Intelligent light switching method and system and related devices

JP7582581B2Active Publication Date: 2024-11-13YINWANG INTELLIGENT TECHNOLOGIES CO LTD
View PDF 6 Cites 0 Cited by

Patent Information

Application Number
JP2023526210
Authority / Receiving Office
JP · JP
Patent Type
Patents
Current Assignee / Owner
Priority Date
2020-10-31
Filing Date
2021-08-26
Publication Date
2024-11-13
Estimated Expiration
2041-08-26

AI Technical Summary

Benefits of technology

【0127】 加えて、上記の基準値及び減衰係数を導入するための方法は、本願における例示の方法にすぎない。特定の実施プロセスでは、より良い有益な効果を得るために、方法は実際の要件に基づいて調整又は変更され得る。これは本願では限定されない。

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure 0007582581000002
    Figure 0007582581000002
  • Figure 0007582581000003
    Figure 0007582581000003
  • Figure 0007582581000004
    Figure 0007582581000004
Patent Text Reader

Abstract

This application provides an intelligent light switching method and system, and related devices. The method includes using an intelligent light switching system to acquire an image, the image being captured by a video camera disposed at a fixed position on a vehicle and recording light source information in the image, calculate an ambient light brightness value corresponding to the image, classify the light sources contained in the image based on the light source information to obtain a classification result, and switch to high beam or low beam based on the ambient light brightness value corresponding to the image and the classification result. Because the method combines two types of information, namely ambient light information and lamp light source information, more information about the road at night can be used as a reference when deciding whether to switch to high beam or low beam, thereby improving the accuracy of light switching.
Need to check novelty before this filing date? Find Prior Art

Description

[Technical Field]

[0001] This application claims priority to Chinese Patent Application No. 202011197967.5, entitled "Intelligent Light Switching Method and System and Related Apparatus," filed with the State Intellectual Property Office of China on October 31, 2020, which is incorporated herein by reference in its entirety.

[0002] This application relates to the field of intelligent vehicles, and more particularly to intelligent light switching methods and systems and related devices. [Background technology]

[0003] It is well known that nighttime driving requires high and low beams for illumination to help drivers obtain road information. Compared with low beams, high beams have a significant effect of improving visibility and widening the field of view. Particularly on dark roads without streetlights, the visibility range when high beams are turned on is much greater than that when low beams are turned on. However, using high beams is not applicable to all nighttime driving scenarios. For example, if a driver driving a vehicle at night turns on high beams when encountering a vehicle traveling in the opposite direction, the driver of the vehicle traveling in the opposite direction may be instantly blinded or may become less aware of the speed and distance and be unable to judge the width. Therefore, correct use of high beams is important for safe nighttime driving. To fundamentally address the problem of drivers using high beams incorrectly, intelligent light switching technology has emerged.

[0004] Among existing intelligent light switching methods, drivers can be assisted in switching between high and low beams by three methods: distinguishing light sources based on different image features presented by the light sources with different exposure times, directly calculating the brightness of the front ambient light, and obtaining a vanishing point based on the lane. However, the implementation effects of the above methods are not ideal. The implementation of the first method requires relatively high requirements for the camera module, which results in relatively high costs. The second method ignores the influence of the brightness of distant headlights, which leads to relatively large errors in the calculation of the brightness of the ambient light. The implementation of the third method relies on accurate lane detection, but the probability of accurately detecting the lane at night is very low.

[0005] Therefore, how to switch to high beam or low beam accurately is currently an urgent problem to be solved. Summary of the Invention

[0006] This application provides an intelligent light switching method and system, as well as related devices, that can accurately detect the brightness of ambient light and accurately classify light source information. After the two types of information are combined, the high beam or low beam switching function can be accurately performed, effectively solving the problem of not being able to detect distant light sources and avoiding the influence of interfering light sources.

[0007] According to a first aspect, the present application provides an intelligent light-switching method, the method including: acquiring an image, the image being captured by a video camera disposed at a fixed position on a vehicle and having light source information recorded in the image; calculating an ambient light brightness value corresponding to the image; classifying light sources included in the image based on the light source information to obtain a classification result; and switching to high beam or low beam (switching lights) based on the ambient light brightness value corresponding to the image and the classification result.

[0008] The solution provided in this application includes successive steps of obtaining an image, calculating the brightness of ambient light, classifying the light source, and judging whether to switch on the lights based on the result. Since the light switching method combines ambient light information and light source information, more information about the road at night can be used as reference while judging whether to switch on the lights, thereby improving the accuracy of switching on the lights.

[0009] With reference to the first aspect, in a possible implementation of the first aspect, calculating an ambient light brightness value corresponding to the image includes selecting at least one region from the image, calculating a brightness value of the at least one region, and calculating an ambient light brightness value corresponding to the image based on the brightness value of the at least one region.

[0010] In the solution provided herein, the ambient light brightness value is calculated by selecting at least one region, calculating the brightness value of the at least one region, and obtaining the ambient light brightness value based on the brightness value. In this way, information about light sources that may exist in different regions of the image is fully taken into account. Compared with calculating the ambient light brightness based only on the entire image, the method provided herein can obtain the ambient light brightness more accurately.

[0011] Referring to the first aspect, in a possible implementation of the first aspect, classifying the light source included in the image to obtain a classification result includes inputting the image to a light source detection model, and obtaining a light source category in the image based on the light source detection model.

[0012] In the solution provided in this application, light source classification is performed by using a light source detection model, which, after training, can quickly and accurately perform light source classification based on features such as light source size, position, and color.

[0013] Referring to the first aspect, in a possible implementation of the first aspect, before inputting the image into the light source detection model, the method includes training the light source detection model by using a plurality of sample images, the sample images including the light source and annotation information for the light source.

[0014] In the solution provided in this application, the illuminant detection model needs to be trained before it can be officially used. Training is the process of recognizing the initial model and learning the illuminant information in the sample image. After the training is completed, the illuminant detection model can accurately recognize the different characteristics of different illuminants, thereby improving the accuracy of illuminant classification.

[0015] Referring to the first aspect, in a possible implementation of the first aspect, obtaining a light source category in the image based on the light source detection model includes selecting a light point from the image whose brightness value is greater than a preset threshold and setting a lamp box for the light point; pairing the lamp boxes to obtain a plurality of lamp box pairs; determining a VP line based on the plurality of lamp box pairs, wherein the VP line is used to distinguish between a first region and a second region of the image; and classifying the light point based on a positional relationship between the light point and the VP line and a color characteristic of the light point to obtain different light source categories.

[0016] In the solution provided in this application, to perform light source classification, it is first necessary to obtain bright points in the image that have relatively high brightness and to prepare lamp boxes. This is a preliminary filtering of the light sources. Next, light source pairing is performed based on the relationship between vehicle lamp pairs. Compared with single recognition, this pairing method improves classification efficiency and accuracy. Next, a vanishing point horizon line (VP line) is determined based on the distribution of lamp box pairs in the image. Finally, the bright points are classified based on the positional relationship between the VP line and the bright points and their characteristics such as brightness and color. After the VP line is determined, it can be understood that areas including vehicle lamps and areas including street lamps and traffic lights can be roughly separated, making the subsequent classification process more accurate and effective.

[0017] Referring to the first aspect, in a possible implementation of the first aspect, before performing pairing on the lamp boxes, the method further includes performing overlap elimination on the lamp boxes to ensure that there are no overlaps or tangents between the lamp boxes.

[0018] The solution provided in this application requires that the lamp boxes undergo de-duplication before pairing to ensure that there are no overlaps or tangents between them, thus ensuring that the lamp boxes do not influence each other in the subsequent classification process and improving the accuracy of light source classification.

[0019] Referring to the first aspect, in a possible implementation of the first aspect, determining a vanishing point horizon line (VP) based on the plurality of lamp box pairs includes: mapping the centers of the lamp boxes of the lamp box pairs to rows of the image to obtain a row intra-volume distribution map of the lamp box pairs; selecting a VP line based on the row intra-volume distribution map; correcting the VP line based on a preset correction value to obtain a corrected VP line; and adjusting the corrected VP line by using a reference value to obtain a VP line used to classify the bright point, wherein the reference value is used to describe changes in the pitch angle of the video camera.

[0020] In the solution provided in this application, the VP line is first determined based on the in-row dose distribution map of a pair of lamp boxes, and then the VP line needs to be corrected by using the introduced reference value so that the final VP line is more accurate, thus improving the accuracy of the subsequent light source classification.

[0021] According to a second aspect, the present application provides an intelligent light switching system, including: an acquisition unit configured to acquire an image, the image being taken by a video camera disposed at a fixed position on a vehicle, the image having light source information recorded therein; an ambient light detection unit configured to calculate an ambient light brightness value corresponding to the image; a light source classification unit configured to classify light sources included in the image based on the light source information to obtain a classification result; and a switching unit configured to switch lights based on the ambient light brightness value corresponding to the image and the classification result.

[0022] Referring to the second aspect, in a possible implementation of the second aspect, the ambient light detection unit is specifically configured to: select at least one region from the image; calculate a brightness value of the at least one region; and calculate a brightness value of ambient light corresponding to the image based on the brightness value of the at least one region.

[0023] Referring to the second aspect, in a possible implementation of the second aspect, when classifying the light source included in the image to obtain a classification result, the light source classification unit is specifically configured to input the image into a light source detection model, and determine a light source category in the image based on the light source detection model.

[0024] Referring to the second aspect, in a possible implementation of the second aspect, before inputting the image into the light source detection model, the light source classification unit is further configured to train the light source detection model by using a plurality of sample images, the sample images including the light source and a label for the light source.

