Method and apparatus for determining retroreflective luminance coefficient, electronic device, and program product

Abstract

The application discloses a retroreflective luminance coefficient determination method and an electronic device. The retroreflective luminance coefficient determination method comprises determining a calibration height parameter for a target line visibility observation model; determining a calibration observation distance for the target line visibility observation model; setting a calibration projection distance for light projection of target line detection; determining a target line detection height parameter for target line detection from the calibration height parameter based on a proportional relationship between the calibration observation distance and the calibration projection distance, the target line detection height parameter comprising a camera height for image acquisition and a light projection height; or, at least partially setting the target line detection height parameter, determining the calibration projection distance for light projection of target line detection from the calibration observation distance based on a proportional relationship between the calibration height parameter and the at least partially set target line detection height parameter; obtaining target line detection data corresponding to a detected position in target line detection; and determining a retroreflective luminance coefficient according to the target line detection data.

Description

[0001] This application is based on the parent application with the application date of September 30, 2024, the application number of 202411375640.0, and the invention name of "Retroreflective Coefficient Determination Method and Device, Electronic Equipment and Program Product". The parent application is a divisional application. TECHNICAL FIELD

[0002] The present application relates to the field of highway traffic, in particular to a retroreflective coefficient determination method and device, electronic equipment and program product. BACKGROUND

[0003] Road traffic marking is a key traffic safety facility painted on the road surface, aiming to channelize traffic flow, induce vehicle line shape, and ensure traffic safety. These markings have multiple functions of defining lane boundaries, regulating and managing driver driving behavior, inducing driver sight, indicating and warning road conditions ahead, and clarifying road use rights.

[0004] After being crushed for a long time, the visibility of road traffic marking decreases, causing the function of the marking to decrease. At this time, maintenance is needed. An important evaluation criterion for determining whether road marking needs maintenance is the retroreflective coefficient of the road traffic marking, which can also be referred to as the retroreflective coefficient of brightness.

[0005] In China, a handheld retroreflective measuring instrument is commonly used to detect the retroreflective coefficient of brightness of the marking. Static detection and evaluation require the closure of the road, which has a great social impact and an immeasurable cost. Dynamic detection and evaluation are fast and have no impact on traffic, but the detection is easily affected by environmental strong light, dynamic geometric angle changes, and other factors, often involving complex settings and strict operating conditions, resulting in the inability to accurately determine the retroreflective coefficient of brightness of the marking.

[0006] The content described in the background technology is only for the convenience of understanding the related technology in the art, and is not regarded as an acknowledgement of the prior art. SUMMARY

[0007] Therefore, the embodiments of the present application hope to provide a retroreflective coefficient determination solution that can at least partially solve the problems described above.

[0008] In a first aspect, a retroreflective coefficient determination method is provided, comprising:

[0009] Obtaining a first image and a second image corresponding to a detected position in marking detection, wherein the first image is an image collected under single-wavelength laser projection, and the second image is an image collected under single-wavelength laser non-projection;

[0010] Identifying a first marking area in the first image using a given image segmentation algorithm;

[0011] identifying a second reticle region in a second image using a given image segmentation algorithm;

[0012] extracting a first gray value of a first reticle region in the first image;

[0013] extracting a second gray value of a second reticle region in the second image;

[0014] calculating a retroreflective luminance coefficient of a reticle at the detected position according to the first gray value, the second gray value, and a given retroreflective luminance coefficient standard value.

[0015] Optionally, the calculating a retroreflective luminance coefficient of a reticle at the detected position according to the first gray value, the second gray value, and a given retroreflective luminance coefficient standard value comprises:

[0016] determining a reticle gray value difference between the first gray value and the second gray value;

[0017] obtaining the retroreflective luminance coefficient standard value corresponding to a given reticle standard plate;

[0018] obtaining a standard gray value difference corresponding to the given reticle standard plate;

[0019] determining the retroreflective luminance coefficient of the reticle at the detected position according to a proportional relationship among the reticle gray value difference, the standard gray value difference, and the retroreflective luminance coefficient standard value.

[0020] Optionally, the method further comprises, at or after the calculating a retroreflective luminance coefficient of a reticle at the detected position:

[0021] correcting the retroreflective luminance coefficient of the reticle to obtain an ambient light corrected retroreflective luminance coefficient, specifically comprising:

[0022] obtaining an ambient light parameter in the reticle detection;

[0023] determining an ambient light correction coefficient according to the ambient light parameter, the reticle gray value difference, and a retroreflective luminance coefficient standard value of the reticle at the detected position;

[0024] correcting the retroreflective luminance coefficient according to the ambient light correction coefficient to obtain an ambient light corrected retroreflective luminance coefficient.

[0025] Optionally, the obtaining a standard gray value difference corresponding to the given reticle standard plate comprises:

[0026] obtaining a third image and a fourth image corresponding to the given reticle standard plate, wherein the third image is an image collected under single-wavelength laser projection, and the fourth image is an image collected under single-wavelength laser non-projection;

[0027] identifying a third target region in the third image using the given image segmentation algorithm;

[0028] identifying a fourth target region in the fourth image using the given image segmentation algorithm;

[0029] extracting a third gray value of the third target region in the third image;

[0030] extracting a fourth gray value of the fourth target region in the fourth image;

[0031] determining the standard gray difference of the third gray value and the fourth gray value.

[0032] Optionally, when or after calculating the retroreflective luminance coefficient of the target line of the detected position, the method further comprises:

[0033] correcting the retroreflective luminance coefficient of the target line to obtain a distance-corrected retroreflective luminance coefficient, specifically comprising:

[0034] obtaining a calibration projection distance and an actual projection distance of the single-wavelength laser projection;

[0035] determining a distance correction coefficient according to the calibration projection distance and the actual projection distance;

[0036] correcting the retroreflective luminance coefficient of the target line according to the distance correction coefficient to obtain a distance-corrected retroreflective luminance coefficient.

[0037] Optionally, when or after calculating the retroreflective luminance coefficient of the target line of the detected position, the method further comprises:

[0038] correcting the retroreflective luminance coefficient of the target line to obtain an ambient light-corrected and distance-corrected retroreflective luminance coefficient, specifically comprising:

[0039] obtaining an ambient light parameter in the target line detection;

[0040] determining an ambient light correction coefficient according to the ambient light parameter, the target line gray difference, and the standard value of the retroreflective luminance coefficient of the target line of the detected position;

[0041] obtaining a calibration projection distance and an actual projection distance of the single-wavelength laser projection;

[0042] determining a distance correction coefficient according to the calibration projection distance and the actual projection distance;

[0043] correcting the retroreflective luminance coefficient of the target line according to the ambient light correction coefficient and the distance correction coefficient to obtain an ambient light-corrected and distance-corrected retroreflective luminance coefficient.

[0044] Optionally, before acquiring the first image and the second image, the method further comprises:

[0045] determining a calibration height parameter for the target line visibility observation model, the calibration height parameter comprising a calibration driver observation height and a calibration vehicle light height;

[0046] determining a calibration observation distance for the target line visibility observation model;

[0047] setting a calibration projection distance for light projection in target line detection, and determining a target line detection height parameter for target line detection from the calibration height parameter based on a proportional relationship between the calibration observation distance and the calibration projection distance, the target line detection height parameter comprising a camera height for image acquisition and a light source height for light projection; or, at least partially setting the target line detection height parameter, and determining the calibration projection distance for light projection in target line detection from the calibration observation distance based on a proportional relationship between the calibration height parameter and the at least partially set target line detection height parameter;

[0048] wherein the first image and the second image are acquired at the calibration projection distance, the camera height and the light projection height.

[0049] In a second aspect, a retroreflective luminance coefficient determination method is provided, comprising:

[0050] determining a calibration height parameter for the target line visibility observation model, the calibration height parameter comprising a calibration driver observation height and a calibration vehicle light height;

[0051] determining a calibration observation distance for the target line visibility observation model;

[0052] setting a calibration projection distance for light projection in target line detection, and determining a target line detection height parameter for target line detection from the calibration height parameter based on a proportional relationship between the calibration observation distance and the calibration projection distance, the target line detection height parameter comprising a camera height for image acquisition and a light source height for light projection; or, at least partially setting the target line detection height parameter, and determining the calibration projection distance for light projection in target line detection from the calibration observation distance based on a proportional relationship between the calibration height parameter and the at least partially set target line detection height parameter;

[0053] acquiring target line detection data corresponding to a detected position in target line detection, wherein images in the target line detection data are acquired at the calibration projection distance, the camera height and the light projection height;

[0054] determining the retroreflective luminance coefficient from the target line detection data.

