A multi-band adaptive regional intelligent fill light imaging method and device

Abstract

The present invention provides a multi-band adaptive regional intelligent fill light imaging method and device, including a main bracket, an angle adjustment bracket and lamp beads of three bands (810nm, 910nm, 1060nm) and corresponding cameras. The device arranges twelve groups of lamp beads in three rows and four columns, each group contains lamp beads of three bands, and installs them at different inclination angles to achieve accurate fill light for different imaging areas. The system detects the brightness of each area of the image in real time, automatically identifies overexposed areas and their causes, and intelligently adjusts the fill light band and intensity. When strong incident light is detected, it automatically switches to a non-interference band; when it is detected that the reflected light is too strong, the fill light intensity of the corresponding area is reduced. This method effectively avoids interference from ambient light sources and uneven reflection of the fill light itself, greatly improves the imaging quality at night and in low-visibility environments, and significantly enhances the perception ability and safety of the autonomous driving system.

Description

Technical Field

[0001] The present invention belongs to the technical field of intelligent fill light imaging, and specifically relates to a multi-band adaptive regional intelligent fill light imaging method and a fill light device, which provide high-quality data for night driving systems and ensure night driving safety. Background Art

[0002] On highways, the risk of accidents during nighttime driving is far higher than during the day. Despite lower traffic volume at night, the incidence of major traffic accidents is twice that of daytime traffic, with over a third of these accidents occurring at night. The primary safety hazard associated with nighttime driving stems from insufficient road lighting. This is especially true on mixed traffic roads, where timely detection of pedestrians and obstacles is crucial. While improved headlights can alleviate this problem, the glare can impair the vision of oncoming vehicles or pedestrians, increasing the risk of accidents. Therefore, the use of nighttime driving assistance systems, particularly those based on infrared imaging technology, has become increasingly important.

[0003] There are two main types of infrared imaging technology: passive infrared imaging and active near-infrared imaging. Passive technology relies on natural radiation and does not require an additional light source, making it suitable for applications with demanding equipment requirements. However, this technology carries a higher cost. Active technology uses near-infrared light sources to enhance illumination, which is less expensive. However, its imaging is susceptible to interference from highly reflective materials on the road. Furthermore, when meeting other vehicles, the same-band illumination from oncoming vehicles can also interfere with the imaging. Therefore, addressing these lighting interference issues, particularly those caused by strong and reflected light, has become a core challenge for nighttime driving assistance systems.

[0004] Patent No. ZL200710018954.5, a prior art patent, provides a multi-wavelength automatic switching vehicle nighttime driving assistance system. This system aims to improve nighttime imaging clarity by combining variable-wavelength infrared illumination with a hyperfocal infrared imaging optical system. The system automatically switches between near-infrared light sources of different wavelengths to avoid interference from oncoming vehicles. However, prior art fails to fully account for interference from reflected light, resulting in image quality issues under complex lighting conditions.

[0005] Furthermore, patent number ZL201210257215.2 proposes a nighttime driving assistance system for vehicles. This system utilizes a variable-wavelength near-infrared lighting module and automatic wavelength switching to avoid light interference from oncoming vehicles. This system, combined with a hyperfocal imaging optical module and an ECU signal processing module, ensures clear imaging. While this system has achieved some success in addressing oncoming vehicle interference, it does not consider the impact of reflected light on image quality.

[0006] Patent No. ZL 201110009335.6 describes a mechanical infrared filter switching device that uses a stepper motor to drive the switching, reducing the device's size and making it suitable for miniature lenses. Although the device uses a position detector to ensure accurate filter switching, the mechanical adjustment speed is slow, unable to meet the real-time requirements of driving scenarios, and does not consider the impact of reflected light on imaging.

[0007] To address the low efficiency, high cost, and complex operation of traditional manual fill-lighting methods, Patent No. ZL202011636852.1 provides a fully automatic fill-lighting method and device for a warp knitting machine defect detection system. This method captures frame images and performs light intensity detection, automatically adjusting exposure time, gamma enable, automatic gain, and external light source status to achieve fully automatic fill-lighting and dynamically adjust detection intervals, reducing hardware cost and consumption while improving detection performance.

[0008] Patent No. ZL 201720830972.2 describes a glare suppression device that uses a light intensity sensor to detect ambient light intensity and automatically adjusts the light intensity to suppress glare interference. While this device effectively reduces the impact of glare, its mechanical adjustment speed is slow and cannot meet the real-time needs of nighttime driving. Furthermore, it cannot adjust light intensity by region, making it unable to address the problem of localized glare interfering with imaging.