[0025] Referring to the second aspect, in a possible implementation of the second aspect, when obtaining a light source category in the image based on the light source detection model, the light source classification unit is specifically configured to: select a bright point from the image whose brightness value is greater than a preset threshold and set a lamp box for the bright point; perform pairing on the lamp box to obtain a plurality of lamp box pairs; determine a vanishing point horizon line VP line based on the plurality of lamp box pairs, wherein the VP line is used to distinguish between a first region and a second region of the image; and classify the bright point based on the positional relationship between the bright point and the VP line and the color characteristics of the bright point to obtain different light source categories.

[0026] Referring to the second aspect, in a possible implementation of the second aspect, before performing pairing on the lamp boxes, the light source classification unit is further configured to perform overlap elimination on the lamp boxes to ensure that there are no overlaps or tangents between the lamp boxes.

[0027] Referring to the second aspect, in a possible implementation of the second aspect, when determining a vanishing point horizon line VP line based on the plurality of lamp box pairs, the light source classification unit is specifically configured to: map the centers of the lamp boxes of the lamp box pairs to rows of the image to obtain a row intra-volume distribution map of the lamp box pairs; select a VP line based on the row intra-volume distribution map; correct the VP line based on a preset correction value to obtain a corrected VP line; and adjust the corrected VP line by using a reference value to obtain a VP line used to classify the bright point, wherein the reference value is used to describe a change in the pitch angle of the video camera.

[0028] According to a third aspect, there is provided a computing device, the computing device including a processor and a memory, the memory configured to store program code, and the processor configured to execute the program code stored in the memory, thereby performing the intelligent light switching method according to the first aspect, with reference to any implementation of the first aspect.

[0029] According to a fourth aspect, there is provided a computer-readable storage medium having a computer program stored therein, the computer program, when executed by a processor, causing the processor to perform the intelligent light switching method according to the first aspect, with reference to any implementation of the first aspect.

[0030] According to a fifth aspect, there is provided a computer program product comprising instructions that, when executed by a computer, cause the computer to perform the intelligent light switching method according to the first aspect, with reference to any implementation of the first aspect. [Brief explanation of the drawings]

[0031] [Figure 1]FIG. 1 is a schematic diagram of an application scenario of an intelligent light switching method according to an embodiment of the present application. [Figure 2] FIG. 2 is a schematic diagram of the structure of an intelligent vehicle according to one embodiment of the present application. [Figure 3] FIG. 3 is a schematic diagram of the structure of an intelligent light switching system according to one embodiment of the present application. [Figure 4] FIG. 4 is a schematic diagram of the structure of an initial light source detection model according to one embodiment of the present application. [Figure 5] FIG. 5 is a schematic diagram of a method for extracting illuminant features from an image according to an embodiment of the present application. [Figure 6] FIG. 6 is a schematic diagram of a lamp box arrangement for a first bright spot according to one embodiment of the present application. [Figure 7] FIG. 7 is a schematic diagram of performing overlap elimination for lamp boxes according to an embodiment of the present application. [Figure 8] FIG. 8 is a schematic diagram of a lamp box pairing method according to an embodiment of the present application. [Figure 9] FIG. 9 is a schematic diagram of drawing an intra-row volume distribution map according to an embodiment of the present application. [Figure 10] FIG. 10 is a schematic diagram of determining a first row amount according to an embodiment of the present application. [Figure 11] FIG. 11 is a schematic diagram of removing the interference of light reflected by the road surface according to one embodiment of the present application. [Figure 12] FIG. 12 is a schematic diagram of an intelligent light switching method according to an embodiment of the present application. [Figure 13] FIG. 13 is a schematic diagram of the structure of a computer device according to an embodiment of the present application. DETAILED DESCRIPTION OF THE INVENTION

[0032] The technical solutions in the embodiments of the present application are clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only a part of, but not all of, the embodiments of the present application. All other embodiments obtained by those skilled in the art based on the embodiments of the present application without creative efforts shall fall within the protection scope of the present application.

[0033] To help those skilled in the art gain a better understanding, some terms and related art in the present application will first be explained with reference to the accompanying drawings.

[0034] The RGB color space uses different degrees of overlap based on the primary colors red, green, and blue to generate a rich and wide range of colors. Therefore, the RGB color space is commonly known as the three-primary mode, also known as the natural color mode. Red, green, and blue represent the three primary colors of the visible spectrum. Each color is divided into 256 levels based on its brightness. When the three primary colors of colored light overlap, various intermediate colors are generated by different color mixing ratios. Display systems commonly use the RGB color space. Displays equipped with color cathode ray tubes and displaying color raster graphics use R, G, and B values ​​to drive RGB electron guns to emit electrons, which stimulate RGB phosphor powders on a fluorescent screen to emit light beams of different brightness. The overlapping light beams generate various colors.

[0035] The YUV color space contains one brightness signal and two chrominance signals. The brightness signal is usually called Y. The chrominance signals are two independent signals (one representing red and one representing blue). The two chrominance signals are usually called UVPbPr or CbCr. Therefore, the YUV color space is also called the YPbPr color space or the YCbCr color space. The brightness information is separated from the chrominance information. Different sampling rates are used for the brightness and chrominance of an image within the same frame. In the YUV color space, the brightness information Y and the chrominance information U / V are independent of each other. The Y signal component is a black-and-white grayscale map. The U and V signal components are monochrome color maps. Because human vision is more sensitive to brightness than chrominance, the YUV color space is widely used in color television systems.

[0036] The HSI color space uses three parameters, H, S, and I, to describe color characteristics: H represents the color frequency and is called hue; S represents the color depth and is called saturation; and I represents intensity or brightness.

[0037] Region of interest (ROI): In machine vision and image processing, an area isolated from a processed image by using a box, circle, ellipse, irregular polygon, etc. is called a region of interest (ROI). In machine vision software such as Halcon, OpenCV, and MATLAB, various operators and functions are usually used to obtain the region of interest (ROI) for subsequent image processing.

[0038] Intersection over Union (IOU) is the ratio of the intersection to the union of two rectangular areas. IOU is a standard for measuring the accuracy of detecting corresponding objects in a given dataset. IOU is a simple measurement standard. The task of obtaining a predicted bounding box in the output can be measured by IOU.

[0039] Advanced Driver Assistance Systems (ADAS) are active safety technologies that use various sensors installed in vehicles to collect environmental data inside and outside the vehicle in a timely manner and perform technical processing such as recognition, detection, and tracking of static and dynamic objects, allowing the driver to detect potential risks early. This attracts attention and improves safety. Sensors used in ADAS primarily include cameras, radar, lasers, and ultrasound, and can detect light, heat, pressure, or other variables used to monitor the vehicle's condition. They are usually located on the vehicle's front and rear bumpers, side-view mirrors, steering wheel, or windshield. Early ADAS technologies primarily issue alarms passively. When a potential danger to the vehicle is detected, an alarm is issued to alert the driver to abnormal conditions in the vehicle or on the road. Active intervention is also common in the latest ADAS technologies.

[0040] The Adaptive Cruise Control (ACC) system is developed based on the cruise control system. In addition to the functions of the cruise control system, i.e., driving at a speed set by the driver, it can also implement functions such as maintaining a preset following distance and automatically accelerating or decelerating based on changes in the following distance. Compared to the cruise control system, the ACC system helps the driver adjust the brake and throttle.

[0041] Road Side Unit (RSU) is an electronic toll collection system. In ETC (Electronic Collection) systems, it is installed on the road and is used for dedicated short-distance communication. The RSU communicates with an on-board unit (OBU) using DSRC (Digital Short Range Communication) to perform vehicle identification and electronic score deduction. The RSU may include a high-gain directional beam control read / write antenna and a radio frequency controller. The high-gain directional beam control read / write antenna is a microwave transceiver module that performs signal and data transmission and reception, modulation / demodulation, encoding / decoding, and encryption / decryption. The radio frequency controller is a module that controls data transmission and reception and receives information from or transmits information to a host computer.

[0042] The intelligent light switching method provided herein is performed by an intelligent light switching system. To facilitate understanding of the embodiments of the present application, an application scenario of the intelligent light switching system on which the embodiments of the present application are based will be described first. FIG. 1 is a schematic diagram of an application scenario of the intelligent light switching system. The scenario shown in FIG. 1 is an Internet of Vehicles scenario. The scenario includes multiple intelligent vehicles, a radio transmission tower, and an RSU. The intelligent light switching system may be used in an in-vehicle system of an intelligent vehicle, such as an ADAS system or an ACC system, and may intelligently switch to high beams or low beams in multiple scenarios, such as driver assistance and autonomous driving. In addition, the intelligent light switching system may alternatively be installed in the intelligent vehicle as an independent system, separate from other driver assistance systems in the intelligent vehicle.

[0043] Based on the application scenario of the intelligent light switching system, an embodiment of the present application provides an intelligent vehicle 200 for use in the application scenario of the intelligent light switching system. Figure 2 is a schematic diagram of the structure of the intelligent vehicle 200 according to an embodiment of the present application.

[0044] It should be noted that intelligent vehicle 200 may be set to a fully intelligent driving mode or a partially intelligent driving mode. It should be understood that when intelligent vehicle 200 is set to the fully intelligent driving mode, intelligent vehicle 200 may perform corresponding actions without human interaction. The actions include, but are not limited to, accelerating, decelerating, and car-following. When the intelligent vehicle is set to the partially intelligent driving mode, the corresponding actions may be performed automatically by intelligent vehicle 200 as well as by the driver. For example, the corresponding actions may include identifying the vehicle and the surrounding environment, identifying possible behaviors of at least one other vehicle in the surrounding environment, identifying a confidence level corresponding to the likelihood that the other vehicle will perform the possible behavior, and controlling intelligent vehicle 200 based on the identified information.