[0055] In a third aspect, a retroreflective luminance factor evaluation device is provided, comprising:

[0056] an acquisition unit configured to acquire a first image and a second image corresponding to a detected position in a reticle detection, wherein the first image is an image acquired under a single-wavelength laser projection condition, and the second image is an image acquired under a single-wavelength laser non-projection condition;

[0057] a first reticle identification unit configured to identify a first reticle region in the first image using a given image segmentation algorithm;

[0058] a second reticle identification unit configured to identify a second reticle region in the second image using the given image segmentation algorithm;

[0059] a first image processing unit configured to extract a first gray value of the first reticle region in the first image;

[0060] a second image processing unit configured to extract a second gray value of the second reticle region in the second image;

[0061] a retroreflective luminance factor calculation unit configured to calculate a retroreflective luminance factor of a reticle of the detected position according to the first gray value, the second gray value, and a given retroreflective luminance factor standard value.

[0062] Optionally, the acquisition unit is configured to acquire an ambient light parameter value corresponding to the detected position in the reticle detection; and configured to acquire a single-wavelength laser actual projection distance corresponding to the detected position in the reticle detection.

[0063] The retroreflective luminance factor evaluation device further comprises:

[0064] a third processing unit configured to extract an ambient light correction coefficient and a distance correction coefficient of the reticle of the detected position;

[0065] The retroreflective luminance factor calculation unit is configured to calculate a retroreflective luminance factor of the reticle of the detected position according to the first gray value, the second gray value, the ambient light correction coefficient, the distance correction coefficient, and a given retroreflective luminance factor standard value.

[0066] Optionally, the retroreflective luminance factor evaluation device further comprises:

[0067] a retroreflective luminance factor evaluation unit configured to receive the retroreflective luminance factor calculated by the calculation unit, and judge the retroreflective performance of the reticle according to a preset performance index threshold, to evaluate whether the reticle meets a prescribed safety standard.

[0068] In a fourth aspect, an electronic device is provided, comprising: a processor and a memory storing a computer program, wherein the processor is configured to implement the method of embodiments of the present application when running the computer program.

[0069] In a fifth aspect, a program product is provided, comprising a computer program, wherein the computer program, when executed by a processor, implements the method of embodiments of the present application.

[0070] An aspect of the present application provides a retro-reflective luminance coefficient determination scheme, which not only processes two images of single-wavelength laser projection or no projection, but also uses image segmentation algorithms to accurately segment the reticle region for the two images respectively. By extracting the gray value from the reticle region segmented from the reticle image, and combining it with the ambient light parameter, the actual projection distance of light and the standard value of the retro-reflective luminance coefficient of the standard plate, the method can accurately determine the retro-reflective luminance coefficient of the object to be measured.

[0071] Another aspect of the present application provides a retro-reflective luminance coefficient determination scheme, which can construct a reticle detection model based on light projection in proportion to the reticle visibility observation model, for determining the retro-reflective luminance coefficient. This method effectively guarantees the accuracy of the detection result while reducing the complexity of measurement and limiting the measurement equipment, thereby ensuring the accuracy of the determined retro-reflective luminance coefficient.

[0072] Some optional features and other effects of embodiments of the present application are described below, and some can be apparent from reading this document. BRIEF DESCRIPTION OF DRAWINGS

[0073] In the following, embodiments of the present application will be described in detail with reference to the accompanying drawings, wherein the elements shown are not limited by the proportions shown in the drawings and the same or similar reference numerals in the drawings represent the same or similar elements, and wherein:

[0074] Figure 1 A schematic diagram of a reticle detection device according to an embodiment of the present application is shown, which can acquire reticle detection data for the retro-reflective luminance coefficient determination scheme according to an embodiment of the present application;

[0075] Figure 2 An example diagram of an image processed by the retro-reflective luminance coefficient determination scheme according to an embodiment of the present application is shown;

[0076] Figure 3 An example diagram of an image processed by the retro-reflective luminance coefficient determination scheme according to an embodiment of the present application is shown;

[0077] Figure 4 An example flowchart of a retro-reflective luminance coefficient determination method according to an embodiment of the present application is shown;

[0078] Figure 5 An exemplary flowchart of a calibration process of the retroreflective luminance coefficient determination method according to an embodiment of the present application is shown;

[0079] Figure 6 An exemplary structural diagram of a calibration model of the retroreflective luminance coefficient determination method according to an embodiment of the present application is shown; Figure 5

[0080] Figure 7 Another exemplary structural diagram of the calibration model shown in Figure 6

[0081] Figure 8 An exemplary structural diagram of a feature extraction convolution module of the calibration model shown in Figure 6

[0082] An exemplary flowchart of a calibration process of the retroreflective luminance coefficient determination method according to an embodiment of the present application is shown; Figure 9

[0083] An exemplary flowchart of a calibration process of the retroreflective luminance coefficient determination method according to an embodiment of the present application is shown; Figure 10

[0084] An exemplary flowchart of a calibration process of the retroreflective luminance coefficient determination method according to an embodiment of the present application is shown; Figure 11

[0085] An exemplary flowchart of a calibration process of the retroreflective luminance coefficient determination method according to an embodiment of the present application is shown; Figure 12

[0086] An exemplary flowchart of a calibration process of the retroreflective luminance coefficient determination method according to an embodiment of the present application is shown; Figure 13

[0087] An exemplary structural diagram of a calibration model of the retroreflective luminance coefficient determination method according to an embodiment of the present application is shown; and Figure 14

[0088] An exemplary structural diagram of an electronic device that can be used to implement the retroreflective luminance coefficient determination method according to an embodiment of the present application is shown. Figure 15 DETAILED DESCRIPTION

[0089] To make the objects, technical solutions and advantages of the present application clearer, further detailed description of the present application is given below in conjunction with specific embodiments and drawings. Herein, the exemplary embodiments of the present application and their descriptions are used to explain the present application, but not to limit the present application.

[0090] ​​​The term "includes" and its variants are meant to cover a non-exclusive inclusion, such that a process, method, system, product, or apparatus that comprises a list of elements is not necessarily limited to those elements, but can include other elements not expressly listed or inherent to such process, method, system, product, or apparatus. The term "or" means "and / or" unless expressly indicated otherwise. The term "based on" means "based, at least in part, on" unless expressly indicated otherwise. The terms "one example embodiment" and "an embodiment" mean "at least one example embodiment." The term "another embodiment" means "at least one additional example embodiment." For ease of understanding, the ordinal terms "first," "second," and the like do not by themselves indicate any sequence or order in time or in plan. Rather, they serve as identifiers

[0091] Provided in the present application is a retroreflective luminance coefficient determination scheme capable of accurately determining the retroreflective luminance coefficient of a reticle, particularly relates to a retroreflective luminance coefficient determination method, a retroreflective luminance coefficient determination device, and an electronic device, a program product, and a storage medium capable of implementing the retroreflective luminance coefficient determination method.

[0092] As described above, the retroreflective luminance coefficient determination scheme in one aspect of the present application not only processes two images with or without single-wavelength laser projection, but also precisely segments the reticle region using an image segmentation algorithm for each of the two images. The retroreflective luminance coefficient determination scheme in another aspect of the present application constructs a reticle detection model based on light projection in proportion to a reticle visibility observation model, for determining the retroreflective luminance coefficient. It will be appreciated that embodiments of the present application are intended to cover any combination scheme involving any of the above aspects or both aspects, which falls within the scope of the present application.

[0093] In the retroreflective luminance coefficient determination scheme of the embodiments of the present application, optional or preferred features / schemes can also be included, which are innovative by themselves and can also be applied to other applications beyond the determination of retroreflective luminance coefficient. Therefore, those skilled in the art will appreciate that these optional or preferred features can be claimed in separate or subsequent applications, independent of the retroreflective luminance coefficient determination scheme. These optional or preferred features / schemes include but are not limited to: a reticle detection scheme for acquiring reticle detection data, a reticle identification scheme for identifying a reticle region, a correction scheme for correcting retroreflective luminance coefficient. Conversely, those skilled in the art will also appreciate that the retroreflective luminance coefficient determination scheme of different embodiments of the present application can or can not include (involve) these optional or preferred features / schemes, and can include (involve) all or part of these optional or preferred features / schemes, as will be further explained below.

[0094] As mentioned before, in the preferred embodiments of the present application, the reticle detection data for determining the reticle retroreflective luminance coefficient can be detected by a reticle detection device according to the embodiments of the present application.

[0095] Referring to Figure 1 , a reticle detection device 100 of this embodiment is shown, which is configured to detect reticle detection data for determining the reticle retroreflective luminance coefficient. However, as mentioned before, the retroreflective luminance coefficient determination method according to the embodiments of the present application can acquire the relevant reticle detection data using other devices or other means, which falls within the scope of the present application.

[0096] Continuing to refer to Figure 1 , the reticle detection device 100 can include a camera 110 and a laser source 120. As shown in Figure 1 , the camera 110 can be configured to capture images of the detected location at a downwardly tilted viewing angle. The laser source 120 can be configured to project single-wavelength laser light at a downwardly tilted projection angle toward the detected location. In one preferred embodiment, the single-wavelength laser light can be infrared laser light or green laser light. In one preferred example, the wavelength of the single-wavelength laser light can be about 808 nm. The detection data of the embodiments of the present application, which include both single-wavelength laser light projection and no projection of images, are particularly advantageous for detecting or identifying reticles, especially in images under outdoor natural light environment.