[0009] To address the issue of partial overexposure or underexposure in images, Patent No. ZL201711378137.0 proposes a fill-light method and device. This method utilizes multiple fill-lights to divide the captured area into multiple sub-areas. The light intensity of the fill-light corresponding to each sub-area is adjusted separately, keeping the brightness difference between each sub-image and the overall image within a certain range. While this technology can address local overexposure or underexposure caused by the intensity of the light itself, it cannot address direct interference from other light sources in the same wavelength band.

[0010] Patent No. ZL202010400682.0 provides a fill light adjustment method, device, electronic device, and storage medium. By combining exposure parameter adjustment with fill light intensity adjustment, this method achieves refined fill light adjustment, significantly improving shooting quality, particularly at night or in low-light scenes. However, this technology can only adjust the overall brightness of the image and cannot address uneven lighting conditions or regional adjustments.

[0011] In summary, the first three patents (ZL200710018954.5, ZL201210257215.2, and ZL 201110009335.6) are limited to solving the interference problem of opposing light sources in the same band, and the last four patents (ZL202011636852.1, ZL201720830972.2, ZL201711378137.0, and ZL202010400682.0) can only solve the interference problem of uneven lighting in the environment. The existing technology fails to simultaneously solve the impact of uneven lighting and opposing light sources in the same band on imaging quality. During night driving, the uncertainty of the lighting environment makes vision-based perception difficult. Therefore, providing clear and stable imaging under complex lighting conditions has become a problem that needs to be solved urgently by current technology. Summary of the Invention

[0012] The purpose of the present invention is to provide a multi-band adaptive regional intelligent fill light imaging method and fill light device, aiming to solve the problem of uneven illumination and the influence of the coupling of light sources in the same band on the imaging quality, provide high-quality data for the night driving system, and ensure night driving safety.

[0013] The objectives of the present invention are achieved through the following technical solutions.

[0014] A multi-band adaptive regional intelligent fill-light imaging device includes: a condenser lens, a lamp panel, an angle adjustment bracket, a main bracket, a controller, a narrow-band 810nm camera, a narrow-band 910nm camera, a narrow-band 1060nm camera, narrow-band 810nm lamp beads, narrow-band 910nm lamp beads, narrow-band 1060nm lamp beads, a driver, and a detector. The narrow-band 810nm camera, narrow-band 910nm camera, and narrow-band 1060nm camera are horizontally mounted in the middle of the lower portion of the main bracket, with their optical axes parallel to each other and perpendicular to the mounting plane of the main bracket. The lamp beads are divided into three rows and four columns, totaling twelve groups, and are welded to twelve lamp panels, with the optical axis of each lamp bead perpendicular to the mounting plane of the lamp panel.

[0015] The twelve light panels are fixed on twelve angle adjustment brackets respectively, and the angle adjustment brackets in different positions are installed on the main bracket at different inclination angles, located above the camera;

[0016] The four tilted angle adjustment brackets in the bottom row tilt the light axes of the lamp beads on them downward, forming an angle of β degrees with the camera optical axis in the vertical plane; the four tilted angle adjustment brackets in the middle row tilt the light axes of the lamp beads on them and the camera optical axis in the same horizontal plane; the four tilted angle adjustment brackets in the top row tilt the light axes of the lamp beads on them upward, forming an angle of β degrees with the camera optical axis in the vertical plane; the three tilted angle adjustment brackets in the leftmost column tilt the light axes of the lamp beads on them to the left, forming an angle of α1 degrees with the camera optical axis in the horizontal plane; the three tilted angle adjustment brackets in the second column from the left tilt the light axes of the lamp beads on them to the left, forming an angle of α2 degrees with the camera optical axis in the horizontal plane; the three tilted angle adjustment brackets in the third column from the left tilt the light axes of the lamp beads on them to the right, The optical axis forms an angle of α2 degrees in the horizontal plane; the three tilted angle adjustment brackets in the rightmost column tilt the optical axis of the lamp beads on them to the right, forming an angle of α1 degree with the camera optical axis in the horizontal plane; the lamp beads on the same lamp panel include lamp beads of three bands: 810nm, 910nm and 1060nm, and the lamp beads of each band are connected in series into a group to form three groups; the three band groups of lamp beads on the same lamp panel are powered by the same driver in a time-sharing manner. At any time, only one band group of lamp beads is energized. The detector determines the band and intensity, and the controller controls the driver to achieve switching of lamp beads of different band groups and control of light intensity; each lamp bead is covered with a focusing lens; the focusing angle of the focusing lens is smaller than the camera imaging angle of view, which aggregates light and disperses it to cover the entire imaging area, realizing independent fill light and light intensity adjustment for different imaging areas.