[0045] Intelligent vehicle 200 may include various subsystems, such as mobility system 210, sensor system 220, control system 230, one or more peripheral devices 240, one or more computer systems 250, power supply 260, and user interface 270. Optionally, intelligent vehicle 200 may include more or fewer subsystems, and each subsystem may include multiple elements. Additionally, the subsystems and elements of intelligent vehicle 200 may be interconnected in multiple ways, for example, wired or wirelessly.

[0046] Mobility system 210 may include components that provide power to intelligent vehicle 200. In one embodiment, mobility system 210 may include engine 2110, energy source 2120, transmission 2130, and wheels / tires 2140. Engine 2110 may be an internal combustion engine, a motor, an air-compression engine, or a combination of other types of engines, such as a hybrid engine comprised of a gasoline engine and a motor, or a hybrid engine comprised of an internal combustion engine and an air-compression engine. Engine 2110 converts energy from energy source 2120 into mechanical energy.

[0047] Examples of energy source 2120 include gasoline, diesel, other petroleum-based fuels, propane, other compressed gas-based fuels, ethanol, solar panels, batteries, and other power sources. Energy source 2120 may also provide energy to other systems of intelligent vehicle 200.

[0048] The transmission device 2130 can transmit mechanical power from the engine 2110 to the wheels 2140. The transmission device 2130 can include a gearbox, an actuator, and a drive shaft. In one embodiment, the transmission device 2130 can further include other components, such as a clutch. The drive shaft can include one or more shafts that can be coupled to one or more of the wheels 2140.

[0049] Sensor system 220 may include several sensors that sense information about the intelligent vehicle's 200 surrounding environment and obtain vehicle information about the intelligent vehicle. For example, sensor system 220 may include a positioning system 2210, an inertial measurement unit (IMU) 2220, a radar 2230, and a visual sensor 2240. Positioning system 2210 may include a GPS system, a BeiDou system, or another positioning system. Sensor system 220 may further include sensors for monitored internal systems of intelligent vehicle 200, such as an interior air quality monitor, a fuel gauge, and an oil temperature gauge. Data obtained by these sensors may be used to detect objects and their corresponding characteristics, including, but not limited to, position, shape, direction, and speed. Such detection and recognition is significant for the safe subsequent operation of intelligent vehicle 200.

[0050] Positioning system 2210 may be configured to determine the geographic location of intelligent vehicle 200.

[0051] IMU 2220 may sense changes in position and orientation of intelligent vehicle 200 based on inertial acceleration. In one embodiment, IMU 2220 may be a combination of an accelerometer and a gyroscope. In this case, IMU 2220 may be configured to measure the curvature of intelligent vehicle 200.

[0052] Radar 2230 may sense the surrounding environment of intelligent vehicle 200 by using wireless signals. The surrounding environment includes, but is not limited to, surrounding vehicles, infrastructure, and pedestrians. It will be understood that radar 2230 may include, but is not limited to, millimeter wave radar and lidar. In some embodiments, in addition to sensing the surrounding environment, radar 2230 may be further configured to sense the motion state of objects within the environment.

[0053] Vision sensor 2240 may be configured to capture multiple images of the environment surrounding intelligent vehicle 200. Vision sensor 2240 may include, but is not limited to, static cameras and video cameras.

[0054] Control system 230 may be configured to control the operation of intelligent vehicle 200 and components of the intelligent vehicle. Control system 230 may include multiple elements. In one embodiment, control system 230 includes a steering system 2310, an actuator 2320, a braking unit 2330, a computer vision system 2340, a path control system 2350, an obstacle avoidance system 2360, and a high / low beam switching system 2370.

[0055] The steering system 2310 may be manipulated to adjust the direction of travel of the intelligent vehicle 200. For example, in one embodiment, the steering system 2310 may include a steering wheel system.

[0056] The actuator 2320 may be configured to control the engine 2110, thereby controlling the speed of the intelligent vehicle 200. For example, in one embodiment, the actuator 2320 may include a throttle.

[0057] The brake unit 2330 may be configured to control the deceleration of the intelligent vehicle 200. The brake unit 2330 may use friction to slow the rotational speed of the wheels 2140. In another embodiment, the brake unit 2330 may convert the kinetic energy of the wheels 2140 into electric current. The brake unit 2330 may control the speed of the intelligent vehicle 200 by using another method to slow the rotational speed of the wheels 2140.

[0058] It will be appreciated that the actuator 2320 and the brake unit 2330 may be combined into one unit module. The combined unit module may be configured to control the speed of the intelligent vehicle 200. In one embodiment, the combined unit module may include a throttle system and a brake system.

[0059] The computer vision system 2340 may be configured to process and analyze images captured by the vision sensor 2240 to facilitate subsequent operation. The computer vision system 2340 may further recognize the surrounding environment of the intelligent vehicle 200, the characteristics of objects in the surrounding environment, and the motion states of the objects. The surrounding environment may include traffic lights, road boundaries, and obstacles. The characteristics of objects in the surrounding environment include, but are not limited to, surface optical characteristics of the object's surface. The motion states include, but are not limited to, static, accelerating, or decelerating. The computer vision system 2340 may use color space conversion techniques, object recognition algorithms, structure from motion (SFM) algorithms, video tracking, and other computer vision techniques. In some embodiments, the computer vision system 2340 may include an image detection system, a neural network-based processing system, etc., and may be configured to draw a map for the environment, track objects, estimate the object's speed, etc.

[0060] Route control system 2350 is configured to determine a driving route for intelligent vehicle 200. In some embodiments, route control system 2350 may reference data from one or more predetermined maps from positioning system 2210 to determine a driving route for intelligent vehicle 200.

[0061] Obstacle avoidance system 2360 is configured to recognize, evaluate, avoid, or bypass obstacles in the surrounding environment. In one embodiment, obstacle avoidance system 2360 must obtain information about the surrounding environment using radar 2230 and vision sensors 2240, and then analyze the surrounding environment using computer vision system 2340 to recognize potential obstacles. Obstacle avoidance system 2360 then performs evaluation and avoidance.

[0062] The high / low beam switching system 2370 is configured to intelligently switch between high and low beams. In one embodiment, the high / low beam switching system 2370 may be automatically enabled based on ambient light and may automatically switch to high or low beams or may be manually enabled.

[0063] It should be noted that the control system 230 may further include other components, or the above components may be replaced and / or eliminated.

[0064] Intelligent vehicle 200 interacts with external sensors, other vehicles, other computer systems, or users through the use of peripherals 240. Peripherals 240 may include, but are not limited to, a wireless communication system 2410, an on-board computer 2420, a microphone 2430, and / or a speaker 2440.

[0065] It should be noted that in some embodiments, peripheral devices 240 may interact with a user of intelligent vehicle 200. For example, onboard computer 2420 may provide information to the user of intelligent vehicle 200. Additionally, the user of intelligent vehicle 200 may upload data to onboard computer 2420. It will be appreciated that the user of intelligent vehicle 200 may operate the onboard computer 2420 by using a touchscreen. Additionally, peripheral devices 240 may provide a means for intelligent vehicle 200 to communicate with other devices within the vehicle. For example, microphone 2430 may receive audio from the user of intelligent vehicle 200, which may include voice commands and other audio input. Similarly, speaker 2440 may output audio to the user of intelligent vehicle 200.

[0066] The wireless communication system 2410 may wirelessly communicate with one or more devices directly or by using a communication network. For example, the wireless communication system 2410 may use 3G cellular communications such as CDMA, EVDO, GSM / GPRS, 4G cellular communications such as LTE, or 5G cellular communications. The wireless communication system 2410 may communicate with a wireless local area network (WLAN) by using Wi-Fi. In some embodiments, the wireless communication system 2410 may communicate directly with devices by using an infrared link, Bluetooth, or ZigBee. The devices may include, but are not limited to, vehicles and / or roadside stations.

[0067] Additionally, in embodiments of the present application, multiple intelligent vehicles can communicate with each other by using V2X, and therefore, the wireless communication system 2410 may further include one or more DSRC devices and one or more LTE-V2X devices.

[0068] Power supply 260 may provide power to various components of intelligent vehicle 200. In one embodiment, power supply 260 may include one or more battery packs. The batteries in the battery packs may be rechargeable lithium-ion batteries or lead-acid batteries. It will be appreciated that in some embodiments, power supply 260 and energy source 2120 may be implemented together.

[0069] Some or all of the functions of intelligent vehicle 200 are controlled by computer system 250. Computer system 250 may include one or more processors 2520. Processor 2520 executes instructions 25110. Instructions 25110 are stored on a non-transitory computer-readable medium, such as memory 2510. Computer system 250 may alternatively be multiple computing devices providing distributed control of individual components or subsystems of intelligent vehicle 200.

[0070] The processor 2520 may be any conventional processor, such as a commercially available CPU. Optionally, the processor may be a dedicated device, such as an Application-Specific Integrated Circuit (ASIC) or other hardware-based processor. While FIG. 2 functionally illustrates components such as a processor, memory, and computer, those skilled in the art should understand that a processor, computer, or memory may actually include multiple processors, computers, or memories, which may or may not be housed in the same physical housing. For example, the memory may be a hard disk drive or another storage medium located in a different housing than the computer. As such, it will be understood that reference to a processor or computer may include reference to a set of processors, computers, or memories, which may or may not operate in parallel. Unlike using a single processor to perform the steps described herein, some components, such as the steering component and the deceleration component, may include their own processors. The processors perform only calculations related to the component's specific function.

[0071] In various aspects described herein, the processor may be located remotely from the vehicle and in wireless communication with the vehicle. In other aspects, some processes described herein are performed by a processor located within the vehicle, while other processes, including the execution of steps necessary for a single operation, are performed by a remote processor.