[0097] In some embodiments of the present application, the reticle detection device further includes a filter mounted on the camera, the wavelength of the filter corresponding to the wavelength of the single-wavelength laser light projected by the laser source. This further helps to detect or identify reticles in images under outdoor natural light environment.

[0098] Referring to Figures 1 to 3, the reticle detection device 100 can include a first mode and a second mode. In the first mode, the camera 110 captures images with the laser source 120 projecting laser light (not shown in the figure) Figure 1 ), and in the second mode, the camera 110 captures images without the laser source 120 projecting laser light (not shown in the figure). In this way, the reticle detection device 100 is able to detect reticle detection data corresponding to the detected position, which includes a first image (as shown in Figure 2 ) captured in the first mode and a second image (as shown in Figure 3 ) captured in the second mode.

[0099] With continued reference to Figure 1 , the camera 110 can be positioned above the laser source 120. Although not shown in the figure, the camera 110 and the laser source 120 have substantially the same lateral position in a direction transverse to the projection direction of the laser source 120, i.e., the camera 110 is substantially positioned directly above the corresponding laser source 120.

[0100] As previously mentioned, a reticle visibility observation model can be used to construct a light projection based reticle detection model in a scale. Further, the setting angle and height of the camera 110 and the laser source 120 of the reticle detection device 100 can be determined according to the reticle detection model.

[0101] In a preferred embodiment, the reticle visibility observation model can include a calibration height parameter, which includes a calibration driver observation height H d and a calibration vehicle light height H cl . In an embodiment, the calibration driver observation height can represent the height of the eyes of a simulated driver driving a small passenger car. In an example, the calibration driver observation height H d may be 120 cm. In an embodiment, the calibration vehicle light height H cl may represent the height of the front light of a small passenger car. In an example, the calibration vehicle light height H cl may be 65 cm.

[0102] In a preferred embodiment, the reticle visibility observation model can include a calibration observation distance D d , which can represent the farthest distance of visual observation of the eyes of a simulated driver driving a small passenger car, and assumes that the vehicle light projects the same position, i.e., the vehicle light projection distance D cl is equal to the calibration observation distance. In an example, the calibration observation distance Dd and vehicle light projection distance D cl are both 30 m.

[0103] In a preferred embodiment, the light projection based lane detection model can comprise a calibration projection distance D s and a lane detection height parameter of the light projection based lane detection model can comprise a camera height H c and a light projection height of the laser source H l .

[0104] In a preferred embodiment of the present application, the calibration projection distance D s may be pre-set as needed, and the lane detection height parameter can be determined from the calibration height parameter according to a proportional relationship between the pre-set calibration observation distance and the calibration projection distance. Further, the camera height can be determined from the driver observation height, and / or the light projection height can be determined from the calibration vehicle light height, according to the proportional relationship between the pre-set calibration observation distance and the calibration projection distance.

[0105] In this embodiment, in combination with reference to Figure 1 , the lane detection height parameter can be determined according to the following formulas (1) and (2):

[0106] H c =H d *D s / D d (1)

[0107] H l =H cl *D s / D d (2)

[0108] In an example, the camera height of the camera and the light source height of the laser source are in a proportion of 24 / 13±5%, more preferably, the camera height is 24 cm±0.5 cm, and the light source height is 13 cm±0.5 cm.

[0109] In another preferred embodiment of the present application, the alignment mark detection height parameter can be at least partially preset as needed, for example, the camera height and / or the light source height can be preset, and the calibration projection distance can be determined according to the proportional relationship between the preset alignment mark detection height parameter (camera height and / or light source height) and the calibration height parameter (calibration driver observation height and / or calibration vehicle light height). In this preferred embodiment, the light source height can be preset, and the calibration projection distance can be determined according to the proportional relationship between the alignment mark detection height parameter and the calibration height parameter; or the camera height can be preset, and the light source height can be determined according to the proportional relationship between the calibration driver observation height and the calibration vehicle light height, and then the calibration projection distance can be determined according to the proportional relationship between the alignment mark detection height parameter and the calibration height parameter. However, other specific solutions are also conceivable as long as they conform to the determination of the calibration projection distance according to the proportional relationship between the at least partially preset alignment mark detection height parameter and the calibration height parameter.

[0110] In a preferred embodiment, the alignment mark visibility observation model can include an observation angle parameter, and the calibration angle parameter can include a driver observation angle and a vehicle light projection angle. In an embodiment, the driver observation angle can represent the angle of the farthest position (also the vehicle light projection position) of the simulated driver visual observation of the small passenger car relative to the road surface. In an embodiment, the vehicle light projection angle can be the angle of the front headlight to the vehicle light projection position relative to the road surface.

[0111] In a preferred embodiment, the light projection based alignment mark detection model can further include an alignment mark detection angle parameter, which can include a camera observation angle and a laser source projection angle.

[0112] In a preferred embodiment of the present application, the camera observation angle can also be determined according to the driver observation angle. In an example, the camera observation angle is 2.29°±0.05°. In a preferred embodiment of the present application, the laser source projection angle can also be determined according to the vehicle light projection angle. In an example, the laser source projection angle is 1.24°±0.05°.

[0113] In the above preferred embodiments, the correspondence of the parameters of the light projection based alignment mark detection model to the alignment mark visibility observation model is described. It will be appreciated that the above correspondence of the parameters is a general correspondence, which is intended to cover any equivalents and / or fluctuations falling within the scope of the application. Specifically, the above correspondence explicitly covers a fluctuation range within ±5% of the exact corresponding value.

[0114] In this embodiment, the detection location of the road marking detection device 100 can include various road surfaces, including but not limited to roads with road markings, and may also include roads with road marking standard plates placed on them. In an exemplary embodiment, the road marking standard plate can have any suitable length, for example, 50 cm. Figure 1 In the illustrated embodiment, when light is projected onto the marking standard board, the light projection position can be the width of the marking standard board. W s The midpoint, but this application is not limited to this.

[0115] like Figure 1 As shown, the marking detection device 100 may further include a distance sensing unit 130, which is configured to acquire the actual projection distance of the laser beam projected by the laser source to the detected position. The real-time acquisition of the light projection distance by the distance sensing unit 130 is particularly beneficial for implementing the distance correction processing of the embodiments of this application. In some embodiments, the distance sensing unit 130 may be configured synchronously with the light projection of the laser source, for example, it may acquire the light projection time of the laser source. t 0 And the moment when the distance sensing unit 130 receives the reflected light of the projected light. t 1 The light projection distance can be determined accordingly. In one example, the distance sensing unit 130 is, for example, a TOF sensor, but this application is not limited to this.

[0116] like Figure 1 As shown, the road marking detection device 100 may further include an ambient light sensing unit 140, which is configured to acquire ambient light parameters when acquiring the first and / or second images. Accordingly, the road marking detection data will also include the ambient light parameters.

[0117] In a comparative example, the following tests were conducted on the light source involved and its visibility analysis. More specifically, the effectiveness of the marking detection device under illumination by conventional light sources such as laser light sources, xenon lamp light sources, halogen lamp light sources, and LED light sources according to the embodiments of this application can be compared and analyzed through tests.

[0118] Specifically, images of the marking lines illuminated by different types of light sources were taken using a camera in both a dark room and outdoors under natural daylight conditions, as well as images of the marking lines with the lighting source turned off. Five images were taken consecutively each time. In this test, besides the different light sources used, the marking line detection device employed... Figure 1 The example of a road marking detection model shares similar components, configurations, and parameters (however, this does not imply that a road marking detection device using a conventional light source is prior art). The road marking images are then processed to extract the grayscale values ​​of the road marking regions in each image, and their characteristics are analyzed.

[0119] The research results show that in a dark room, the camera captured reticle images and their gray values can be clearly identified under the light environment of xenon lamp, halogen lamp, and LED lamp, and there is a difference in the gray values of the reticle pictures captured when the light source is turned off. However, under the outdoor natural light environment, there is no significant difference in the gray values of the reticle images captured when the xenon lamp, halogen lamp, and LED lamp light sources are turned on and when the conventional light source is turned off. After subtraction (removing the interference of natural light), it is almost impossible to obtain the gray values of the reticle under the illumination of artificial light source. In contrast, the reticle detection device using the laser light source of the embodiments of the present application, more specifically, the camera with a filter (only through 808 nm) under the illumination of infrared laser (wavelength of 808 nm), can well identify the reticle and extract the gray values of the reticle under the illumination of artificial light source in both cases, as shown in Figure 2 . The test results of green laser and infrared laser are consistent, as shown in Figure 3 .