[0017] Furthermore, the multi-band adaptive regional intelligent fill light imaging device uses three narrow-band near-infrared lamp beads with luminous center bands of 810nm, 910nm and 1060nm, respectively, and a half-wave width of 20nm; the imaging center wavelengths of the three narrow-band near-infrared cameras are 810nm, 910nm and 1060nm, respectively, and the imaging band range is 20nm (i.e., the center wavelength ±10nm); a large band interval is maintained between the three imaging bands to avoid mutual interference between the bands, and the device has high luminous efficiency and imaging quantum efficiency, thereby improving imaging quality.

[0018] Furthermore, in the multi-band adaptive regional intelligent fill light imaging device, the horizontal and vertical field of view angles of the camera are H and V respectively, and H is greater than V. Then the focusing angle 2γ of the focusing lens is determined according to the camera field of view, satisfying the formula 2γ=3V / 5, where γ is the light beam divergence half-angle of the focusing lens; the angle α1 in the horizontal plane of the lamp panel is 2(H-2γ) / 3, the angle α2 in the horizontal plane is 2(H-2γ) / 9, and the angle β in the vertical plane is 3(V-2γ) / 4; so that the imaging area has progressive lighting coverage from the center to the edge, ensuring uniform distribution of light intensity in the imaging area.

[0019] Furthermore, the multi-band adaptive regional intelligent fill light imaging device uses a two-stage amplifier consisting of a transistor and a MOS tube to form a driver; a single-chip microcomputer is used as a controller to provide a PWM signal to the transistor base of the driver to adjust the driving current of each group of lamp beads.

[0020] The method comprises the following core steps: acquiring scene images using an imaging device comprising narrowband near-infrared cameras of three different wavelengths and narrowband near-infrared lamps of corresponding wavelengths; detecting the average brightness of the image in real time by region, and automatically identifying the imaging areas of reflected light and incident light; employing deep learning technology to distinguish different types of illumination sources, and determining the wavelengths for the next frame of imaging and fill-light based on whether there is a region of strong incident light, thereby achieving multi-band adaptive intelligent fill-light imaging and resolving interference from incident light of the same wavelength as the imaging; adjusting the fill-light intensity of the corresponding region based on the average brightness of each region, thereby achieving regional intelligent fill-light, ensuring uniform fill-light imaging, and avoiding local overexposure. The central wavelengths of the narrowband near-infrared cameras and lamps of the three different wavelengths are 810 nm, 910 nm, and 1060 nm, respectively; and the imaging device comprises multiple groups of lamps of three different wavelengths distributed in different spatial locations, installed at different angles to provide independent illumination adjustment for each imaging region and switch imaging bands.

[0021] Furthermore, the specific working process of the multi-band adaptive regional intelligent fill light imaging method is as follows:

[0022] Step 1: Initialize and set the detector to collect 810nm band camera images; set the controller to control the driver to supply the maximum current to the lamp bead group in this band; then proceed to step 2 in sequence;

[0023] Step 2: The detector collects an image frame from the camera in the current working band and detects the average brightness of each area in the image; the image is divided into 5 brightness levels from dark to bright based on the average brightness; the number of areas N with brightness levels exceeding level 3 is counted; and the process proceeds to step 3 in sequence;

[0024] Step 3: The detector determines that N>0, which indicates that there is a local strong light imaging area, then proceed to step 4 to further determine the cause of the strong light imaging area, otherwise jump to step 9;

[0025] Step 4: The detector uses the deep learning model to predict whether the area with brightness exceeding level 3 is reflected light or incident light imaging; and counts the number M of imaging areas with brightness exceeding level 3 due to strong incident light; then proceeds to step 5 in sequence;

[0026] Step 5: The detector determines that M>0. If it is greater than 0, it indicates that there is a strong interfering light source in the same wavelength band as the imaging in the environment. Then jump to step 10 and switch the imaging and illumination bands to avoid interference from the strong light source in the same wavelength band. Otherwise, go to step 6.

[0027] Step 6: Check whether the detector receives a stop signal. If yes, go to step 7; otherwise, jump to step 11.

[0028] Step 7: Set the detector to collect 810nm band camera images; set the controller to control the driver to not power the lamp bead group in this band; and send the working band signal to the controller; then proceed to step 8 in sequence;

[0029] Step 8: The detector stops detecting;

[0030] Step 9: Set the detector's operating band to 810nm, set the maximum current for each zone's lamp group, and send the operating band signal to the controller. Use the 810nm band lamps and camera to achieve high-efficiency, low-energy imaging. Go to step 12.