[0072] In some embodiments, memory 2510 may include instructions 25110 (e.g., program logic) that may be executed by processor 2520 to perform various functions, including the above-described functions, of intelligent vehicle 200. Memory 2510 may also include additional instructions, including instructions used to transmit data to, receive data from, interact with, and / or control one or more of mobility system 210, sensor system 220, control system 230, and peripheral devices 240.

[0073] In addition to storing instructions 25110, memory 2510 may further store data such as road maps, route information, vehicle data such as vehicle position, direction and speed, and other relevant information. It will be appreciated that in one embodiment, computer system 250 of intelligent vehicle 200 may use the data to perform relevant operations when intelligent vehicle 200 is in autonomous, semi-autonomous, and / or manual mode. For example, the current speed of the intelligent vehicle may be adjusted based on the road information for the target road segment and the received target vehicle speed range so that the intelligent vehicle can travel at a constant speed while following other vehicles.

[0074] User interface 270 is configured to provide information to or receive information from a user of intelligent vehicle 200. Optionally, user interface 270 may include interfaces required for one or more input / output devices of peripherals 240, such as a USB interface, an AUX interface, or an OBD interface.

[0075] Computer system 250 may control functions of intelligent vehicle 200 based on data from various subsystems (e.g., mobility system 210, sensor system 220, and control system 230) and data received from user interface 270. For example, computer system 250 may control steering system 2310 to avoid obstacles detected by sensor system 220 and obstacle avoidance system 2360.

[0076] Optionally, the above components may be mounted as subsystems within intelligent vehicle 200. Alternatively, one or more of the components may be mounted separately from intelligent vehicle 200. For example, memory 2510 may be partially or completely separate from intelligent vehicle 200. The above components may be coupled via wires and / or wirelessly.

[0077] It should be noted that the above modules and components within the modules can be added, replaced or deleted based on actual requirements, which is not limited in this application.

[0078] An intelligent driving vehicle traveling on a road, such as intelligent vehicle 200 of FIG. 2, may recognize its surrounding environment in order to adjust its current speed. The surrounding environment may include, but is not limited to, other vehicles, pedestrians, traffic control devices, other infrastructure, and other types of objects. In some embodiments, the intelligent driving vehicle may consider each recognized object independently and determine the speed that the autonomous vehicle needs to adjust based on the object's characteristics, such as its speed, acceleration, or relative distance from the vehicle.

[0079] Optionally, intelligent vehicle 200 or a computing device associated with intelligent vehicle 200 (e.g., computer system 250, computer vision system 2340, and data storage device 2510 of FIG. 2 ) may predict the behavior of a recognized object based on the recognized object's characteristics and the state of the surrounding environment (e.g., traffic, rain, ice on the road). It will be understood that all recognized objects are related. Therefore, by analyzing the state of all other objects in the surrounding environment, the behavior of a single object may be further predicted. Intelligent vehicle 200 may adjust the speed of intelligent vehicle 200 based on the recognized object's predicted behavior. That is, intelligent vehicle 200 may determine how the vehicle needs to be adjusted (e.g., accelerate, decelerate, or stop) and the stable state to which the vehicle should be adjusted based on the object's predicted behavior. Other factors may also be considered in this process, such as the horizontal position within the roadway on which intelligent vehicle 200 is traveling, the curvature of the road, and the effects of proximity to static and dynamic objects.

[0080] In addition to instructions to adjust the speed of intelligent vehicle 200, the computing device may further provide instructions to change the steering angle of intelligent vehicle 200 so that the autonomous vehicle follows a predetermined track and / or maintains a safe horizontal distance and a safe vertical distance from objects near the autonomous vehicle (e.g., vehicles in adjacent lanes).

[0081] The intelligent vehicle 200 may be a car, truck, motorcycle, bus, boat, recreational vehicle, amusement park vehicle, construction equipment, streetcar, train, etc. This is not limiting in this embodiment of the present application.

[0082] It will be understood that the schematic diagram of the intelligent vehicle structure shown in Figure 2 is merely an example implementation in this embodiment of the present application. The intelligent vehicle in this embodiment of the present application includes, but is not limited to, the structures described above.

[0083] The present application provides an intelligent light switching system for intelligent light switching during night driving. The intelligent light switching system first acquires an image of a scenario in front of the system, then acquires ambient light information, classifies light sources in the image, and finally determines whether to switch to high / low beams based on the ambient light information and light source classification information. The units of the intelligent light switching system may be divided in multiple ways, which are not limited in the present application. Figure 3 is an example of a division method. As shown in Figure 3, the functions of each functional unit are described below.

[0084] The intelligent light switching system 300 includes an acquisition unit 310, an ambient light detection unit 320, a light source classification unit 330, and a switching unit 340. The acquisition unit 310 is configured to acquire images, which are captured by a video camera disposed at a fixed position on the vehicle, and light source information is recorded in the images. The ambient light detection unit 320 is configured to calculate an ambient light brightness value corresponding to the image. The light source classification unit 330 is configured to classify the light sources included in the image based on the light source information to obtain a classification result. The switching unit 340 is configured to switch to high beam or low beam based on the ambient light brightness value corresponding to the image and the classification result.

[0085] From the above, it can be seen that light source classification is a key step in the solution of the present application and is implemented by the light source classification unit 330. Specifically, the light source classification unit 330 includes a light source detection model, which performs detection and light source classification on the captured image.

[0086] It should be understood that the light source detection model may be an AI model, and that before using the AI ​​model to perform detection, an initial AI model needs to be trained. In this application, the initial AI model is trained by using sample images that contain light source information and are obtained by a video camera, so that the trained AI model has a light source classification function and can perform light source classification on images obtained by the video camera.

[0087] In addition, the light source detection model in this application can further identify the location of the light source (detection box information / target location in the image) and detect the actual distance between the light source and the ego-vehicle.

[0088] In the training process, special training data needs to be used for training. When the analysis is based on the model capability requirements, sample images with annotation information taken by a video camera need to be used for training. The sample images record the target, and the annotation information includes the target's category information in the sample image, position information (the detection box position / location of the target in the image), and distance information (the distance between the target and the vehicle).

[0089] The target category information is used to indicate the category of the target. In this embodiment of the present application, the target is a light source, such as a vehicle headlight, a vehicle taillight, a street lamp, or a traffic light. The target category information is the category information of the light source.

[0090] In addition, detection boxes need to be used during classification. Detection boxes are used to mark targets in sample images. In this embodiment of the present application, the detection boxes are lamp boxes, and the detection box information includes, but is not limited to, target category information and pixel coordinate information. For example, in this embodiment of the present application, rectangular detection boxes are used for annotation. The detection box information includes category information and pixel coordinate information of the annotated target. The category information includes features such as the shape and color of the detection box. The pixel coordinate information includes four pixel coordinates, namely, the top-left horizontal coordinate, the top-left vertical coordinate, the bottom-right horizontal coordinate, and the bottom-right vertical coordinate of the detection box.

[0091] In one embodiment of the present application, the detection box information may directly display text to indicate the category information of the target, or may indicate the category of the target by using features such as the shape and color of the detection box. The annotation information may be stored as a file in a format such as extensible markup language (XML), JavaScript object notation (JSON), etc.

[0092] The light source detection model is described in detail below.

[0093] In this application, the training data (sample images) required to train the light source detection model are acquired by a video camera at a fixed position in the vehicle. A target detection algorithm is used to perform detection on the sample images, and the light source category information and detection box information recorded in the sample images can be obtained. Alternatively, the above information can be obtained by manual labeling. It should be understood that the sample images can include images acquired at different times and with different exposure times.

[0094] After acquiring a plurality of sample images and obtaining annotation information for the plurality of sample images, the plurality of sample images with annotation information form a training set, and training is performed using the training samples in the training set. First, an initial lamp light source detection model is determined. From the above, it can be seen that the initial lamp light source detection model of the present application can be an AI model, specifically a deep neural network model. This model can not only detect the type and position of the light source in the image, but also calculate the actual distance between the light source and the host vehicle. This means that the initial light source detection model of the present application is structurally improved.

[0095] 4, the structure of the initial illuminant detection model 400 in this application mainly includes three parts: a backbone network 410, a detection network 420, and a loss function calculation unit 430. The backbone network 410 is configured to extract features from the input sample image, and includes a visual geometry group (VGG) network, a residual network, and a loss function calculation unit 430. The backbone network 410 includes several convolution layers, which may be a convolution network, a dense convolution network, etc. The detection network 420 is configured to detect and recognize the features extracted by the backbone network 410, and output light source category information and light source position information (i.e., detection box information). The detection network 420 is also basically formed by several convolution layers, and therefore further performs convolution calculation on the output result of the backbone network 410.

[0096] As mentioned above, the light source detection model of the present application has a different structure compared to common detection models (e.g., yolo and faster RCNN). Because there is a proportional relationship between the distance between vehicle lights of the same vehicle in an image, the vehicle's actual width, the focal length of the video camera used to capture the image, and the distance between the vehicle and the ego vehicle, if three of the four values ​​are known, the remaining one can be calculated based on the proportional relationship. Therefore, the backbone network 410 may use the same type of network. However, in the detection network 420 of the present application, multiple channels are added to each convolutional layer that plays the role of regressing the detection box. Two channels are preferentially added to indicate the horizontal and vertical coordinates of the light source in the sample image. The distance between the vehicle lights of vehicles in the image can be obtained by using the horizontal and vertical coordinates of the light source of the same vehicle to obtain the distance between the vehicle and the ego vehicle. Of course, more channels may be added. Each channel is assigned a corresponding physical meaning, which is not a limitation of the present application.