[0120] In some embodiments of the present application, the reticle detection device includes a plurality of cameras and a plurality of laser light sources, each of the cameras being matched with one of the laser light sources. The plurality of image acquisition units are arranged in a transverse direction to the projection direction of the laser light sources, and each of the cameras and the laser light source matched therewith have substantially the same transverse position in the transverse direction to the projection direction of the laser light source. In some embodiments, the reticle detection device can be a vehicle-mounted reticle detection device. In the vehicle-mounted reticle detection device, for example, cameras and matched laser light sources can be installed on the left front side and the right front side of the vehicle, and optionally in the middle position.

[0121] Next, the retroreflective luminance coefficient determination method according to the embodiments of the present application will be described.

[0122] Referring to Figure 4 , a retroreflective luminance coefficient determination method according to an embodiment of the present application is shown, which includes steps S410 to S460:

[0123] S410: Obtain a first image and a second image corresponding to a detected position in reticle detection.

[0124] The first image is an image acquired under the projection of single-wavelength laser light, and the second image is an image acquired under the non-projection of single-wavelength laser light.

[0125] In one embodiment, the retroreflective luminance coefficient determination method can use the reticle detection data detected by the reticle detection device 100 of the above-mentioned embodiment. However, it is conceivable that in other embodiments, other first and second images related to the reticle can be used, as long as the first and second images are images obtained under the light projection and non-projection, respectively.

[0126] In some embodiments of the present application, the retroreflective luminance coefficient determination method can be acquired in real time by the on-vehicle lane marking detection device when the vehicle is in free driving. In further embodiments, the retroreflective luminance coefficient determination method can be executed in real time by an electronic device located on the vehicle or by the electronic device of the vehicle itself. In still further embodiments, the retroreflective luminance coefficient determination method can be executed in real time by a cloud or remote electronic device. When the detection data is acquired by the lane marking detection device 100 of the above-mentioned embodiments, the cloud or remote electronic device can be communicatively connected to the lane marking detection device 100, for example, directly or through a communication unit of the vehicle.

[0127] In some embodiments, the matched first and second images of lane marking detection data can be acquired in real time at a certain interval frequency. For example, the camera can take 30 images per minute at equal time intervals, and the laser light source is alternately turned on and off when the camera takes pictures, and adjacent images can in turn constitute the first image and the corresponding second image. In a preferred embodiment, the first image can be an image acquired under the projection of an 808 nm infrared laser.

[0128] However, it is contemplated that the retroreflective luminance coefficient determination method of the embodiments of the present application is not limited thereto. For example, the retroreflective luminance coefficient determination method of the embodiments of the present application can be executed non-real-time using stored images.

[0129] S420: identifying a first lane marking region in the first image using a given image segmentation algorithm.

[0130] S430: identifying a second lane marking region in the second image using a given image segmentation algorithm.

[0131] In the embodiments of the present application, especially in steps S420 and S430, the image segmentation algorithm is an artificial intelligence image segmentation algorithm based on machine learning or deep learning, especially an end-to-end image segmentation algorithm capable of identifying lane marking regions in real time. By accurately identifying and calibrating the lane marking detection region, it facilitates rapid and accurate determination of the lane marking retroreflective luminance coefficient. By way of explanation and not limitation, compared to the scheme of determining the retroreflective luminance coefficient by filtering the gray value of the entire image, the accuracy of the gray value of the lane marking region is higher after accurately identifying the lane marking region, and the accuracy of the determined retroreflective luminance coefficient is also higher.

[0132] In preferred embodiments, in the embodiments of the present application, especially in steps S420 and S430, the lane marking identification can be processed using a lane marking recognition model based on the YOLO image segmentation algorithm, and more preferably, a lane marking recognition model based on the improved YOLOv8n-seg image segmentation algorithm. As Figures 6 to 8As shown, the reticle recognition model based on the YOLO image segmentation algorithm includes a backbone network 620, a neck network 630, and a plurality of segment heads 640 for processing different size feature maps.

[0133] Correspondingly, in a preferred embodiment, the reticle region recognition in the embodiments of the present application, especially in steps S420 and S430, includes: inputting an image 610, here an image related to a reticle (such as a first image or a second image), into a reticle recognition model based on the YOLO image segmentation algorithm for reticle recognition processing, thereby identifying a reticle region in the image. More specifically, with reference to Figure 5 and Figures 6-8 , the input image 610 is processed in turn by the backbone network 620, the neck network 630, and the plurality of segment heads 641, 642, 643, and 644.

[0134] As shown in Figure 6 and Figure 7 , the backbone network 620 can be used to extract a plurality of feature maps from the input image 610 and output the plurality of feature maps through a plurality of feature channels to the neck network 630. As shown in Figure 6 and Figure 7 , the backbone network 620 includes a plurality of feature extraction convolution modules 621, 622, 623, and 624 corresponding to a plurality of feature channels, respectively. As shown in Figure 7 In particular, in addition to the plurality of feature extraction convolution modules 621, 622, 623, and 624 of the plurality of feature channels, the backbone network 620 can also include additional layers or modules, which are not described here.

[0135] As shown in Figure 6 and Figure 7 , the neck network 630 can include a bottom-up and top-down sampling structure, including a plurality of necessary neural network layers, such as various layers with sampling functions and required splicing layers.

[0136] As shown in Figure 6 and Figure 7 , after sampling processing, the neck network 630 can correspondingly output the sampled feature maps to the corresponding segment heads 641 to 644. As shown in Figure 7In the specific embodiment, each segmentation head uses a dual loss function to jointly detect and segment, i.e., a bounding box loss function (Bbox Loss) and a target classification loss function (ClsLoss).

[0137] In the embodiments of the present application, the (improved) YOLO image segmentation algorithm refers to a YOLO image segmentation algorithm for which the novel features of any of the embodiments of the present application are used to optimize the target line recognition, and in particular refers to a YOLOv8n-seg image segmentation algorithm for which the novel features of any of the embodiments of the present application are used to optimize the target line recognition. It can be envisaged that the features of the embodiments of the present application for target line identification processing can be combined into or used to improve other end-to-end image segmentation algorithm frameworks, and the present application or subsequent applications can cover such novel combinations and improvements.

[0138] In one specific example, the target line detection device can be configured according to the target line detection model described above to obtain multiple road images, and the training set, test set and validation set can be divided according to a certain ratio, such as an 8:1:1 ratio, to verify the recognition result.

[0139] Before the target line recognition processing or as a preprocessing step, it can also include annotating the images as the training set. In one example, the road target line images in the training set can be annotated using an image annotation tool (Labelme image label annotation software). In order to accurately obtain the boundary coordinate points of the road target line and the category of the road target line during deep learning, an irregular boundary box mask is used to mark the target in the Labelme software during annotation, and then the category of the target road target line is created and named, and different colors of masks are used to mark different categories of targets.

[0140] In further embodiments, as Figure 5 As shown, the processing using the target line recognition model based on the YOLO image segmentation algorithm can specifically include steps S510 to S540:

[0141] S510: using a backbone network to perform feature extraction on an input image, and outputting multiple feature maps through multiple feature channels.

[0142] The multiple feature channels correspond to the multiple segmentation heads.

[0143] S520: adding an attention mechanism to part of the multiple feature channels, thereby applying attention processing to part of the multiple feature maps and not applying attention processing to other parts of the multiple feature maps.

[0144] S530: using a neck network to perform sampling processing on the multiple feature maps that have been partially subjected to attention processing, and outputting the multiple feature maps that have been subjected to sampling processing to the multiple segmentation heads;

[0145] S540: respectively processing the plurality of feature maps subjected to sampling processing by using the plurality of segmentation heads to segment and extract the lane line region in the plurality of feature maps.

[0146] In combination with reference Figures 6 to 8 , the plurality of segmentation heads can include a large-size segmentation head 641, a plurality (such as two) of medium-size segmentation heads 642, 643, and a small-size segmentation head 644.

[0147] In this preferred embodiment, the lane line recognition model based on the improved YOLO image segmentation algorithm adds a small-size segmentation head, which can also be referred to as a small target segmentation head.

[0148] Therefore, in the above step S510, the plurality of feature maps are output through the plurality of feature channels, including: A1: outputting the large-size feature map through the first feature channel, A2: respectively outputting the plurality of medium-size feature maps through the plurality of second feature channels, and A3: outputting the small-size feature map through the third feature channel.

[0149] Further, as Figures 6 to 8 indicated, the plurality of channels also correspondingly include the first feature channel corresponding to the large-size segmentation head, the plurality of second feature channels respectively corresponding to the plurality of medium-size segmentation heads, and the third feature channel corresponding to the small-size segmentation head.

[0150] Therefore, in the above step S540, the plurality of feature maps subjected to sampling processing are correspondingly output to the plurality of segmentation heads, including: B1: correspondingly outputting the large-size feature map subjected to sampling processing to the large-size segmentation head, B2: correspondingly outputting the plurality of medium-size feature maps subjected to sampling processing to the plurality of medium-size segmentation heads, and B3: correspondingly outputting the small-size feature map subjected to sampling processing to the small-size segmentation head.