[0031] Step 10: The detector automatically switches the working band to any of the other two bands and sends the working band signal to the controller; jump to step 12;

[0032] Step 11: The detector determines the driving current intensity of each area according to the brightness level of each area and sends it to the controller; among them, the current intensities corresponding to the five brightness levels [0,50), [50,100), [100,150), [150,200), and [200,255) are 2.0A, 1.5A, 1.0A, 0.5A, and 0A respectively. The current intensity is inversely proportional to the brightness. As the brightness level increases, the set current decreases successively; jump to step 12;

[0033] Step 12: The controller controls the driver to supply power to the lamp beads in the current working band according to the working band and the current intensity of each area; after completing a round of imaging fill light, return to step 2.

[0034] Furthermore, the multi-band adaptive regional intelligent fill light imaging method is mainly used in security monitoring, autonomous driving and intelligent transportation in low-visibility environments such as night, haze, and heavy rain. By dynamically switching fill light and imaging in different bands, it effectively solves the problems of ambient light source interference and uneven lighting, overcomes the technical defects of traditional single-band imaging systems such as local overexposure, overall underexposure and light source interference under complex lighting conditions, significantly improves the image's information acquisition capability, contrast and clarity, enhances the target recognition accuracy and real-time perception capability of the autonomous driving system in harsh environments, and thus improves traffic safety.

[0035] Furthermore, the device for implementing the multi-band adaptive regional intelligent fill light imaging method is the above-mentioned multi-band adaptive regional intelligent fill light imaging device; wherein, the detector performs the function of real-time detection of the image brightness mean by region and automatic identification of the reflected light and incident light imaging areas; the controller performs the function of determining the next frame imaging and fill light band according to the strong incident light area and adjusting the fill light intensity according to the brightness mean; the driver performs the function of switching between different bands of lamp beads and controlling the light intensity; the narrow-band camera and narrow-band lamp beads respectively realize multi-band adaptive intelligent fill light imaging and regional intelligent fill light.

[0036] This invention proposes a novel intelligent dimming method and device that uses intelligent regional fill light and band switching to address the impact of multiple interference factors on imaging quality, such as the uneven reflection of the same-band incident light source and the environment. Specific technologies include:

[0037] Intelligent band switching: A deep learning model is used to predict whether high-brightness areas are imaged by reflected light or incident light. When facing strong incident light, the imaging and lighting bands are automatically switched to avoid overlap with interfering ambient light bands, thereby improving image quality.

[0038] Regional light intensity adjustment: In the face of strong reflected light, the system automatically detects the local exposure intensity of different areas in the image in real time, identifies overexposed areas, and intelligently adjusts the fill light intensity to ensure uniform imaging and avoid overexposure.

[0039] Intelligent fill light imaging device: The present invention designs a precise regional intelligent fill light device that can adjust the fill light intensity in real time according to the light intensity differences in different areas, improve the imaging effect, and support switching of fill light bands.

[0040] Through the above technology, the present invention not only takes into account the interference from the same-band light sources in the environment, but also solves the problem of uneven reflection of the environment on its own light source. Through intelligent adjustment and band switching technical means, it effectively reduces the impact of complex lighting on imaging quality, significantly improves the imaging quality of the night driving assistance system, and ensures that clear and stable visual information is obtained in complex traffic environments, thereby improving the safety of night driving. BRIEF DESCRIPTION OF THE DRAWINGS

[0041] Figure 1 This is the main view of the intelligent fill light imaging device;

[0042] Figure 2 This is a top view of the intelligent fill light imaging device;

[0043] Figure 3 This is the left view of the intelligent fill light imaging device;

[0044] Figure 4 This is a flow chart of the intelligent fill light imaging method;

[0045] Figure 5 This is the intelligent fill light control block diagram.

[0046] Figure 1-Figure 3 In the figure, 1 is the focusing lens, 2 is the lamp panel, 3 is the angle adjustment bracket, 4 is the main bracket, 5 is the controller, 6 is the narrow-band 810nm camera, 7 is the narrow-band 910nm camera, 8 is the narrow-band 1060nm camera, 9 is the narrow-band 1060nm lamp bead, 10 is the narrow-band 910nm lamp bead, 11 is the narrow-band 810nm lamp bead, 12 is the driver, and 13 is the detector. DETAILED DESCRIPTION

[0047] In order to make the purpose, technical solutions and benefits of the present invention more clear and explicit, the present invention is further described in detail below with reference to the accompanying drawings and examples.