[0097] First, the parameters of the initial illuminant detection model 400 are initialized. Next, a sample image is input to the initial illuminant detection model 400. The backbone network 410 extracts the features of targets recorded in the sample image to obtain abstract features, and then inputs the abstract features to the detection network 420. The detection network performs further detection and recognition, predicts the type and location of the target and the distance from the target to the host vehicle, and outputs the prediction results to the loss function calculation unit 430 via corresponding channels. Then, annotation information corresponding to the sample image is also input to the loss function calculation unit 430. The loss function calculation unit 430 compares the prediction results obtained by the detection network 420 with the annotation information corresponding to the sample image to calculate a loss function, and updates and adjusts the parameters in the model by using the loss function as a target function and a backpropagation algorithm. Sample images carrying annotation information are input sequentially, and the above training process is continuously and iteratively performed until the value of the loss function converges, i.e., until the value of the loss function calculated each time fluctuates around a specific value. In this case, the light source classification model is fully trained, i.e., it has the type and location of the target in the image and the distance from the target to the ego-vehicle, and can be used for light source classification.

[0098] In this application, two channels are added to each convolutional layer responsible for regressing the detection box, so the construction of the loss function needs to be redesigned. The target positioning model in this application is improved based on a classical target detection model (e.g., yolo or faster RCNN), and the loss function of the classical target detection model is assumed to be Loss1. In this case, the loss function Loss used to construct the target positioning model in this application may be expressed as Loss = Loss1 + Loss2, where Loss2 is the loss function corresponding to the two newly added channels.

[0099] As mentioned above, the extraction of light source features in an image is mainly performed by the backbone network 410 and the detection network 420. The method for extracting features for classification is specifically described below with reference to Fig. 5. As shown in Fig. 5, the method includes, but is not limited to, the following steps:

[0100] S501: Select a bright point from the image whose brightness value is greater than a preset threshold, and set a lamp box for the bright point.

[0101] Specifically, a bright spot whose brightness value is equal to or greater than a preset value is selected as a first bright spot, and a lamp box is set for the first bright spot, the size of which can be adaptively changed based on the size of the first bright spot to be marked.

[0102] For example, as shown in FIG. 6, which is a schematic diagram of setting up a lamp box for a first bright spot, the distribution of bright spots with different characteristics on an image can be clearly displayed by recognizing elements such as the position, brightness, size, and color of the first bright spots on the image, and marking the first bright spots located at the bottom of the image with a red frame, the first bright spots located at the top of the image with specific distribution characteristics with a blue frame, and the first bright spots located at the top of the image with unique color characteristics with a purple frame.

[0103] The specific distribution feature means that the connecting line of the multiple first bright points on the left side of the upper region of the image and the connecting line of the multiple first bright points on the right side of the upper region of the image intersect with a specific point in the center of the image within a specific error range.

[0104] It should be noted that the above lamp box setting method is merely an example in this application. Specific parameters such as the size, shape, and color of the lamp box can be set by researchers and developers based on actual requirements and experimental data. For example, during lamp box setting, lamp boxes of different colors and shapes can be used to mark bright spots with different characteristics. This is not limited in this application.

[0105] S502: Filtering and deduplication are performed on the lamp boxes.

[0106] After the lamp boxes are set, there may be intersections or tangents between the lamp boxes, which may affect subsequent operations. Therefore, it is necessary to filter and remove overlaps from the lamp boxes. There are several methods for filtering and removing overlaps from the lamp boxes. For example, in one embodiment of the present application, the lamp boxes are removed overlaps by using an IOU algorithm. As shown in FIG. 7, a bright point A is marked by a lamp box C, and a bright point B is marked by a lamp box D. If an overlapping area E is detected between the lamp boxes C and D, a second lamp box is generated. The second lamp box is larger than the original lamp box and may contain both bright points. Lamp box F in FIG. 7 is the second lamp box. It should be understood that the size of the second lamp box may be determined based on the size of the original lamp box. In other cases, specific parameters of the second lamp box, such as the size, shape, and color, may also be preset.

[0107] S503: Pairing is performed on the lamp boxes obtained after the overlap removal to obtain a first pair of lamp boxes.

[0108] In one embodiment of the present application, any two lamp boxes in an image can be successfully paired only if at least three conditions are met. The first condition is that the absolute value of the difference in height between the two lamp boxes is less than a first threshold. The second condition is that the distance between the two lamp boxes has a linear relationship with the height of the lamp boxes, i.e., the ratio of the distance between the two lamp boxes to the height of the lamp boxes is within a ratio range. The third condition is that the angle between the connecting line between the center points of the two lamp boxes and the horizontal line is less than a second threshold. It should be understood that in the second condition, the distance between the two lamp boxes is the distance between the center points of the lamp boxes (i.e., the length of the connecting line between the center points of the two lamp boxes), and the height of the lamp boxes is the average height of the two lamp boxes.

[0109] 8 is a schematic diagram of lamp box pairing. In the figure, lamp box A and lamp box B are two lamp boxes selected from an image, the height of lamp box A is a, the height of lamp box B is b, and the distance between the center points of the lamp boxes is c. If |ac| is less than a first threshold,

[0110]

number

[0111] It is understood that the first threshold, the second threshold and the ratio section are set by researchers and developers based on actual requirements and experimental data, which is not limited in the present application.

[0112] S504: The distance between the distant vehicle lamp and the host vehicle is determined.

[0113] From the above, it can be understood that the central region is a region where vehicle lamps of a distant vehicle may appear. A first pair of lamp boxes within the central region is selected. The ratio of the distance between the centers of the first bright points in the first pair of lamp boxes to the actual vehicle width is a first ratio. The ratio of the focal length of the video camera used to obtain the image to the distance from the vehicle to which the first pair of lamp boxes belongs to the host vehicle is a second ratio. It can be understood that the first ratio is equal to the second ratio within a certain error range. In one embodiment of the present application, the two ratios can be considered to be approximately equal to each other. Therefore, the distance from the vehicle to which the first pair of lamp boxes belongs to the host vehicle can be determined by knowing the distance between the centers of the first bright points in the first lamp boxes, the actual vehicle width, and the focal length of the video camera. Alternatively, a value is introduced as an error value. If the distance between the center of the first bright points in the first pair of lamp boxes and the actual vehicle width is known, the first ratio can be calculated. The error value is added to or subtracted from the first ratio to obtain the second ratio. Since the focal length of the video camera is known, the distance from the vehicle to which the first lamp box pair belongs to the ego vehicle can be obtained.

[0114] It should be noted that the above method is merely an example method in this application, and there may be other methods for determining the distance between the lamp of the distant vehicle and the own vehicle, which is not limited in this application.

[0115] It will be understood that certain parameters such as the shape and size of the central region may be preset by researchers and developers based on experimental data, and this is not a limitation of the present application.

[0116] In addition, in another embodiment of the present application, in order to obtain a more accurate light source classification result, the distance between all vehicles in the obtained image and the own vehicle can be calculated according to the above method. In this way, not only the light source information but also the light source distance information can be obtained, and more accurate switching can be performed by referring to the illumination distance of the high beam and low beam.

[0117] S505: The vanishing point horizon (VP line) is identified.

[0118] In one embodiment of the present application, the obtained image is divided into N rows (N is a positive integer), the row in which the center point of the two first bright points of the first lamp box pair is located is identified, the number of first lamp box pairs in each row of the obtained image is counted, and the number in the row is used as the horizontal coordinate and the number of first lamp box pairs is used as the vertical coordinate to draw an in-row quantity distribution map. The in-row quantity distribution map may take the form of a smooth scatter plot, a line scatter plot, a histogram, etc. When the in-row quantity distribution map is a smooth scatter plot, the trough with the largest horizontal coordinate among all troughs in the in-row quantity distribution map is selected, and the in-row quantity corresponding to that trough is used as the first row quantity, or the central position with the smallest interval on the left half side of the peak with the largest horizontal coordinate among all peaks in the in-row quantity distribution map is selected, and the in-row quantity corresponding to that position is used as the second row quantity.

[0119] FIG. 9 is a schematic diagram illustrating how to plot a row quantity distribution map. The obtained image is divided into seven rows. When the image width is 280 pixels, the height of each row is 40 pixels. The row in which the center point of the first luminous point of each first lamp box pair is located is identified, and the number of first lamp box pairs in each row in the obtained image is counted. In FIG. 9, the number of lamp pairs in rows 1, 2, and 4 is 1, the number of lamp pairs in row 3 is 5, the number of lamp pairs in row 5 is 3, and the number of lamp pairs in rows 6 and 7 is 0. The row quantity distribution map is plotted by using the row quantity as the horizontal coordinate and the number of first lamp box pairs as the vertical coordinate. The row quantity distribution map is a histogram.

[0120] 10 is a schematic diagram of determining the quantity of the first row. The position marked with a black dot is the trough with the largest horizontal coordinate in the in-row quantity distribution map. That is, the in-row quantity corresponding to the position marked with a black dot is the quantity of the first row. It will be understood that the method of determining the quantity of the second row is the same as the method of determining the quantity of the first row. The details will not be described again here.

[0121] If the centers of the two first light points of a pair of first lamp boxes are in different rows, 0.5 is added to each row where the centers of the two first light points are located during the subsequent counting of the number of pairs of first lamp boxes in each row of the image. It should be understood that the above process can alternatively be performed according to another statistical method. For example, 1 is added to each row where the centers of the two first light points are located, or 1 is added to either row where the centers of the two first light points are located. This is not a limitation of the present application.

[0122] In another embodiment of the present application, there is also another method for drawing the in-row quantity distribution map, which may be to first identify the row in which the center point of the first bright point in the image and marked on the lamp box is located, count the number of first bright points in each row in the image, and draw the in-row quantity distribution map by using the in-row quantity as the horizontal coordinate and the number of first bright points as the vertical coordinate.

