tester according to claim 3 , wherein the process of determining the light-dark boundary point group includes a process of determining a point at which a change in luminous intensity of the light-dark profile is a maximum value as the light-dark boundary point.

5. 4. The headlamp tester according to claim 3, wherein the process of calculating the derivative of the light-dark profile includes a process of convoluting each luminous intensity included in the light-dark profile and a plurality of luminous intensities adjacent to each luminous intensity using a weighting coefficient table.

6. The process of determining the light-dark boundary point group includes: A process of determining a segment included in the light-dark boundary point group by fitting a piecewise linear function to the light-dark boundary point group; determining, as a provisional elbow point, a bending point having the highest score calculated according to a predetermined criterion among the bending points defining the determined section; and determining the elbow point by fitting a two-piecewise linear function to a range including the provisional elbow point and bending points adjacent to both sides of the provisional elbow point.

7. The process of determining the light-dark boundary point group includes: The headlamp tester according to claim 6 , further comprising a process of integrating two adjacent sections having similar slopes from among the determined sections, and repeating the process of determining the sections.

8. acquiring a measurement image obtained by capturing an image of light from a vehicle headlight projected onto a screen; correcting the measured image with human contrast sensitivity characteristics; determining a light-dark boundary point group present in the corrected measurement image; and determining an elbow point based on the determined light-dark boundary point cloud.

9. On the computer, acquiring a measurement image obtained by capturing an image of light from a vehicle headlight projected onto a screen; correcting the measured image with human contrast sensitivity characteristics; determining a light-dark boundary point group present in the corrected measurement image; and determining an elbow point based on the determined light-dark boundary point group.

Images