[0048] The present invention proposes a multi-band adaptive regional intelligent fill light imaging method and fill light device, which includes a condenser lens 1, a lamp panel 2, an angle adjustment bracket 3, a main bracket 4, a controller 5, a narrow-band 810nm camera 6, a narrow-band 910nm camera 7, a narrow-band 1060nm camera 8, a 1060nm narrow-band near-infrared lamp bead 9, a 910nm narrow-band near-infrared lamp bead 10, a narrow-band 810nm lamp bead 11, a driver 12 and a detector 13. Figure 1-Figure 3As shown, the horizontal and vertical field angles of view of the narrowband 810nm camera 6, narrowband 910nm camera 7, and narrowband 1060nm camera 8 are 30 degrees and 20 degrees respectively, and they are installed in the lower center of the main bracket 4. The optical axes of the three band cameras are parallel to each other and perpendicular to the mounting plane of the main bracket 4; the lamp beads are divided into three rows and four columns, a total of twelve groups, which are welded to twelve lamp panels 2 respectively, and the optical axis of each lamp bead is perpendicular to the mounting plane of the lamp panel 2; the twelve lamp panels 2 are respectively fixed on twelve angle adjustment brackets 3, and the angle adjustment brackets 3 are tilted at different angles. The four angle adjustment brackets 3 installed at the bottom row are installed so that the optical axis of the lamp beads thereon is offset downward by β=6 degrees relative to the optical axis of the camera in the vertical direction; the four angle adjustment brackets 3 installed at the middle row are installed so that the optical axis of the lamp beads thereon is offset vertically by 0 degrees relative to the optical axis of the camera; the four angle adjustment brackets 3 installed at the top row are installed so that the optical axis of the lamp beads thereon is offset upward by β=6 degrees relative to the optical axis of the camera in the vertical direction; the three angle adjustment brackets 3 installed at the leftmost column are installed so that the optical axis of the lamp beads thereon is offset vertically by β=6 degrees relative to the optical axis of the camera The camera optical axis is offset to the left by an angle of α1 = 12 degrees in the horizontal direction; the three tilted angle adjustment brackets 3 in the second row on the left make the optical axis of the lamp beads on them offset to the left by an angle of α2 = 4 degrees in the horizontal direction relative to the camera optical axis; the three tilted angle adjustment brackets 3 in the third row on the left make the optical axis of the lamp beads on them offset to the right by an angle of α2 = 4 degrees in the horizontal direction relative to the camera optical axis; the three tilted angle adjustment brackets 3 in the rightmost row make the optical axis of the lamp beads on them offset to the right by an angle of α1 = 12 degrees in the horizontal direction relative to the camera optical axis; a large group of lamp beads on the same lamp panel 2 It is further divided into three bands, and the lamp beads of the same band group are connected in series into a small group; the three band group lamp beads on the same lamp panel 2 are powered by the same driver 12-time sharing, and only one band group lamp bead is energized at the same time, and the energized band and its intensity are controlled by the controller 5 to realize the control of the on and off of the three band group lamp beads and the light intensity; each lamp bead is covered with a focusing lens 1; the focusing angle of the focusing lens 1 is smaller than the camera imaging angle of view, which is 12 degrees in this case, aiming to aggregate light and disperse it to cover the entire imaging area, so as to realize independent fill light and light intensity adjustment for different imaging areas.

[0049] Specifically, the three bands of narrow-band near-infrared lamp beads and cameras are 810nm, 910nm and 1060nm respectively; the half-wave width of the light emitted by the three lamp beads is 20nm; the imaging wavelength range of the three narrow-band near-infrared cameras is 20nm (i.e., the center wavelength ±10nm); a large wavelength interval is maintained between the three imaging bands to avoid mutual interference between the bands, and they have high luminous efficiency and imaging quantum efficiency to ensure imaging quality.

[0050] The present invention also provides another embodiment, a multi-band adaptive regional intelligent fill light imaging method. First, the brightness average of the imaging image is detected in real time by region, and the overexposed or underexposed areas are automatically identified; then, the fill light intensity of the corresponding area is adjusted according to the brightness average of each area to ensure uniform fill light imaging and avoid local overexposure; finally, the imaging and fill light bands are automatically switched according to the brightness average of the entire image; the intelligent fill light imaging device cooperates with the intelligent fill light imaging method to avoid interference from strong light sources in the same band in the environment, and avoid interference from uneven reflection of the environment on the fill light itself, thereby improving imaging quality.