[0123] If the intra-row quantity distribution map does not contain peaks and / or troughs, the default vanishing point horizon is selected as the first vanishing point horizon (VP-line I). If the absolute value of the difference between the quantity of the first row and the default vanishing point horizon is greater than a third threshold, or if the absolute value of the difference between the quantity of the second row and the default vanishing point horizon is greater than a third threshold, the default vanishing point horizon is selected as the first vanishing point horizon (VP-line I). Because vehicle lamps are usually located at the center of the vehicle body, to obtain the second vanishing point horizon (VP-line II), VP-line I must be corrected based on a preset correction value, i.e., a correction value must be added to the vanishing point horizon. In addition, when the vehicle brakes, accelerates, or travels uphill or downhill, the pitch angle of the video camera changes, which can result in significant changes in the position of the light source in the image. For this reason, a reference value is introduced. The third vanishing point horizon (VP-line III) is obtained by multiplying the absolute value of the difference between VP-line I and the reference value by an attenuation coefficient α and adding the resultant result to the reference value. In this case, the obtained VP line III is a VP line that can be used to classify bright spots.

[0124] It is understood that the correction value is a relatively small positive integer, and the specific value of the correction value is set by researchers and developers based on experimental data and practical requirements, which is not limited in the present application.

[0125] It is understood that the third threshold value is set by researchers and developers based on actual requirements and experimental data, which is not limited in this application.

[0126] It should be noted that in another embodiment of the present application, the reference value and damping coefficient values ​​may be adaptively adjusted based on the particular driving environment, for example, braking, accelerating, driving uphill or driving downhill.

[0127] In addition, the above-mentioned reference value and the method for introducing the attenuation coefficient are only exemplary methods in this application. In a specific implementation process, the method can be adjusted or modified based on actual requirements to achieve better beneficial effects, which is not limited in this application.

[0128] S506: Divide the image into regions by using VP lines III.

[0129] After obtaining the VP line III, the obtained image is divided into a first region and a second region by using the row in the obtained image corresponding to the intra-row quantity of the VP line III as a dividing line, where the first region is the image region above the dividing line and the second region is the image region below the dividing line.

[0130] S507: Removes interference from light reflected by the road surface.

[0131] Because the distance between the light source and the road surface in the first region is relatively large, the reflected light generated between the light source and the road surface is relatively dark, and it is highly likely that the condition "brightness is equal to or greater than the preset threshold" is not met. However, because the distance between the light source and the road surface in the second region is relatively small, it is highly likely that the condition "brightness is equal to or greater than the preset threshold" is met. Therefore, the light emitted by the first bright spot in the second region and reflected by the road surface is canceled.

[0132] A first bright point in the second region is selected. A separate position analysis is performed on the first bright point to detect whether there is another first bright point in the region vertically below the first bright point, with a left-right offset equal to or less than a preset distance. If there is another first bright point in the region, the other first bright point is removed to obtain a pair of a second bright point and a second lamp box. It should be understood that the pair of first lamp boxes remaining after the filtering operation is the second lamp box pair. In addition, the preset distance is set by researchers and developers based on actual requirements and experimental data, which is not a limitation of the present application.

[0133] FIG. 11 is a schematic diagram of removing the interference of light reflected by the road surface. In the figure, a position analysis is performed on a bright point A, and the preset distance between the left offset and the right offset is the distance from point M to point O (or the distance from point N to point O). It should be understood that the distance from point M to point O is equal to the distance from point N to point O. There are two solutions to remove the interference of light reflected by the road surface. The first solution is to detect whether there is another first bright point in area X. The second solution is to detect whether there is another first bright point within a sector area consisting of area X and area Y. The sector area is an area drawn using point P as the center of a circle and the distance from point P to point M as the radius. The angle of the sector area is (α + β). It should be understood that α = β. When the first solution is used, bright point D is removed because it is within area X. When the second solution is used, bright point E is removed because it is within the sector area.

[0134] S508: Identify a pair of headlights and a pair of taillights.

[0135] A brightness detection is performed on a second bright spot in a second pair of lamp boxes within the second region, and a color analysis is performed on a halo of the second bright spot in the second pair of lamp boxes within the second region. If the brightness of the second bright spot is within the first section and the color of the halo is within the second section, the second pair of lamp boxes is a headlight pair. If the brightness of the second bright spot is within the third section and the color of the halo is within the fourth section, the second pair of lamp boxes is a taillight pair.

[0136] Note that when an image obtained in the RGB color space is converted to an image in the YUV color space, the component V (red component) or the component U (blue component) is primarily considered during color analysis. Therefore, when brightness detection is performed on the second bright spot, the value of the component Y (brightness value) of each pixel of the second bright spot is obtained, thereby obtaining an average brightness value of the second bright spot. When color analysis is performed on the second bright spot, the value of the component V of each pixel of the second bright spot is obtained, thereby obtaining an average value of the component V of the second bright spot. If the average brightness value is within the third interval and the average value of the component V is within the fourth interval, the second lamp box pair is a taillight pair. Of course, color distinction can alternatively be performed by calculating the average value of the component U. This is not a limitation of the present application.

[0137] It will be understood that for images in different color spaces, the methods for performing brightness detection and color analysis may be different. The above are only exemplary methods in this application and should not be considered as limitations on this application.

[0138] It should be noted that only when both of the two luminous points in the second lamp box pair satisfy the above condition can the second lamp box pair be determined to be a headlight pair or a taillight pair. In one embodiment of the present application, an error value is introduced. If the brightness information and color information of the two luminous points in the second lamp box pair satisfy the above condition within the error value range, the second lamp box pair can be considered to be a headlight pair or a taillight pair. It should be understood that the specific value of the error value can be set by researchers and developers based on actual requirements and experimental data, and is not limited by this application.

[0139] It should be understood that the first, second, third and fourth sections are set by researchers and developers based on actual requirements and experimental data, which is not limited in the present application.

[0140] A brightness detection is performed on a second bright spot in the first region, and a color analysis is performed on a halo of the second bright spot in the first region. If the brightness of the second bright spot is within the fifth interval and the color of the halo is within the sixth interval, the second bright spot is a street light.

[0141] S509: Determine the street light.

[0142] The method for determining streetlights is the same as the method for determining headlight pairs and taillight pairs, and the details will not be described again here. Please refer to the above related content about determining headlight pairs and taillight pairs.

[0143] It should be noted that the fifth and sixth sections are set by researchers and developers based on actual requirements and experimental data, which is not limited in the present application.

[0144] S510: Determine the traffic light.

[0145] A bright point in the first region whose brightness is equal to or greater than a fourth threshold is selected as a third bright point. Brightness detection is performed on the third bright point in the first region. Color analysis is performed on the halo of the third bright point in the first region. If the brightness of the third bright point is within the seventh section and the color of the halo is within the eighth section, the third bright point is a red light of the traffic light. If the brightness of the third bright point is within the ninth section and the color of the halo is within the tenth section, the third bright point is a green light of the traffic light. If the brightness of the third bright point is within the eleventh section and the color of the halo is within the twelfth section, the third bright point is a yellow light of the traffic light.

[0146] The method for determining the traffic lights is the same as the method for determining the headlight pairs and the taillight pairs, and the details will not be described again here. Please refer to the above related content about determining the headlight pairs and the taillight pairs.

[0147] It is understood that the seventh, eighth, ninth, tenth, eleventh and twelfth sections are set by researchers and developers based on actual requirements and experimental data, which is not limited in the present application.

[0148] It is understood that the fourth threshold value is set by researchers and developers based on actual requirements and experimental data, which is not limited in this application.

[0149] S511: Correct misclassification.

[0150] Misclassification can occur in the light source classification process. For example, large vehicles such as buses and trucks have two lights on the top and bottom of the front and rear, respectively, so there may be a situation where two lights are in the first region and two lights are in the second region. This needs to be corrected.

[0151] A vertical analysis is performed on a second lamp box pair in the second region to find a lamp box pair that is vertically above the second lamp box pair and whose distance from the second lamp box pair is within interval 13. If the brightness of the lamp box pair is within interval 14 and the halo color of the lamp box pair is within interval 15, the lamp box pair is a headlight of a large vehicle. If the brightness of the lamp box pair is within interval 16 and the halo color is within interval 17, the lamp box pair is a taillight of a large vehicle.

[0152] It is understood that the 13th, 14th, 15th, 16th and 17th sections are set by researchers and developers based on actual requirements and experimental data, which is not limited in the present application.

[0153] The process for performing the above brightness detection and color analysis is the same as the method for determining the headlight pair and the taillight pair, and the details will not be described again here. Please refer to the above related content about determining the headlight pair and the taillight pair.

[0154] Note that S504 (determining the distance between the lamps of distant vehicles and the host vehicle) can be performed after S507 (removing interference from light reflected by the road surface), or after S508 (determining headlight pairs and taillight pairs), S509 (determining street lights), S510 (determining traffic lights), and S511 (correcting misclassification).

[0155] In addition, in one embodiment of the present application, the obtained original image is subjected to the operations of selecting a central region of the image as an ROI region and compressing the image. The central region is an area where headlights of distant vehicles may appear. Specific parameters such as the shape and size of the central region may be preset by researchers based on experimental data. Performing subsequent operations (e.g., calculating brightness weights) after compressing the image may reduce the amount of calculation, but some details in the image may be lost. Therefore, light source classification is performed based on a combination of the compressed image and the central region image of the original image. Specifically, the compressed image is compared with the selected central region image of the original image, and light sources in the central region are classified based on the central region image, and light sources in other regions are classified based on the compressed image.

