Self-adaptive vehicle lamp control method and system based on ultrasonic detection assistance and vehicle

Abstract

The invention discloses a self-adaptive vehicle lamp control method and system based on ultrasonic detection assistance and a vehicle, and relates to the technical field of vehicle lamp intelligent adjustment, and the method comprises the following steps: obtaining an external environment image in a vehicle driving process in real time to judge a current driving environment, starting vehicle front ultrasonic detection if the current driving environment is abnormal, and otherwise, starting vehicle front ultrasonic detection if the current driving environment is abnormal; if yes, performing identification according to the external environment image; wherein the obstacle in front of the vehicle is identified according to the ultrasonic detection signal, the position and the distance of the obstacle in front of the vehicle relative to the vehicle are calculated, the signal is input into the classification model based on deep learning, and the category of the obstacle in front of the vehicle is identified; the position, the distance and the category of the obstacle in front of the vehicle are integrated into the self-adaptive front lighting system, and the direction, the intensity and the range of irradiation of the vehicle lamp are automatically adjusted. According to the method, more efficient and more accurate self-adaptive adjustment of the vehicle lamp can be realized in severe weather or environment.

Description

Technical Field

[0001] The present invention relates to the technical field of intelligent headlight adjustment, and particularly to an adaptive headlight control method, system and vehicle based on ultrasonic detection assistance. Background Technique

[0002] The statements in this part only provide background technical information related to the present invention and do not necessarily constitute prior art.

[0003] Vehicle lighting systems need to meet various regulatory requirements for beam patterns and lighting levels to improve the driving performance and safety of vehicles in dark or low visibility conditions. In recent years, with the development of intelligent vehicles, adaptive headlight systems and related technologies that can automatically change the illumination according to changes in the external environment or the vehicle itself have also been proposed. For example, the Adaptive Front Lighting System (AFS) of automobiles, as a lighting device, can automatically change the working mode of the front lighting system according to weather conditions, external light, road conditions, and driving information, etc., adjust the light shape of the irradiated light, provide a wider range and more reliable lighting field of vision for the driver, and ensure the safety of the driver and road pedestrians. Among them, Dynamic Bending Light (DBL) belongs to the core sub-function module of the AFS system. It judges whether the vehicle turns left or right through the steering wheel rotation angle, combines speed parameters, etc., and dynamically adjusts the horizontal irradiation direction or the lateral offset angle of the headlight to achieve synchronous lighting of the curve trajectory; Adaptive Driving Beam (ADB) identifies oncoming vehicles or pedestrians through a camera, and generates a non-glare lighting dark area in real time to achieve intelligent control of the high beam.

[0004] The existing AFS system mainly obtains driving images and external environment information through a camera, uses image recognition algorithms such as YOLO series algorithms for target detection to distinguish pedestrians or vehicles, etc. On this basis, relying on the Ackermann steering model, according to real-time data such as the steering wheel steering sensor, vehicle speed sensor, and body height sensor, it automatically adjusts and optimizes the basic lighting of the headlight. For example, automatically adjusting the irradiation angle of the low beam to eliminate the inner blind area when the vehicle is driving on a curve, or extinguishing the corresponding LED lights according to the identified pedestrian trajectory and vehicle driving trajectory to prevent pedestrians and oncoming vehicle drivers from being dazzled and causing traffic accidents, etc.

[0005] That is, the existing AFS system relies on visual sensors and image recognition algorithms for implementation, and it is necessary to adjust the vehicle lights according to the trajectories of recognized obstacles such as pedestrians and vehicles. However, in harsh weather or environments (such as low visibility environments like rain and fog), the existing AFS system is affected by the external environment, and the image recognition effect is poor. It is unable to quickly and accurately judge obstacles, and thus unable to achieve fast and precise adaptive adjustment of vehicle lights. That is, the response speed of the entire system is limited, and there are problems of adjustment delay and poor accuracy. Summary of the Invention

[0006] To solve the deficiencies of the above-mentioned existing technologies, the present invention provides an adaptive vehicle light control method, system, and vehicle based on ultrasonic detection assistance. On the basis of the existing AFS system, ultrasonic detection technology and a deep learning model are introduced, which can achieve fast and accurate recognition of obstacles in front of the vehicle in harsh environments, provide more dimensional information for the adaptive vehicle light system, and assist the traditional visual recognition method to achieve more efficient and precise adaptive control of vehicle lights under any circumstances, effectively shortening the reaction time of the entire control system, avoiding problems of adjustment delay and poor accuracy, and improving driving safety.