[0051] Specifically, the workflow is as follows Figure 4 As shown, its control block diagram is as follows Figure 5 As shown below, the specific workflow analysis is as follows:

[0052] Step 1: Initialization, setting the detector 13 to collect the camera image of the 810nm band; setting the controller 5 to control the driver 12 to supply the maximum current (for example, 2.0A) to the lamp bead group of the band;

[0053] Step 2: The detector 13 captures an image frame from the camera in the current working band and detects the average brightness of each area in the image; the image is divided into 5 brightness levels from dark to bright according to the average brightness; and the number N of areas with brightness levels exceeding level 3 is counted;

[0054] Based on the acquired image, the detector 13 calculates the average brightness of each imaging area; the system divides the average brightness into five levels: brightness level 1: [0, 50] (very dark); brightness level 2: [50, 100] (dark); brightness level 3: [100, 150] (medium); brightness level 4: [150, 200] (bright); brightness level 5: [200, 255] (very bright); brightness level 1 indicates the weakest light, and level 5 indicates the strongest light.

[0055] Region division and analysis: Based on brightness grading, different areas in the image will be assigned different brightness levels. The system will analyze areas with brightness levels exceeding level 3, which are usually areas where strong light interference may occur.

[0056] Step 3: The detector 13 determines that N>0, which indicates that there is a local strong light imaging area, then the process goes to step 4 to further determine the cause of the strong light imaging area, otherwise jump to step 9;

[0057] Step 4: The detector 13 calls the deep learning model to predict whether the area with brightness exceeding level 3 is reflected light or incident light imaging. The model can automatically analyze and identify different types of illumination; and count the number M of imaging areas with brightness exceeding level 3 due to incident light; and proceed to step 5 in sequence; the deep learning model is trained to distinguish between reflected light areas (such as ground reflection or object surface reflection) and incident light areas (such as direct light generated by other light sources) in the image. The model receives image data as input and outputs the classification results of the illumination type of each area; through deep learning analysis, if a strong interfering light source is detected in the environment (i.e., a strong light source with the same imaging band), the system will automatically switch to another band for imaging and fill light; for example, when a strong interfering light source is detected in the 810nm band, the system will switch to the 910nm or 1060nm band to avoid interference;

[0058] Step 5: The detector 13 determines that M>0, which indicates that there is a strong interfering light source in the same wavelength band as the imaging in the environment, then jumps to step 10, switching the imaging and illumination bands to avoid interference from the strong light source in the same wavelength band. Otherwise, go to step 8;

[0059] Step 6: Check whether the detector 13 receives the stop signal. If yes, go to step 7; otherwise, jump to step 11.

[0060] Step 7: The detector 13 is set to collect the camera image of the 810nm band; the controller 5 is set to control the driver 12 to not power the lamp bead group in this band; and send the working band signal to the controller 5;

[0061] Step 8: The detector 13 stops detecting;

[0062] Step 9: The detector 13 sets the operating band to 810nm, sets the maximum current for each zone lamp group, and sends the operating band signal to the controller 5. This is because the 810nm band lamps have the highest luminous efficiency, and the 810nm band narrow-band near-infrared camera has the highest quantum efficiency, so its imaging quality is the best at night and the energy consumption is the lowest.

[0063] Step 10: The detector 13 automatically switches the working band to any one of the other two bands and sends the working band signal to the controller 5;

[0064] Step 11: The detector 13 determines the driving current intensity of each area according to the brightness level of each area, which is inversely proportional to each other, and sends it to the controller 5;

[0065] The relationship between current and brightness is inversely proportional, that is, the higher the brightness, the lower the current intensity; the specific relationship is as follows:

[0066] For brightness level 1 ([0, 50)), the current intensity is 2.0A;

[0067] For brightness level 2 ([50, 100)), the current intensity is 1.5A;

[0068] Corresponding to brightness level 3 ([100, 150)), the current intensity is 1.0A;

[0069] Corresponding to brightness level 4 ([150, 200)), the current intensity is 0.5A;

[0070] Corresponding to brightness level 5 ([200, 255]), the current intensity is 0A (that is, the lamp beads in this area are turned off);

[0071] The controller 5 adjusts the current intensity according to the brightness level. For example, in an area with a brightness level of 1 (very dark), the current is 2.0A to ensure that the area receives sufficient light; while in an area with a brightness level of 5 (very bright), the current is reduced to 0A to avoid overexposure.

[0072] Step 12: The controller 5 controls the driver 12 to supply power to the lamp beads of the current working band according to the working band and the current intensity of each area;

[0073] During the entire process, the system continuously adjusts the current, band switching and fill light intensity through real-time feedback from the detector to ensure uniform lighting in the entire imaging area and avoid local overexposure or insufficient lighting.

[0074] During nighttime autonomous driving, the 810nm band LEDs offer high luminous efficiency, providing clear imaging with low power consumption. Through the multi-band adaptive intelligent fill-light device, the autonomous driving system can cope with complex nighttime lighting conditions, enhance environmental perception, and ensure driving safety.

[0075] In adverse weather conditions (such as fog or heavy rain), the system can automatically adjust the fill light intensity to ensure that the camera obtains clear images in low-visibility environments, avoiding driving safety affected by insufficient light or reflected light.