[0151] In the lane line recognition model based on the improved YOLO image segmentation algorithm in the embodiments of the present application, after processing by the backbone network and the neck network, not only the large-size feature map and the plurality of medium-size feature maps are output to the corresponding segmentation heads for instance segmentation, but also the small-size feature map is output to the corresponding small-size segmentation head (small target segmentation head) for segmenting small target lane lines, especially for the specific scene of lane line recognition. As an explanation but not as a limitation, the pixel area scale of the road lane line photographed by the camera at a small angle in the picture is quite different, and the general image segmentation algorithm not specially modified for lane line recognition has weak instance detection capability for various small-scale targets, so the detection of road lane lines far from the camera is prone to missed detection.

[0152] In a preferred embodiment of this application, the ratio of the large-size feature map corresponding to the large-size segmentation head to the small-size feature map corresponding to the small-size segmentation head is greater than or equal to 8. In one example, an input image of size 640*640 can be downsampled by a sampling factor of 20 times (corresponding to the first feature channel of the large-size segmentation head), 40 times, 80 times, and 160 times (corresponding to the third feature channel of the small-size segmentation head) to obtain feature maps of sizes 32*32, 16*16, 8*8, and 4*4, respectively, which are used for segmentation by multiple feature heads. By using a segmentation head for a small target (and the corresponding backbone network and neck network structure, as further described below) in the improved YOLO image segmentation algorithm's road marking recognition model, the instance detection capability for small target road markings is effectively improved.

[0153] Reference Figure 5 and Figures 6-8 In step S530, an attention mechanism, such as channel attention (CA), is added to the multiple feature channels, thereby applying attention processing to a portion of the multiple feature maps while not applying attention processing to other portions of the multiple feature maps. According to a preferred embodiment, the feature map sizes processed by the segmentation head corresponding to the feature channels to which the attention mechanism is applied are not adjacent. More specifically, referring to the reference... Figures 6 to 8 The above step S520 may accordingly include: B1: adding an attention mechanism to the first feature channel and the third channel to apply attention processing to the large-size feature map and the small-size feature map, and not applying an attention mechanism to the second channel to apply no attention processing to the intermediate-size feature map. In other words, channel attention mechanism 660 is applied only to the first and third feature channels corresponding to the large-size segment head and the small-size segment head, while no channel attention mechanism is applied to the second feature channel corresponding to the intermediate-size segment head. According to a preferred embodiment, the attention mechanism includes an encoder for embedding the spatial coordinate information of the gradation region into the feature and a coordinate attention generation module for generating coordinate attention based on the embedded spatial coordinate information of the gradation region.

[0154] Channel attention mechanisms increase the importance of a channel to the information of interest by assigning weights to the feature maps of each channel to which attention is applied. The applicant's research yielded the following unexpected finding: adding partial channel attention mechanisms, such as adding them non-adjacently, and especially applying them only to the largest and smallest feature maps, can minimize the interference of various road disturbances with the target information of road markings. As an explanation, and not a limitation, this involves a bottom-up approach using a backbone network (e.g.,... Figure 6 and Figure 7) The deeper layer features are extracted layer by layer, the feature information in the uppermost feature map is the richest and the position information is the most fuzzy, while the feature information in the lowermost feature map is the opposite. By applying the channel attention mechanism to the feature channels at both ends, the clearest position information and feature information can be extracted and fused to obtain higher accuracy. In addition, the partially applied channel attention mechanism according to the embodiments of the present application, such as the non-adjacent channel attention mechanism, especially the channel attention mechanism applied only to the largest and smallest feature maps, can also be combined with further preferred features to obtain further beneficial effects of optimizing the lane recognition.

[0155] With continued reference to Figures 6 to 8 As described previously, the backbone network 620 includes a plurality of feature extraction convolutional modules 621, 622, 623, 624 corresponding to a plurality of feature channels respectively. In the preferred embodiments of the present application, at least part of, preferably all of the feature extraction convolutional modules corresponding to the feature channels without attention mechanism can include deformable convolution extraction layers, and the feature extraction convolutional modules corresponding to the feature channels with attention mechanism include fixed convolution extraction layers. As Figure 7 Specifically, in the feature extraction convolutional modules 621, 624 corresponding to the first and third feature channels, C2F (Coarse to Fine) fixed convolution extraction layers can be included; while in the feature extraction convolutional modules 622, 623 corresponding to the second feature channel without attention mechanism, deformable convolution extraction layers 6221 and 6231 can be included. In a preferred embodiment, the deformable convolution extraction layer can include a DCNv3 deformable convolution layer. Referring to Figure 8 , a DCNv3 deformable convolution functional layer is shown in the deformable convolution extraction layer 6221. In the embodiments of the present application, it will be understood that the convolution extraction layer can be broadly interpreted to include functional convolution (sub) layers with extraction functions, and can also include additional (sub) layers such as batch layers, bottleneck layers, splicing layers, etc. according to needs, and the embodiments of the present application do not limit the different numbers and configurations of additional (sub) layers in the convolution extraction layer.

[0156] When collecting road markings, the road marking collection vehicle may undergo turning, lane changing and other operations, causing the road markings to have serious shape distortion, and the angle of the road markings in the image may also change unknownly. The present inventors have made the following unexpected findings through research: by making at least part of the feature extraction convolutional modules corresponding to the feature channels without attention mechanism contain deformable convolution extraction layers, the features of these shape-distorted road markings can be effectively and accurately extracted.

[0157] As described previously, as Figure 7As shown, each segmentation head uses a dual loss function to jointly detect and segment, i.e., a bounding box loss function (Bbox Loss) and a target classification loss function (Cls Loss). In the embodiments of the present application, the bounding box loss function (Bbox Loss) is further improved to optimize the lane recognition. In the preferred embodiments, before the lane recognition processing is performed, the related method can include a training step, specifically comprising: C1: inputting a training image containing a labeled lane into a lane recognition model framework to be trained to obtain a trained lane recognition model, wherein the training image has a true value related to the labeled lane, and the true value includes a real bounding box of the labeled lane; C2: the training includes iteratively performing the following steps until a preset training completion condition is reached: C11: inputting a training sample into the lane recognition model framework to obtain a prediction value related to the labeled lane, and the prediction value includes a predicted bounding box of the labeled lane; C22: based on a given loss function, calculating a loss value between the prediction value and the true value, wherein the given loss function includes a key point intersection over union (MPDIOU) loss function, and the calculation of the loss value between the prediction value and the true value includes: C221: determining an area intersection over union of the predicted bounding box and the real bounding box, C222: determining distances between key points of the predicted bounding box and corresponding key points of the real bounding box according to a plurality of given predicted bounding box key points, and C223: determining a key point intersection over union loss value according to the area intersection over union and the distances; C23: reversely updating parameters of the lane recognition model framework based on the loss value.

[0158] In the preferred embodiments, the plurality of predicted bounding box key points include diagonal points. The selection of the bounding box diagonal points (bounding box diagonal lines) is determined according to the pointing direction and / or shape distortion of the labeled lane.

[0159] In some known schemes, the bounding box loss function Box_Loss=DFL_Loss+CIOU_Loss is used to measure the coincidence degree between the predicted bounding box and the real bounding box. Wherein DFL_Loss is a deep feature loss function, and CIOU_Loss is a complete intersection over union loss function, which uses the distance ratio of the “real bounding box” and the “predicted bounding box” to measure the coincidence degree between the predicted bounding box and the real bounding box. However, the applicant found through research that this loss function cannot well improve the loss value when performing lane recognition (training).

[0160] To this end, the embodiments of the present application propose to use a key point intersection over union (MPDIOU) loss function, and preferably, use the bounding box diagonal points as the key points of the key point intersection over union (MPDIOU) loss function.

[0161] According to a more optional embodiment, the selection of the prediction box diagonal point (box diagonal line) located on the box diagonal line is determined according to the pointing direction and / or shape distortion of the labeled marking line. In one example, for a marking line turning right. The upper right and lower left prediction box diagonal points can be used as key points. In one example, for a marking line turning left, the upper left and lower right prediction box diagonal points can be used as key points. In one example, for a straight marking line in an image taken when turning right (and thus causing shape distortion), the upper right and lower left box diagonal points can be used as key points. In one example, for a straight marking line in an image taken when turning left (and thus causing shape distortion), the upper left and lower right box diagonal points can be used as key points.

[0162] As an explanation but not as a limitation, the loss value is difficult to improve for the prediction box and the real box in the case of the same aspect ratio but different width and height values; the embodiment of the present application uses the key point intersection over union (MPDIOU) loss function, and uses the box diagonal point as the key point, which can accurately segment the road marking line under the interference of vehicle, pedestrian shielding, etc.; further, the embodiment of the present application also determines the selection of the box diagonal point (box diagonal line) direction according to the pointing direction and / or shape distortion of the labeled marking line, which further improves the marking line accurate detection capability.