[0156] The above content describes the light source classification model and its training process. The process of the intelligent light switching method provided in this application is specifically described below. As shown in Figure 12, this method includes, but is not limited to, the following steps:

[0157] S1210: An image is acquired.

[0158] Specifically, the intelligent light switching system uses a video camera module to capture images of the scenario ahead of the vehicle in real time, including but not limited to ambient light information and light source information such as light information of the vehicle ahead, street lights on both sides of the road, and traffic lights.

[0159] It will be appreciated that the video camera module may capture images in real time based on a preset time, for example, may capture an image every 1 / 20th of a second.

[0160] In one embodiment of the present application, images obtained in real time by the intelligent light switching system using a video camera module include three frame images, which are a short-exposure image, a medium-exposure image, and a long-exposure image, respectively, where the short-exposure image has the shortest exposure time, the long-exposure image has the longest exposure time, and the medium-exposure image has an exposure time intermediate between the previous two exposure times.

[0161] It should be understood that the exposure time can be set in multiple ways. For example, the short-term exposure time is set to 10 ms, the medium-term exposure time is set to 20 ms, and the long-term exposure time is set to 30 ms; or the short-term exposure time is set to 5 ms, the medium-term exposure time is set to 10 ms, and the long-term exposure time is set to 20 ms. Therefore, the above short-term exposure, medium-term exposure, and long-term exposure are only relative. The specific exposure times are set by researchers and developers based on actual requirements and experimental data and are not limited here.

[0162] Note that the image characteristics exhibited by a light source change with exposure time. Using three frames of images for analysis helps obtain more light source information. In other embodiments of the present application, alternatively, more or fewer images may be acquired for subsequent analysis.

[0163] In one embodiment of the present application, the image acquired by the intelligent light switching system is an image in RGB color space, which needs to be converted to an image in YUV color space.

[0164] Note that the above conversion from the RGB color space to the YUV color space is merely an example of the present application. In actual cases, other conversion methods, such as conversion to the HSI color space, may be used.

[0165] S1220: Calculate the ambient light brightness value corresponding to the image.

[0166] Specifically, the intelligent light switching system selects at least one ROI region from the acquired image, obtains the brightness value of each pixel in the selected region, calculates an average brightness value, and then calculates a brightness weighted value based on a specific weight (wherein the brightness weighted value is the brightness value of the ambient light). It should be understood that the specific parameters, such as the size and shape of the selected ROI region, are preset by researchers based on the image, actual requirements, and experimental data, and this is not limited in the present application. It should be understood that the weights are set by researchers based on experimental data, and this is not limited in the present application.

[0167] In one embodiment of the present application, the central region and the entire image can be selected as the ROI region. The central region is the region where the headlights of a distant vehicle may appear. As mentioned above, specific parameters such as the shape and size of the central region can be preset by researchers based on experimental data.

[0168] In one embodiment of the present application, the central region, the upper region, and the entire image can be selected as the ROI region. The central region is a region where vehicle lights of distant vehicles may appear. The upper region is a region where street lights may appear. As mentioned above, specific parameters such as the shape and size of the central region and the upper region can be preset by researchers based on experimental data.

[0169] Note that for images in the YUV color space, the component Y represents brightness information, so obtaining the value of the component Y of a pixel of an image can be considered as obtaining the brightness value of that pixel.

[0170] For example, the central region and the entire image are selected as the ROI region. Here, the average brightness value of the central region is calculated as 200, and the average brightness value of the entire image is calculated as 120. If the specific weights of the central region and the entire image are 3 and 7, respectively, the brightness weight value is calculated as 200×0.3+120×0.7=144. Therefore, the brightness weight value obtained by the calculation is 144.

[0171] S1230: Classify the light sources included in the image.

[0172] Specifically, an image is input to an illuminant detection model, and an illuminant category in the image is obtained based on the illuminant detection model.

[0173] Note that there is no clear time relationship between the ambient light brightness calculation and the light source classification. In practical application, the ambient light brightness calculation may be performed first, the light source classification may be performed first, or the two may be performed simultaneously.

[0174] S1240: Determine whether to switch to high beam or low beam based on the brightness value of the ambient light corresponding to the image and the classification result.

[0175] Specifically, if the brightness value of the ambient light corresponding to the image is equal to or greater than the sixth threshold and the light source classification result indicates that a vehicle lamp is indeed present within the high beam illumination distance, the brightness value of the ambient light may be considered not to be affected by an interfering light source. In this case, a decision is made to switch to low beam. If the brightness value of the ambient light corresponding to the image is equal to or greater than the sixth threshold and the light source classification result indicates that a vehicle lamp is not present within the high beam illumination distance, the brightness value of the ambient light may be considered to be affected by an interfering light source. In this case, a decision is made to switch to high beam. If the brightness value of the ambient light corresponding to the image is less than the sixth threshold and the light source classification result indicates that a vehicle lamp is not present within the high beam illumination distance, a decision is made to switch to high beam. If the brightness value of the ambient light corresponding to the image is less than the sixth threshold and the light source classification result indicates that a vehicle lamp is present within the high beam illumination distance, a decision is made to switch to low beam.

[0176] It should be understood that the high beam illumination distance can be set by researchers and developers based on actual requirements and vehicle lamp parameters. This is not a limitation of the present application. In addition, the video camera captures images in real time, and the intelligent light switching system also calculates the brightness of ambient light and classifies the light source in real time, so light source information and corresponding distance information can be obtained in real time. In one embodiment of the present application, a time-dependent switching mode can be set. If the light source classification result indicates that the difference between the illumination distance of the high beam of the host vehicle and the distance between the light source and the host vehicle is within a preset range, it is determined to switch to the high beam with a delay. It should be understood that the preset range and the delay time can be set by researchers and developers based on actual requirements. This is not a limitation of the present application.

[0177] The intelligent light switching system can be manually or automatically enabled. The manual activation method is applicable to the switching method of S1240. The intelligent light switching system is triggered when the driver determines that the lamp needs to be turned on. When the automatic activation method is used, the brightness of the ambient light can be calculated first. If the brightness is equal to or greater than the seventh threshold, it is deemed that the lamp does not need to be turned on in the environment in which the host vehicle is currently located, and no further light source classification is required. Otherwise, light source classification is performed to determine whether to switch to high beam or low beam, thereby reducing energy consumption. Alternatively, a module for calculating the brightness of the ambient light can be additionally provided. If the brightness of the ambient light obtained by the module is equal to or greater than the seventh threshold, it may be deemed that the intelligent light switching system does not need to be enabled. Otherwise, the intelligent light switching system is triggered. In this case, the intelligent light switching system can switch to high beam or low beam according to the method described in S1210 to S1240. It should be understood that the seventh threshold is set by researchers and developers based on actual requirements and experimental data. This is not a limitation of the present application.

[0178] The method in the embodiment of the present application has been described in detail above. In order to facilitate better implementation of the above solution in the embodiment of the present application, the following further provides a related apparatus for cooperative implementation accordingly.

[0179]

[0033] Figure 3 is a schematic diagram of the structure of an intelligent light switching system according to the present application. The intelligent light switching system is configured to perform the intelligent light switching method of Figure 12. The division of functional units in the intelligent light switching system is not limited herein, and units in the intelligent light switching system may be added, deleted, or combined as needed. In addition, the operations and / or functions of the units in the intelligent light switching system are each used to perform the corresponding steps of the method described in Figure 12. For the sake of brevity, the details will not be described again here. Figure 3 shows an example of a division of functional units.

[0180] The intelligent light switching system 300 includes an acquisition unit 310 , an ambient light detection unit 320 , a light source classification unit 330 and a switching unit 340 .

[0181] Specifically, the obtaining unit 310 is configured to perform the above step S1210, and optionally perform any method of the above steps.

[0182] Ambient light detection unit 320 is configured to perform step S1220 above, and optionally any method of the steps above.

[0183] The light source classification unit 330 is configured to perform the above steps S501 to S511 and S1230, and optionally perform any method of the above steps.

[0184] The switching unit 340 is configured to perform step S1240 above, and optionally any method of the steps above.

[0185] The four units may communicate data with each other via a communication channel. It should be understood that the units included in the intelligent light switching system 300 may be software units or hardware units, or some units may be software units and some units may be hardware units.

[0186] 13 is a schematic diagram of the structure of a computer device according to an embodiment of the present application. As shown in FIG. 13, the computer device 1300 includes a processor 1310, a communication interface 1320, and a memory 1330. The processor 1310, the communication interface 1320, and the memory 1330 are connected to each other using an internal bus 1340.

[0187] The computing device 1300 may be the intelligent light switching system 300 of Figure 3. The functions performed by the intelligent light switching system 300 of Figure 3 are actually performed by a processor 1310 of the intelligent light switching system 300.

[0188] The processor 1310 may include one or more independent processors, such as a central processing unit (CPU), or a combination of a CPU and a hardware chip. The hardware chip may be an application-specific integrated circuit (ASIC), a programmable logic device (PLD), or a combination thereof. The PLD may be a complex programmable logic device (CPLD), a field-programmable gate array (FPGA), a generic array logic (GAL), or any combination thereof.

[0189] The communication interface 1320 is an interface for connecting an Ethernet, a wireless access network, The network is configured to communicate with another device or communication network, such as a Radio Access Network (RAN), a Core Network, or a Wireless Local Area Network (WLAN).

[0190] The bus 1340 may be a peripheral component interconnect (PCI) bus, an extended industry standard architecture (EISA) bus, etc. The bus 1340 may be classified as an address bus, a data bus, a control bus, etc. For clarity, only one thick line is used to represent the bus in Figure 13, but this does not mean that there is only one bus or only one type of bus.