[0076] This example describes in detail the implementation of a multi-band adaptive, regionalized intelligent fill-light imaging device, including brightness grading, current adjustment, the application of deep learning methods, and band switching. Through these optimizations and controls, the system achieves efficient fill-light imaging, ensuring image quality under varying lighting conditions, with significant value in autonomous driving and intelligent transportation applications.

Claims

1. A multi-band adaptive regional intelligent fill light imaging device, characterized in that: include: Focusing lens (1), lamp panel (2), angle adjustment bracket (3), main bracket (4), controller (5), narrow-band 810nm camera (6), narrow-band 910nm camera (7), narrow-band 1060nm camera (8), narrow-band 1060nm lamp beads (9), narrow-band 910nm lamp beads (10), narrow-band 810nm lamp beads (11), driver (12) and detector (13); wherein the narrowband 810nm camera (6), the narrowband 910nm camera (7) and the narrowband 1060nm camera (8) are horizontally mounted in the middle of the lower portion of the main support (4), with their optical axes being parallel to each other and perpendicular to the mounting plane of the main support (4); The lamp beads are divided into three rows and four columns, with a total of twelve groups, which are respectively welded on twelve lamp panels (2), and the optical axis of each lamp bead is perpendicular to the installation plane of the lamp panel (2); Twelve light panels (2) are respectively fixed on twelve angle adjustment brackets (3), and the angle adjustment brackets (3) at different positions are installed on the main bracket (4) at different tilt angles and are located above the camera to achieve regional intelligent fill light; Lamp beads with three wavelength bands of 810nm, 910nm and 1060nm are welded on the same lamp panel (2), and the lamp beads of each wavelength band are connected in series into a group, forming three groups to realize multi-band adaptive intelligent imaging; The three wavelength group lamp beads on the same lamp panel (2) are powered by the same driver (12) in a time-sharing manner. At any time, only one wavelength group lamp bead is powered. The detector (13) determines the wavelength and intensity. The controller (5) controls the driver (12) to switch between the different wavelength group lamp beads and control the light intensity. Each lamp bead is covered with a condensing lens (1); the condensing angle of the condensing lens (1) is smaller than the camera imaging angle of view, and the condensing lens (1) condenses light and disperses it to cover the entire imaging area, thereby achieving independent fill light and light intensity adjustment for different imaging areas; Assuming that the horizontal and vertical field angles of the camera are H and V respectively, and H is greater than V, the focusing angle 2γ of the focusing lens (1) is determined according to the camera field angle, satisfying the formula 2γ=3V / 5, wherein γ is the light beam divergence half angle of the focusing lens (1); the horizontal tilt angles α1=2(H-2γ) / 3, α2=2(H-2γ) / 9 of the twelve angle adjustment brackets (3) for fixing the lamp panel (2), and the vertical tilt angle β=3(V-2γ) / 4; so that the imaging area has progressive illumination coverage from the center to the edge, ensuring uniform illumination intensity distribution in the imaging area.

2. The multi-band adaptive regional intelligent fill light imaging device according to claim 1, characterized in that: The horizontal tilt angles α1 and α2 and the vertical tilt angle β of the twelve angle adjustment brackets (3) mounted on the main bracket (4) are set as follows: the four angle adjustment brackets (3) mounted at an angle in the bottom row are arranged so that the optical axis of the lamp beads thereon tilts downward, forming an angle of β degrees with the optical axis of the camera in a vertical plane; The four angle adjustment brackets (3) in the middle row are installed so that the optical axis of the lamp beads thereon and the optical axis of the camera are in the same horizontal plane; The four angle adjustment brackets (3) in the top row are installed so that the optical axis of the lamp beads thereon is tilted upward, forming an angle of β degrees with the optical axis of the camera in the vertical plane; The three angle adjustment brackets (3) in the leftmost column are installed so that the optical axis of the lamp beads thereon is tilted to the left, forming an angle of α1 degrees with the optical axis of the camera in the horizontal plane; The three angle adjustment brackets (3) in the second column on the left are installed so that the optical axis of the lamp beads on them is tilted to the left, forming an angle of α2 degrees with the optical axis of the camera in the horizontal plane; The three angle adjustment brackets (3) in the third column on the left are installed so that the optical axis of the lamp beads on them is tilted to the right, forming an angle of α2 degrees with the optical axis of the camera in the horizontal plane; The three angle adjustment brackets (3) installed in the rightmost column are used to tilt the optical axis of the lamp beads thereon to the right, forming an angle of α1 degrees with the optical axis of the camera in the horizontal plane.