[0163] In one embodiment, the accuracy, recall rate and precision of marking line detection and segmentation in the image verification set are analyzed. The test results show that the model according to the embodiment of the present application not only can meet the real-time marking line recognition of end-to-end, but also the accuracy on the detection category reaches 98.4%, the recall rate reaches 94.7%, and the precision reaches 96.7%; the accuracy on the segmentation reaches 98.3%, the recall rate reaches 94.2%, and the precision reaches 96.4%. The model realizes accurate detection and segmentation of road marking lines.

[0164] However, it should be understood that other image segmentation algorithms can also be used to obtain new embodiments.

[0165] S440: Extracting a first gray value of a first marking line region in the first image.

[0166] S450: Extracting a second gray value of a second marking line region in the second image.

[0167] In this step S440 and S450, the gray value of the marking line region can be extracted in various ways, such as but not limited to averaging, median, etc., which is not limited by the present application. As shown in the foregoing, the embodiment of the present application extracts the gray value of the accurately segmented marking line region, rather than extracting the gray value from the entire image by threshold screening, which improves the accuracy of the calculated retroreflective luminance coefficient.

[0168] S460: calculating the retroreflective luminance factor of the detected position's scale according to the first gray value, the second gray value and a given retroreflective luminance factor standard value.

[0169] In some embodiments, as shown in FIG. 4B, the step S460 can comprise steps S461-S464: Figure 9

[0170] S461: determining the scale gray value difference between the first gray value and the second gray value;

[0171] S462: obtaining the retroreflective luminance factor standard value corresponding to the given scale standard board;

[0172] S463: obtaining the standard gray value difference corresponding to the given scale standard board;

[0173] In some embodiments, as shown in FIG. 4B, the step S463 can comprise steps S4631-S4636: Figure 10

[0174] S4631: obtaining the third image and the fourth image corresponding to the given scale standard board, wherein the third image is an image collected under the condition of single-wavelength laser projection, and the fourth image is an image collected under the condition of no single-wavelength laser projection;

[0175] S4632: identifying the third scale region in the third image by using a given image segmentation algorithm;

[0176] S4633: identifying the fourth scale region in the fourth image by using a given image segmentation algorithm;

[0177] S4634: extracting the third gray value of the third scale region in the third image;

[0178] S4635: extracting the fourth gray value of the fourth scale region in the fourth image;

[0179] S4636: determining the standard gray value difference between the third gray value and the fourth gray value.

[0180] In Figure 9 and Figure 10 In the embodiments shown in FIGS. 4B and 4C, similar or different methods as described in steps S420 and S430 can be used to identify the third and fourth scale regions, and similar or different methods as described in steps S440 and S450 can be used to extract the third and fourth gray values.

[0181] In addition, it will be appreciated that in the step S463, different methods for obtaining the standard gray value difference corresponding to the given scale standard board can also be used. For example, the known or stored standard gray value difference can be obtained by looking up a table. ​​

[0182] S464: determining the retroreflective luminance coefficient of the marker line of the detected position according to the proportional relationship of the marker line gray difference, the standard gray difference and the retroreflective luminance coefficient standard value.

[0183] In the embodiments of the present application, when or after the retroreflective luminance coefficient of the marker line of the detected position is calculated, the method further comprises a correction step S470 (not shown): correcting the retroreflective luminance coefficient of the marker line to obtain a corrected retroreflective luminance coefficient.

[0184] More specifically, as shown in FIG. 4, the correction can include distance correction, and in particular, the correction step S470 can comprise: Figure 11

[0185] S471: obtaining the calibration projection distance and the actual projection distance of the single-wavelength laser projection.

[0186] S472: determining a distance correction coefficient according to the calibration projection distance and the actual projection distance.

[0187] S473: correcting the retroreflective luminance coefficient of the marker line according to the distance correction coefficient to obtain a distance-corrected retroreflective luminance coefficient.

[0188] In some embodiments, the determination of the distance correction coefficient according to the calibration projection distance and the actual projection distance comprises:

[0189] The distance correction coefficient is determined according to the following formula (3) K D :

[0190] K D = D 2 / D N 2 (3)

[0191] wherein, D the actual projection distance is d, D N the calibration projection distance is D.

[0192] In some embodiments, the correction can include ambient light correction, and the correction step S470 can comprise: correcting the retroreflective luminance coefficient of the marker line to obtain an ambient light-corrected retroreflective luminance coefficient, and in particular comprising: D1: obtaining an ambient light parameter at the time of obtaining the marker line detection data; D2: determining an ambient light correction coefficient according to the ambient light parameter, the marker line gray difference and the retroreflective luminance coefficient standard value of the marker line of the detected position; D3: correcting the retroreflective luminance coefficient according to the ambient light correction coefficient to obtain an ambient light-corrected retroreflective luminance coefficient. ​

[0193] As described above, the correction process is optional, and in some embodiments, the correction process, such as the correction processes described in steps S471, S472, S473 and D1, D2, D3 above, can be combined with the calculation of the retroreflective luminance coefficient, or can be performed separately after the calculation of the (original) retroreflective luminance coefficient. In one illustration, the correction process combined with the calculation of the retroreflective luminance coefficient will be described:

[0194] In this example, the detection device of this application embodiment can be used to detect the test marking standard board (for ease of distinction, the test marking standard board is temporarily regarded as a marking) and determine the gray values ​​of the marking area in the first image and the second image as B1 and B2, respectively. Then the gray difference of the marking is... B = B 1 -B 2 .

[0195] Retroreflection luminance coefficient without distance / environment correction: Assuming the standard value of the retroreflection luminance coefficient is... RL s =150mcd.m-2.lx-1, its grayscale difference is standard B 150 The retroreflection luminance coefficient without distance / environment correction is RL raw = B * RL s / B 150 Since the inspection is performed using a standard caliper in this example, the process can conveniently be called caliper calibration. However, it will be understood that when the actual caliper is processed, the process corresponds to the retroreflective luminance coefficient of the caliper at the inspection location (without distance and environmental correction) as determined by steps S461 to S464 above.

[0196] Distance correction: as previously stated K D = D 2 / D N 2 .

[0197] Ambient light correction: The applicant discovered that, especially under natural light conditions, some ambient light remains in the grayscale difference B. To completely eliminate the influence of ambient light, a correction factor is introduced. K E Specifically, the correlation between ambient illuminance E, grayscale difference B, and the true value of retroreflection luminance coefficient RL* can be obtained and analyzed. This correlation can be derived from a model / algorithm. f(E) Characterize and determine the correction coefficients. K E= f(E) .

[0198] Thus, in this example, the laser irradiation-based determination method of the retroreflective luminance coefficient of the target line can determine the corrected retroreflective luminance coefficient as RL = K D *K E * RL raw = D 2 / D N 2 * f(E) * (B 1 -B 2 ) * RL s / B 150 .

[0199] In this example, the correction coefficient K E = f(E) may be dynamically determined based on the acquired ambient light intensity E, and the introduction of the correction coefficient K E further realizes dynamic detection and correction of the retroreflective luminance coefficient, and advantageously further eliminates the influence of ambient light.

[0200] In combination with reference to Figure 1 and Figure 5 , before acquiring the first image and the second image, the method further comprises: S400 (not shown): constructing a target line detection model based on a proportional relationship, as shown in Figure 12 S400 can specifically include:

[0201] S401: determining a calibration height parameter for the target line visibility observation model, the calibration height parameter including a calibration driver observation height and a calibration vehicle light height.

[0202] S402: determining a calibration observation distance for the target line visibility observation model.

[0203] S403: setting a calibration projection distance for light projection for target line detection, and determining a target line detection height parameter for target line detection from the calibration height parameter based on a proportional relationship between the calibration observation distance and the calibration projection distance.

[0204] wherein the target line detection height parameter includes a camera height for image acquisition and a light source height for light projection.

[0205] As an alternative to step S403, S403' (not shown): setting at least partly a calibration height parameter for the calibration observation, and determining a calibration projection distance for the light projection of the calibration based on a proportional relationship between the calibration height parameter and the at least partly set calibration height parameter.

[0206] wherein the first image and the second image are acquired at the calibration projection distance, the camera height and the light projection height.

[0207] The above describes a retroreflective luminance coefficient determination method according to an aspect of the present application, which can process two images with or without single-wavelength laser projection, and precisely segment the target line region by using image segmentation algorithms for the two images respectively; and as shown in Figure 12 the preferred embodiment of the retroreflective luminance coefficient determination method of this aspect can further construct a target line detection model based on a proportional relationship. However, it will be appreciated that in another aspect of the present application, a retroreflective luminance coefficient determination method for constructing a target line detection model can be provided, in which the retroreflective luminance coefficient of the target line can be determined based on or not based on the laser projection image and / or the image segmentation feature, for example, the retroreflective luminance coefficient determination method of this aspect can determine the retroreflective luminance coefficient based on a general light projection image. Accordingly, as shown in Figure 13 the present application further provides a retroreflective luminance coefficient determination method, comprising:

[0208] S1310: determining a calibration height parameter for the target line visibility observation model, the calibration height parameter including a calibration driver observation height and a calibration vehicle light height.