[0007] In the first aspect, the present invention provides an adaptive vehicle light control method based on ultrasonic detection assistance.

[0008] An adaptive vehicle light control method based on ultrasonic detection assistance includes:

[0009] Obtain the external environment image during the vehicle driving process in real time;

[0010] Judge the current driving environment according to the external environment image. If the current driving environment is abnormal, start ultrasonic detection in front of the vehicle, and identify the obstacles in front of the vehicle, their positions, distances, and categories according to the real-time obtained ultrasonic detection signals; otherwise, identify according to the external environment image;

[0011] Among them, to identify the obstacles in front of the vehicle according to the ultrasonic detection signals, calculate the position and distance of the obstacles in front of the vehicle relative to the vehicle, and then input the signals into a classification model based on deep learning to identify the categories of the obstacles in front of the vehicle;

[0012] Integrate the position, distance, and category of the obstacles in front of the vehicle into the adaptive front lighting system, and automatically adjust the irradiation direction, intensity, and range of the vehicle lights.

[0013] In the second aspect, the present invention provides an adaptive vehicle light control system based on ultrasonic detection assistance.

[0014] An adaptive vehicle light control system based on ultrasonic detection assistance includes:

[0015] An environmental information acquisition module for real-time acquisition of external environmental images during vehicle driving;

[0016] An obstacle recognition module for judging the current driving environment according to the external environmental image. When the current driving environment is abnormal, ultrasonic detection in front of the vehicle is started, and obstacles in front of the vehicle, their positions, distances, and categories are recognized according to the real-time obtained ultrasonic detection signals; otherwise, recognition is performed according to the external environmental image;

[0017] Among them, obstacles in front of the vehicle are recognized according to the ultrasonic detection signals, the positions and distances of the obstacles in front of the vehicle relative to the vehicle are calculated, and then the signals are input into a classification model based on deep learning to recognize the categories of the obstacles in front of the vehicle;

[0018] An adaptive headlight control module for integrating the positions, distances, and categories of obstacles in front of the vehicle into the adaptive front lighting system to automatically adjust the direction, intensity, and range of headlight irradiation.

[0019] In a third aspect, the present invention also provides a vehicle that executes the adaptive headlight control method assisted by ultrasonic detection proposed in the first aspect, or includes the adaptive headlight control system assisted by ultrasonic detection proposed in the second aspect.

[0020] The above one or more technical solutions have the following beneficial effects:

[0021] 1. Considering that the existing automotive adaptive front lighting system (i.e., AFS system) highly depends on cameras and image detection algorithms, and there are problems such as difficulty in quickly and effectively recognizing obstacles in harsh environments and thus difficulty in accurately adapting and adjusting, the present invention proposes an adaptive headlight control method, system, and vehicle assisted by ultrasonic detection. Considering the characteristics that ultrasonic detection is less affected by environmental factors such as light, ultrasonic detection technology is introduced as an auxiliary. When facing harsh environments, ultrasonic sensor arrays are used to emit signals and receive the returned signals, and environmental noise in the returned signals is filtered by Fourier transform. Obstacles in front of the vehicle are recognized according to the filtered ultrasonic signals, and the obstacle position, distance, and category data obtained by ultrasonic detection are fused with existing CAN network data such as vehicle speed and steering angle to construct a multimodal control model, and the AFS system is used to achieve more precise adaptive adjustment of the direction, intensity, and range of the headlights. Compared with traditional visual recognition methods, the present invention can effectively improve the recognition speed and accuracy in bad weather or environments by introducing ultrasonic assistance on the basis of visual recognition. The obtained distances and categories of obstacles in front can provide more dimensional information for the adaptive headlight system, making the adaptive algorithm more efficient and accurate, effectively shortening the reaction time of the entire system, and ensuring the accuracy and effectiveness of headlight adaptive adjustment.

[0022] 2. When the present invention identifies obstacles based on ultrasonic signals, it first measures the distance according to the time difference between the emission and reflection of ultrasonic signals to determine whether there are obstacles in front of the vehicle, and obtains the position and distance information of the obstacles with centimeter-level accuracy, avoiding the problem of large errors in traditional visual ranging. Secondly, the present invention also introduces a deep learning algorithm. Considering that the spectral characteristics of sound waves reflected by objects of different materials are significantly different, the ultrasonic signals are used for deep learning and training of the neural network model, so as to quickly and accurately identify the types of obstacles in front, which is convenient for the subsequent application of the AFS system. Compared with the traditional YOLO general image detection algorithm, this identification method of the present invention focuses more on the driving scenario and has a higher recognition accuracy.