3. The multi-band adaptive regional intelligent fill light imaging device according to claim 1, characterized in that: The luminous center bands of the three narrow-band near-infrared lamp beads are 810nm, 910nm and 1060nm, respectively, and the half-wave width is 20nm; the imaging center wavelengths of the three narrow-band near-infrared cameras are 810nm, 910nm and 1060nm, respectively, and the imaging band range is 20nm.

4. The multi-band adaptive regional intelligent fill light imaging device according to claim 1, characterized in that: A two-stage amplifier consisting of a triode and a MOS tube is used to form a driver (12); a single chip microcomputer is used as a controller (5) to provide a PWM signal to the triode base of the driver (12) to adjust the driving current of each group of lamp beads.

5. A multi-band adaptive regional intelligent fill light imaging method, characterized in that: A multi-band adaptive regional intelligent fill-light imaging device according to any one of claims 1 to 4 is used, and the method comprises the following core steps: using the narrow-band 810nm camera (6), the narrow-band 910nm camera (7), the narrow-band 1060nm camera (8) and the corresponding narrow-band 1060nm lamp beads (9), the narrow-band 910nm lamp beads (10), and the narrow-band 810nm lamp beads (11) to acquire a scene image; detecting the brightness mean of the image in real time by region, and automatically identifying the imaging area of reflected light and incident light; determining the band of the next frame of imaging and fill-light according to whether there is a strong incident light area, realizing multi-band adaptive intelligent fill-light imaging, and solving the interference of incident light in the same band as the imaging; adjusting the fill-light intensity of the corresponding area according to the brightness mean of each area, realizing regional intelligent fill-light, ensuring uniform fill-light imaging, and avoiding local overexposure.

6. The multi-band adaptive regional intelligent fill light imaging method according to claim 5, characterized in that: The specific workflow is as follows: Step 1: Initialization, setting the detector (13) to collect the camera image of the 810nm band; setting the controller (5) to control the driver (12) to supply the maximum current to the lamp beads of the band group; and sequentially entering step 2; Step 2: The detector (13) collects a frame of image from the camera of the current working band, detects the brightness mean of each area in the image; divides the image into 5 brightness levels from dark to bright according to the brightness mean; counts the number N of areas with brightness levels exceeding level 3; and sequentially proceeds to step 3; Step 3: The detector (13) determines that N>0, which indicates that there is a local strong light imaging area, then proceeds to step 4 to further determine the cause of the strong light imaging area, otherwise jumps to step 9; Step 4: The detector (13) uses a deep learning model to predict whether the area with brightness exceeding level 3 is reflected light or incident light imaging, and the model can classify and judge the type of illumination based on image features; and counts the number M of imaging areas with brightness exceeding level 3 due to incident light; Go to step five in sequence; Step 5: The detector (13) determines that M>0, which indicates that there is a strong interfering light source in the same wavelength band as the imaging in the environment, then jump to step 10, switch the imaging and illumination bands to avoid the strong light source in the same wavelength band interfering with the imaging, otherwise go to step 6; Step 6: The detector (13) determines whether a stop signal is received, and if so, proceeds to step 7, otherwise, jumps to step 11; Step 7: Set the detector (13) to collect the camera image of the 810nm band; set the controller (5) to control the driver (12) to not power the lamp beads of the band group; and send the working band signal to the controller (5); and proceed to step 8 in sequence; Step 8: The detector (13) stops detecting; Step 9: Set the detector (13) to work in a wavelength band of 810nm, set the lamp bead groups in each area to supply the maximum current, and send the working wavelength band signal to the controller (5); use the 810nm wavelength lamp bead and camera to achieve high-efficiency and low-energy imaging; jump to step 12; Step 10: The detector (13) automatically switches the working band to any one of the other two bands and sends the working band signal to the controller (5); Skip to step 12; Step 11: The detector (13) determines the driving current intensity of each area according to the brightness level of each area and sends it to the controller (5); wherein, the current intensities corresponding to the five brightness levels [0, 50), [50, 100), [100, 150), [150, 200), and [200, 255) are 2.0A, 1.5A, 1.0A, 0.5A, and 0A, respectively. The current intensity is inversely proportional to the brightness. As the brightness level increases, the set current decreases successively; jump to step 12; Step 12: The controller (5) controls the driver (12) to supply power to the lamp beads of the current working band according to the working band and the current intensity of each area; after completing a round of imaging fill light, the process returns to step 2.

7. The multi-band adaptive regional intelligent fill light imaging method according to claim 5, characterized in that: It is used in security monitoring, autonomous driving and intelligent transportation in low-visibility environments such as night, fog, haze and heavy rain.

Images