[0209] S1320: determining a calibration observation distance for the target line visibility observation model.

[0210] S1330: setting a calibration projection distance for the light projection of the target line detection, and determining a target line detection height parameter for the target line detection based on a proportional relationship between the calibration observation distance and the calibration projection distance.

[0211] the target line detection height parameter includes a camera height and a light projection height for acquiring images

[0212] S1330' (not shown): setting at least partly a target line detection height parameter, and determining a calibration projection distance for the light projection of the target line detection based on a proportional relationship between the calibration height parameter and the at least partly set target line detection height parameter;

[0213] S1340: acquiring target line detection data corresponding to a detected position in the target line detection.

[0214] The image in the reticle detection data is acquired under the calibration projection distance, the camera height, and the light projection height.

[0215] S1350: Determine the retroreflective luminance coefficient according to the reticle detection data.

[0216] In summary, the embodiments of the present application provide the following several innovative invention subjects, any of which can be combined with or without other innovative subjects or features:

[0217] (1) The embodiments of the present application provide a reticle detection device, and the reticle detection data of the reticle detection device includes two modes of pictures with single-wavelength laser projection and without projection, so that the data detected by the reticle detection device can be particularly beneficial to accurately calculating the retroreflective luminance coefficient of the reticle.

[0218] Those skilled in the art will understand that the reticle detection data obtained by the reticle detection data can not be limited to determining the retroreflective luminance coefficient of the reticle, or can be additionally used for other purposes of reticle detection / identification, such as automatic driving, etc. Those skilled in the art will understand that the reticle detection device and the retroreflective luminance coefficient determination scheme can be provided or implemented by the same or different entities, for example, the reticle detection device is provided in the vehicle, and the retroreflective luminance coefficient determination scheme can be implemented by other electronic devices belonging to or not belonging to the vehicle.

[0219] (2) The embodiments of the present application provide a retroreflective luminance coefficient determination scheme in one aspect, which not only processes two images with single-wavelength laser projection or without projection, but also uses image segmentation algorithm to accurately segment the reticle area for the two images respectively. By extracting the gray value from the reticle area segmented from the reticle image, and combining it with the standard value of the retroreflective luminance coefficient of the standard plate, the method can accurately determine the retroreflective luminance coefficient of the object to be measured.

[0220] (3) The embodiments of the present application provide another aspect of the retroreflective luminance coefficient determination scheme, which can construct a reticle detection model based on light projection in proportion to the reticle visibility observation model, for determining the retroreflective luminance coefficient. This method effectively guarantees the accuracy of the detection result while reducing the measurement complexity and limiting the measurement equipment, thereby ensuring the accuracy of the determined retroreflective luminance coefficient.

[0221] Here, the retroreflective luminance coefficient determination scheme of the first aspect or its features can be combined with the retroreflective luminance coefficient determination scheme of the second aspect.

[0222] (4) The application embodiment provides a marking line identification method, especially based on artificial intelligence, which can efficiently (real-time, end-to-end) and accurately (especially avoiding occlusion and shape distortion) identify marking lines. The marking line identification method provided by the application embodiment can be advantageously combined with the retroreflective luminance coefficient determination scheme of the first aspect to further ensure the accurate and rapid determination of the retroreflective luminance coefficient. However, the marking line identification method can not be limited to the scheme for determining the retroreflective luminance coefficient of the marking line, and can be additionally used for other purposes of marking line detection / identification, such as automatic driving, etc.

[0223] (5) The application embodiment provides a marking line retroreflective luminance coefficient correction method based on distance and / or ambient light correction, which solves the problems of strong light interference and dynamic detection angle change, and further ensures the high accuracy of the retroreflective luminance coefficient. Here, the correction method can cover various modifications of only distance correction, only ambient light correction, and both correction.

[0224] In addition, the correction method provided by the application embodiment can be advantageously combined with the retroreflective luminance coefficient determination scheme and / or the marking line detection device of the application embodiment to further ensure the accurate determination of the retroreflective luminance coefficient. However, the correction method provided by the application embodiment can be used to correct the retroreflective luminance coefficient obtained by other methods. For example, a known ordinary light projection retroreflective luminance coefficient measurement device can be modified to add a distance sensing unit and / or an ambient light sensing unit to determine the actual distance of light projection and / or the ambient light when the image is collected, so that the retroreflective luminance coefficient value determined by the known measurement device can be more accurately corrected in the subsequent retroreflective luminance coefficient determination process.

[0225] As an example, reference is made to Figure 14 A retroreflective luminance coefficient evaluation device 1400 of the application embodiment is also shown, which can include an acquisition unit 1401, a first marking line identification unit 1402, a second marking line identification unit 1403, a first image processing unit 1404, a second image processing unit 1405, and a retroreflective luminance coefficient calculation unit 1407, wherein:

[0226] The acquisition unit 1401 is configured to acquire a first image and a second image corresponding to a detected position in marking line detection, wherein the first image is an image collected under single-wavelength laser projection, and the second image is an image collected under single-wavelength laser non-projection.

[0227] The first marking line identification unit 1402 is configured to identify a first marking line region in the first image by using a given image segmentation algorithm.

[0228] The second marking line identification unit 1403 is configured to identify a second marking line region in the second image by using a given image segmentation algorithm.

[0229] The first image processing unit 1404 is configured to extract a first gray scale value of a first reticle region in the first image.

[0230] The second image processing unit 1405 is configured to extract a second gray scale value of a second reticle region in the second image.

[0231] The retroreflective luminance coefficient calculation unit 1407 is configured to calculate a retroreflective luminance coefficient of a reticle at the detected position according to the first gray scale value, the second gray scale value, and a given retroreflective luminance coefficient standard value.

[0232] In some embodiments, the acquisition unit 1401 is further configured to acquire an ambient light parameter value corresponding to the detected position in reticle detection; and configured to acquire a single-wavelength laser actual projection distance corresponding to the detected position in reticle detection.

[0233] In this embodiment, with reference to Figure 14 , the retroreflective luminance coefficient evaluation device 1400 can further include a third processing unit 1406, wherein the third processing unit 1406 is configured to extract an ambient light correction coefficient and a distance correction coefficient of the reticle at the detected position.

[0234] In this embodiment, the retroreflective luminance coefficient calculation unit 1407 is further configured to calculate a retroreflective luminance coefficient of the reticle at the detected position according to the first gray scale value, the second gray scale value, the ambient light correction coefficient, the distance correction coefficient, and a given retroreflective luminance coefficient standard value.

[0235] In some embodiments, with reference to Figure 14 , the retroreflective luminance coefficient evaluation device 1400 can further include a retroreflective luminance coefficient evaluation unit 1408, wherein the retroreflective luminance coefficient evaluation unit 1408 is configured to receive the retroreflective luminance coefficient calculated by the calculation unit, and judge the retroreflective performance of the reticle according to a preset performance index threshold to evaluate whether it meets the prescribed safety standard.

[0236] Under the teachings of the present application, the features of the method embodiments can be combined with the features of the device embodiments in a non-contradictory manner to obtain new embodiments, and the features of the device embodiments can also be combined with the features of the method embodiments in a non-contradictory manner to obtain new embodiments, which fall within the scope of the present application.

[0237] In the embodiments of the present application, the retroreflective luminance coefficient determination method and device can be implemented by a computer integrating related functional modules or components. Figure 15 A structure diagram of a computer that can be used to implement the retroreflective luminance coefficient determination method and device of the embodiments of the present application is shown.

[0238] As Figure 15 indicated, the computer 1500 includes a processor 1501, which can execute the various suitable operations and processes according to the program and / or data stored in the read-only memory (ROM) 1502 or loaded from the storage section 1508 into the random access memory (RAM) 1503. The processor 1501 can include a central processing unit (CPU) and can be one of a single-core processor, a multi-core processor, or a plurality of processors. In some embodiments, the processor 1501 can include a general-purpose main processor and one or more special-purpose coprocessors, such as a graphics processing unit (GPU), a neural processing unit (NPU), a digital signal processor (DSP), etc. In the RAM 1503, various programs and data required for the operation of the electronic device 1500 are also stored. The processor 1501, the ROM 1502, and the RAM 1503 are connected to each other through a bus 1504. An input / output (I / O) interface 1505 is also connected to the bus 1504.

[0239] The above processor and memory are used together to execute the program stored in the memory, which, when executed by the computer, can implement the steps or functions of the methods described in the above embodiments.