[0023] Advantages of additional aspects of the present invention will be given in part in the following description, become apparent in part from the following description, or be learned through the practice of the present invention. BRIEF DESCRIPTION OF THE DRAWINGS

[0024] The accompanying drawings forming a part of this specification are used to provide a further understanding of the present invention. The schematic embodiments and descriptions thereof of the present invention are used to explain the present invention and do not constitute an improper limitation of the present invention.

[0025] Figure 1 is the overall flowchart of the adaptive headlight control method based on ultrasonic detection assistance described in the embodiments of the present invention;

[0026] Figure 2 is the flowchart of the adaptive regulation of the headlight by means of ultrasonic detection in the embodiments of the present invention. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0027] It should be noted that the following detailed description is exemplary only for the purpose of describing specific embodiments, aiming to provide further explanation of the present invention and is not intended to limit the exemplary embodiments of the present invention. Unless otherwise specified, all technical and scientific terms used herein have the same meaning as commonly understood by those of ordinary skill in the technical field to which the present invention belongs. In addition, it should also be understood that when the terms "comprising" and / or "including" are used in this specification, they indicate the presence of features, steps, operations, devices, components, and / or combinations thereof.

[0028] The present invention proposes an adaptive headlight control method assisted by ultrasonic detection. By introducing ultrasonic detection technology and deep learning algorithms, considering the characteristics that ultrasonic detection is less affected by environmental factors such as light, an ultrasonic sensor array is used to emit ultrasonic signals and receive the reflected signals. The environmental noise in the signals is filtered by Fourier transform, and then the time difference between the emission and reflection of the ultrasonic signals is used for ranging to obtain the position and distance information of obstacles with centimeter-level accuracy, avoiding the problem of large errors in traditional visual ranging. At the same time, considering that the reflection spectrum characteristics of sound waves are significantly different when encountering objects of different materials, model learning and training are also carried out using ultrasonic signals, which can quickly and accurately identify the categories of obstacles in front (such as pedestrians, vehicles, potholes, road signs, etc.). Compared with the traditional YOLO general detection algorithm, the above-mentioned recognition model for ultrasonic signals can focus more on the driving scenario and has a higher recognition accuracy. Finally, the obstacle distance and category data obtained by ultrasonic detection are fused with the existing CAN network data such as vehicle speed and steering angle, and the AFS system is used to achieve more accurate adaptive adjustment of the headlight direction, intensity and range.

[0029] Embodiment 1

[0030] This embodiment provides an adaptive headlight control method based on ultrasonic detection, as Figure 1 shown, which specifically includes the following steps:

[0031] Step S1: Real-time obtain the external environment image during the vehicle driving process.

[0032] Step S2: Judge the current driving environment according to the external environment image. If the current driving environment is abnormal, start the ultrasonic detection in front of the vehicle, and identify the obstacles in front of the vehicle, their positions, distances and categories according to the real-time obtained ultrasonic detection signals; otherwise, identify according to the external environment image.

[0033] Specifically, considering that existing vehicles usually have multiple cameras installed around the vehicle to obtain images of the external environment during vehicle driving, for this external environmental impact, the current driving environment can be quickly and accurately judged through existing image recognition algorithms, such as recognizing bad weather or environments like rain and fog. Considering the difficulty of recognizing obstacles in the image in the above-mentioned bad environment, the recognition accuracy and efficiency are low, which will affect the subsequent adaptive adjustment of vehicle lights. Therefore, this embodiment also introduces ultrasonic detection technology. An ultrasonic sensor array is installed at the front vehicle lights (including the left and right vehicle lights). During driving, when it is detected that the current driving environment is abnormal, the ultrasonic sensor array is activated for detection, that is, the ultrasonic sensor array continuously emits ultrasonic signals. If an obstacle is encountered during driving, when the emitted ultrasonic signal encounters an obstacle ahead, the signal will be reflected back and captured by the ultrasonic sensor array to obtain the ultrasonic detection signal of real-time reflection. This ultrasonic detection signal contains the category and distance information of the obstacle ahead, so that the obstacle in front of the vehicle can be accurately and quickly detected.

[0034] Preferably, the sensor array is deployed with symmetrically distributed MEMS ultrasonic sensors. The horizontal spacing is designed to be 1 / 2 of the vehicle width in front, and the vertical installation angle deflects inwards by 15°, thereby forming an intersection detection area to ensure the accuracy of ultrasonic detection.