[0191] The memory 1330 may be a volatile memory, such as a random access memory. The memory 1330 may alternatively include non-volatile memory (RAM). memory, such as read-only memory (ROM), flash memory, hard disk drive (HDD) or solid state drive (solid The memory 1330 may include a storage device (e.g., a hard disk drive, a solid-state drive, a solid-state drive (SSD)). The memory 1330 may alternatively include a combination of the above types of memory. The memory 1330 is configured to store program code for performing the above-described embodiments of the intelligent light switching method. In one implementation, the memory 1330 also buffers other data, and the processor 1310 controls the execution of the data to perform the functional units of the intelligent light switching system 300 or the method steps performed by the intelligent light switching system 300 in the method embodiment shown in FIG. 12 .

[0192] An embodiment of the present application further provides a computer-readable storage medium, which stores a computer program, which, when executed by a processor, can implement some or all of the steps recorded in any one of the above method embodiments and the functions of any functional units shown in FIG.

[0193] An embodiment of the present application further provides a computer program product, which, when executed on a computer or processor, causes the computer or processor to perform one or more of the method steps performed by the intelligent light switching system 300 in any one of the above methods. When the above modules in the apparatus are implemented in the form of software functional units and sold or used as independent products, the modules may be stored in a computer-readable storage medium.

[0194] In each of the above embodiments, the description of the embodiment focuses on each, and for the parts not described in detail in the embodiment, please refer to the related descriptions of other embodiments.

[0195] It should be understood that the terms "first," "second," "third," "fourth," and various numbers used herein are merely used for distinction purposes to facilitate explanation and are not to be construed as limiting the scope of the present application.

[0196] It should be understood that the term "and / or" herein describes only an associative relationship between related objects, and that three relationships may exist. For example, A and / or B may represent three cases: when only A is present, when both A and B are present, and when only B is present. Additionally, the symbol " / " herein generally indicates an "or" relationship between related objects.

[0197] It should be understood that the sequence numbers of the above processes do not refer to the order of execution in the embodiments of the present application. The order of executing the processes should be determined according to the functions and internal logic of the processes, and should not be construed as a limitation on the implementation process of the embodiments of the present application.

[0198] Those skilled in the art will recognize that, in combination with the examples described in the embodiments disclosed herein, the units and algorithm steps can be implemented by electronic hardware or a combination of computer software and electronic hardware. Whether a function is implemented by hardware or software depends on the specific application and design constraints of the technical solution. Those skilled in the art may use different methods to implement the described functions for each specific application, but such implementation should not be considered to go beyond the scope of the present application.

[0199] Those skilled in the art can clearly understand that for the sake of convenient and simple description, the detailed operation processes of the above systems, devices and units should refer to the corresponding processes in the above method embodiments, and the details will not be described again here.

[0200] In some embodiments provided herein, it should be understood that the disclosed systems, devices, and methods may be implemented in other ways. For example, the described device embodiments are merely examples. For example, the division into units is merely a logical division, and other divisions may be used in actual implementation. For example, multiple units or components may be combined or integrated into other systems, or some functions may be ignored or not performed. In addition, the shown or described mutual couplings or direct couplings or communication connections may be implemented using some interfaces. Indirect couplings or communication connections between devices or units may be implemented in electronic, mechanical, or other forms.

[0201] The units described as separate components may or may not be physically separated, and the components shown as units may or may not be physical units, and may be located in one location or distributed across multiple network units. Some or all of the units may be selected based on actual requirements to achieve the objectives of the solutions of the embodiments.

[0202] Additionally, the functional units in the embodiments of the present application may be integrated into one processing unit, each of the units may exist physically alone, or two or more units may be integrated into one unit.

[0203] When a function is implemented in the form of a software functional unit and sold or used as an independent product, the function may be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present application may essentially or a part that contributes to the prior art or a part of the technical solution may be implemented in the form of a software product. A computer software product is stored in a storage medium and includes some instructions for instructing a computer device (which may be a personal computer, a server or a network device) to perform some or all of the steps of the method described in the embodiments of the present application. The above storage medium may include a USB flash drive, a removable hard disk, a read-only memory (ROM), a random access memory (RRAM), a This includes any medium capable of storing program code, such as memory, RAM, magnetic disk, or optical disk.

[0204] The steps of the methods in the embodiments of the present application may be sequentially adjusted, combined or deleted based on actual requirements.

[0205] The modules of the apparatus in the embodiments of the present application can be combined, divided or deleted based on actual requirements.

[0206] In conclusion, the above embodiments are only intended to illustrate the technical solutions of the present application, and are not intended to limit the present application. Although the present application has been described in detail with reference to the above embodiments, those skilled in the art should understand that the technical solutions described in the above embodiments may be modified or some technical features may be replaced with equivalents without departing from the scope of the technical solutions of the embodiments of the present application.

Claims

1. 1. An intelligent light switching method, the method comprising: acquiring an image, the image being taken by a video camera disposed at a fixed position on the vehicle, the image having light source information recorded therein; calculating an ambient light brightness value corresponding to the image; Classifying light sources included in the image based on the light source information to obtain a classification result, the image being inputted into a light source detection model, and obtaining a light source category in the image based on the light source detection model; switching lights based on the ambient light brightness value corresponding to the image and the classification result; Including, Obtaining a light source category in the image based on the light source detection model, selecting a bright point from the image whose brightness value is greater than a preset threshold and setting a lamp box for the bright point; performing pairing on the lamp boxes to obtain a plurality of pairs of lamp boxes; determining a vanishing point horizon (VP line) based on the plurality of lamp box pairs, the VP line being used to distinguish a first region from a second region of the image; Classifying the bright points according to a positional relationship between the bright points and the VP line and color characteristics of the bright points to obtain different light source categories; A method comprising:

2. Calculating an ambient light brightness value corresponding to the image includes: selecting at least one region from the image and calculating a brightness value for the at least one region; calculating an ambient light brightness value corresponding to the image based on the brightness value of the at least one region; The method of claim 1 , comprising:

3. Prior to inputting the image into a light source detection model, the method comprises: training the light source detection model by using a plurality of sample images, the sample images including the light sources and annotation information for the light sources; The method of claim 1 , comprising:

4. Prior to pairing with the lamp box, the method further comprises: performing overlap elimination on the lamp boxes to ensure there are no overlaps or tangents between the lamp boxes; The method of claim 1 further comprising:

5. determining a vanishing point horizon (VP line) based on the plurality of lamp box pairs, mapping lamp box centers of the lamp box pairs onto rows of the image to obtain a row quantity distribution map of the lamp box pairs; selecting a VP line based on the row quantity distribution map; correcting the VP line based on a preset correction value to obtain a corrected VP line; adjusting the corrected VP line by using a reference value to obtain a VP line used to classify the bright point, the reference value being used to describe a change in pitch angle of the video camera; The method of claim 1 , comprising:

6. 1. An intelligent light switching system, comprising: an acquisition unit configured to acquire an image, the image being captured by a video camera disposed at a fixed position on the vehicle, the image having light source information recorded therein; an ambient light detection unit configured to calculate an ambient light brightness value corresponding to the image; a light source classification unit configured to classify a light source included in the image based on the light source information to obtain a classification result, the light source classification unit inputting the image into a light source detection model, and obtaining a light source category in the image based on the light source detection model; a switching unit configured to switch lights based on an ambient light brightness value corresponding to the image and the classification result; Including, When obtaining a light source category in the image based on the light source detection model, the light source classification unit: selecting a bright point from the image whose brightness value is greater than a preset threshold and setting a lamp box for the bright point; performing pairing on the lamp boxes to obtain a plurality of pairs of lamp boxes; determining a vanishing point horizon (VP line) based on the plurality of lamp box pairs, the VP line being used to distinguish a first region from a second region of the image; Classifying the bright points according to a positional relationship between the bright points and the VP line and color characteristics of the bright points to obtain different light source categories; The system is specifically configured to:

7. The ambient light detection unit includes: selecting at least one region from the image and calculating a brightness value for the at least one region; calculating an ambient light brightness value corresponding to the image based on the brightness value of the at least one region; The system of claim 6 , specifically configured to:

8. Before inputting the image into the illuminant detection model, the illuminant classification unit: training the illuminant detection model using a plurality of sample images, the sample images including the illuminants and labels of the illuminants; The system of claim 6 , further configured to:

9. Before performing pairing with the lamp box, the light source classification unit: performing overlap elimination on the lamp boxes to ensure there are no overlaps or tangents between the lamp boxes; The system of claim 6 , further configured to:

10. When determining a vanishing point horizontal line (VP line) based on the plurality of lamp box pairs, the light source classification unit: mapping lamp box centers of the lamp box pairs onto rows of the image to obtain a row quantity distribution map of the lamp box pairs; selecting a VP line based on the row quantity distribution map; correcting the VP line based on a preset correction value to obtain a corrected VP line; adjusting the corrected VP line by using a reference value to obtain a VP line used to classify the bright point, the reference value being used to describe a change in pitch angle of the video camera; The system of claim 6 , specifically configured to:

11. A computing device comprising a memory and a processor, the processor executing computer instructions stored in the memory enabling the computing device to perform the method of any one of claims 1 to 5.

12. 6. A computer readable storage medium storing a computer program which, when executed by a processor, performs the method according to any one of claims 1 to 5.

13. A vehicle including an intelligent light switching system according to any one of claims 6 to 10.

Citation Information

Patent Citations

  • Control circuit for image array sensor and automatic control of headlight

    JP2001517035A

  • Vehicle light source detection device, light control device and vehicle light source detection program

    JP2014024410A

  • Image processor

    JP2014231301A

  • Out-vehicle environment recognition device

    JP2017097658A

  • Method for determining an object class of an object, from which light is emitted and / or reflected to a vehicle

    US20130079983A1