[0240] The following components are connected to the I / O interface 1505: an input section 1506 including a keyboard, a mouse, a touch screen, etc.; an output section 1507 including a cathode ray tube (CRT), a liquid crystal display (LCD), etc., and a speaker, etc.; a storage section 1508 including a hard disk, etc.; and a communication section 1509 including a network interface card such as a LAN card, a modem, etc. The communication section 1509 performs communication processing via a network such as the Internet. A drive 1510 is also connected to the I / O interface 1505 as needed. A removable medium 1511 such as a magnetic disk, an optical disk, a magneto-optical disk, a semiconductor memory, etc. is mounted on the drive 1510 as needed, so that a computer program read therefrom is installed in the storage section 1508 as needed. Figure 15 Only some components are shown schematically in the computer system 1500, and it is not meant to imply that the computer system 1500 only includes Figure 15 the components shown.

[0241] In some embodiments, the computer 1500 refers to a mobile terminal, including a mobile phone, a vehicle-mounted terminal, etc. Taking the mobile phone as an example, the electronic device 1500 further includes a display screen with touch function, a loudspeaker, a gyroscope, a camera, a 4G / 5G antenna, etc. device modules.

[0242] The systems, apparatuses, modules or units illustrated in the above embodiments can be implemented by the computer or its associated components. The computer may, for example, be a mobile terminal, a personal computer, a laptop computer, an in-vehicle human-machine interaction device, a personal digital assistant, a media player, a navigation device, a tablet computer, or a combination thereof.

[0243] Although not shown, in the embodiments of the present application, a program product is provided, which includes a computer program, wherein the computer program is executed by a processor to implement the method according to any of the embodiments of the present application.

[0244] Although not shown, in the embodiments of the present application, a storage medium is provided, which stores a computer program configured to be executed to perform the method according to any of the embodiments of the present application.

[0245] The storage medium of the embodiments of the present application includes permanent and non-permanent, removable and non-removable information storage articles that can be implemented by any method or technology. Examples of the storage medium include, but are not limited to, phase change memory (PRAM), static random access memory (SRAM), dynamic random access memory (DRAM), other types of random access memory (RAM), read-only memory (ROM), electrically erasable programmable read-only memory (EEPROM), flash memory or other memory technologies, or any other non-transmission medium that can be used to store information that can be accessed by a computing device.

[0246] The methods, programs, systems, apparatuses, etc. of the embodiments of the present application can be executed or implemented in a single or multiple networked computers, or can be practiced in a distributed computing environment. In the embodiments of the present specification, in these distributed computing environments, tasks can be performed by remote processing devices connected through a communication network.

[0247] Unless explicitly stated, the actions or steps of the methods, programs according to the embodiments of the present application do not have to be performed in a specific order and still achieve the desired results. In some embodiments, multi-task processing and parallel processing are possible or can be advantageous.

[0248] In this document, various embodiments of the application are described, but the description is not exhaustive, and not all embodiments of the application have been described, and some embodiments can have been omitted. In this document, "one embodiment," "an embodiment," "some embodiments," "one specific embodiment," or "some embodiments" means that a particular feature, structure, or characteristic described in connection with the embodiment is included in at least one embodiment of the application. The appearances of the phrase "in one embodiment" in various places in the specification are not necessarily all referring to the same embodiment, and some embodiments can be understood as alternatively implying incorporation of the innovative or generative aspects of the application in one or more embodiments.

[0249] Exemplary systems and methods of the application have been specifically illustrated and described herein, but various changes in the systems and methods described herein can be made without departing from the spirit and scope of the application as defined in the appended claims.

Claims

1. A method of determining the coefficient of retroreflective luminance, characterized in that, The method comprises: determining a calibration height parameter for a target line visibility observation model, wherein the target line visibility observation model is used to construct a light projection based target line detection model, the light projection based target line detection model comprises a calibration projection distance and a target line detection height parameter, and the calibration height parameter comprises a calibration driver observation height and a calibration vehicle light height; determining a calibration observation distance for the target line visibility observation model; setting a calibration projection distance for light projection used in target line detection, and determining a target line detection height parameter for target line detection from the calibration height parameter based on a proportional relationship between the calibration observation distance and the calibration projection distance, or at least partially setting the target line detection height parameter, and determining the calibration projection distance for light projection used in target line detection from the calibration observation distance based on a proportional relationship between the calibration height parameter and the at least partially set target line detection height parameter, wherein the target line detection height parameter comprises a camera height for image acquisition and a light projection height; acquiring target line detection data corresponding to a detected position in target line detection, wherein images in the target line detection data are acquired at the calibration projection distance, the camera height and the light projection height; determining a retroreflective luminance coefficient according to the target line detection data.

2. The retroreflective luminance factor determination method of claim 1, wherein, The target line detection data comprises a first image with single-wavelength laser projection and a second image without single-wavelength laser projection.

3. The retroreflective luminance factor determination method according to claim 1 or 2, characterized in that The acquiring of the target line detection data corresponding to the detected position in target line detection comprises: acquiring a first image and a second image corresponding to the detected position in target line detection, wherein the first image is an image acquired under single-wavelength laser projection, and the second image is an image acquired under no single-wavelength laser projection; The determination of the retroreflective luminance coefficient according to the target line detection data comprises: identifying a first target line region in the first image by using a given image segmentation algorithm; identifying a second target line region in the second image by using the given image segmentation algorithm; extracting a first gray value of the first target line region in the first image; extracting a second gray value of the second target line region in the second image; calculating the retroreflective luminance coefficient of the target line at the detected position according to a target line gray value difference determined from the first gray value and the second gray value and a given retroreflective luminance coefficient standard value.

4. The retroreflective luminance factor determination method of claim 3, wherein, The calculation of the retroreflective luminance coefficient of the target line at the detected position according to the first gray value, the second gray value and the given retroreflective luminance coefficient standard value comprises: determining a target line gray value difference of the first gray value and the second gray value; acquiring the retroreflective luminance coefficient standard value corresponding to a given target line standard board; acquiring a standard gray value difference corresponding to the given target line standard board; determining the retroreflective luminance coefficient of the target line at the detected position according to a proportional relationship among the target line gray value difference, the standard gray value difference and the retroreflective luminance coefficient standard value.

5. The retroreflective luminance factor determination method of claim 3, wherein, The method further comprises: correcting the retroreflective luminance coefficient of the target line to obtain an ambient light corrected retroreflective luminance coefficient, specifically comprising: acquiring an ambient light parameter in the reticle detection; determining an ambient light correction coefficient according to the ambient light parameter, the reticle gray scale difference, and a standard value of the retroreflective luminance coefficient of the reticle at the detected position; correcting the retroreflective luminance coefficient according to the ambient light correction coefficient to obtain an ambient light corrected retroreflective luminance coefficient.

6. The retroreflective luminance factor determination method of claim 3, wherein, the acquiring of the standard gray scale difference corresponding to the given reticle standard board comprises: acquiring a third image and a fourth image corresponding to the given reticle standard board, wherein the third image is an image acquired under single-wavelength laser projection, and the fourth image is an image acquired under single-wavelength laser non-projection; identifying a third reticle region in the third image by using a given image segmentation algorithm; identifying a fourth reticle region in the fourth image by using the given image segmentation algorithm; extracting a third gray scale value of the third reticle region in the third image; extracting a fourth gray scale value of the fourth reticle region in the fourth image; determining the standard gray scale difference of the third gray scale value and the fourth gray scale value.

7. The retroreflective luminance factor determination method of claim 3, wherein, When calculating the retroreflective luminance coefficient of the reticle at the detected position or after that, the method further comprises: correcting the retroreflective luminance coefficient of the reticle to obtain a distance corrected retroreflective luminance coefficient, specifically comprising: acquiring a calibration projection distance and an actual projection distance of the single-wavelength laser projection; determining a distance correction coefficient according to the calibration projection distance and the actual projection distance; correcting the retroreflective luminance coefficient of the reticle according to the distance correction coefficient to obtain a distance corrected retroreflective luminance coefficient.

8. The retroreflective luminance factor determination method of claim 3, wherein, When determining the retroreflective luminance coefficient of the reticle or after that, the method further comprises: correcting the retroreflective luminance coefficient of the reticle to obtain an ambient light corrected and distance corrected retroreflective luminance coefficient, specifically comprising: acquiring an ambient light parameter in the reticle detection; determining an ambient light correction coefficient according to the ambient light parameter, the reticle gray scale difference, and a standard value of the retroreflective luminance coefficient of the reticle at the detected position; acquiring a calibration projection distance and an actual projection distance of the single-wavelength laser projection; determining a distance correction coefficient according to the calibration projection distance and the actual projection distance; correcting the retroreflective luminance coefficient of the reticle according to the ambient light correction coefficient and the distance correction coefficient to obtain an ambient light corrected and distance corrected retroreflective luminance coefficient.

9. An electronic device, comprising: comprises: a processor and a memory storing a computer program, the processor being configured to implement the method according to any one of claims 1-8 when running the computer program.

Images