[0035] Further, during the driving process of the vehicle, the ultrasonic detection in front of the vehicle is started, and the ultrasonic detection signal is obtained in real time. According to this ultrasonic detection signal, the obstacle in front of the vehicle and its corresponding information are recognized. The specific process includes:

[0036] Step S2.1: Send and receive ultrasonic detection signals in real time, and dynamically correct the sound speed value of signal propagation according to the current environmental temperature and humidity. Specifically, the formula for dynamic correction of sound speed is:

[0037] v = 331.4 + 0.6 * T + 0.0124 * RH;

[0038] Among them, the basic sound speed is 331.4 m / s, that is, the ultrasonic speed at 0 °C; the temperature coefficient is 0.6 m / (s·°C), that is, for every 1 °C increase in temperature, the sound speed increases by 0.6 m / s; T is the environmental temperature, in degrees Celsius; RH represents humidity, in percentage. For example, if the measured current environmental temperature T = 25 °C and humidity RH = 50%, substituting into the above formula, the corrected sound speed value can be calculated as:

[0039] v = 331.4 + 0.6 × 25 + 0.0124 × 50 = 347.02 m / s.

[0040] Through the above processing, the current accurate sound speed value can be inferred to ensure the accuracy of subsequent obstacle distance detection.

[0041] Step S2.2: Based on the ultrasonic detection signals sent and received in real time, and based on the time difference ratio of the ultrasonic detection signals sent and received, determine the position of the obstacle in front of the vehicle relative to the vehicle.

[0042] Specifically, since ultrasonic sensor arrays are installed at both the left and right headlight positions on the front side of the vehicle, when there is an obstacle in front of the left front headlight, the ultrasonic sensors at both the left and right front headlights can receive the feedback ultrasonic detection signals, and this signal contains the category and distance information of the object in front. In this embodiment, first, based on the time when the ultrasonic sensor arrays on the left and right sides of the vehicle send and receive signals, determine the position of the obstacle relative to the vehicle, that is, according to the time difference ratio of the left and right sensors: Δt ratio =(t left -t right ) / (t left +t right ), determine whether the obstacle is on the left or right side of the vehicle. For example, for the signals emitted at the same moment, if the ultrasonic array on the left obtains the echo signal first, it means that the obstacle is in the left range of the headlight; if the ultrasonic array on the right obtains the echo signal first, it means that the obstacle is in the right range of the headlight. In this way, the position of the obstacle can be quickly determined. On this basis, the AFS system can determine the corresponding adjustment method for the left and right front headlights according to this position, making the headlight adjustment more accurate.

[0043] Step S2.3: Based on the position of the obstacle relative to the vehicle, select the corresponding ultrasonic detection signal, and combine the corrected sound speed value to calculate the distance of the obstacle in front of the vehicle.

[0044] In this embodiment, according to the position of the obstacle, determine whether to select the echo signal received on the left or right side, and then, based on the time difference between the signal emission and reception, combine the corrected sound speed value to calculate the distance of the obstacle in front of the vehicle. The calculation formula is:

[0045] d=(ct) / 2;

[0046] In the above formula, d represents the distance, c represents the sound speed, and t represents the round-trip time of the ultrasonic signal.

[0047] Step S2.4: Input the received ultrasonic detection signal into the classification model based on deep learning to identify the category of the obstacle in front of the vehicle.

[0048] Preferably, considering that the reflected ultrasonic detection signal (hereinafter referred to as the echo signal) contains a certain amount of environmental noise, in this embodiment, the echo signal is first subjected to Fourier transform (i.e., FFT transform) to convert the time-domain signal into a frequency-domain signal. After effectively filtering the frequency-domain signal, the effective signal is extracted through power spectral density analysis. Through the above preprocessing operations, signal noise can be filtered out and clearer signal features can be extracted. Then, the signal is input into the classification model based on deep learning, and the category of obstacles in front of the vehicle can be accurately identified.

[0049] In this embodiment, a classification model based on deep learning applicable to the driving scenario is pre-constructed, and the training process of this model is as follows:

[0050] Step S2.4.1: Collect a number of ultrasonic detection signals reflected under different conditions, label the obstacle category label of each signal, and use the signals with labeled labels as samples to construct a sample training set. Among them, different conditions include setting different categories of obstacles on the road under different harsh weather conditions such as night, rain, and fog. The obstacle categories include obstacles on the road surface such as pedestrians, vehicles, potholes, and other obstacles such as road signs and road edges. Preferably, the above filtering and other preprocessing operations are performed on each sample signal.

[0051] Step S2.4.2: Construct a neural network model, use the sample training set to perform supervised training on the model, and continuously iterate to learn the potential relationship between the reflected signal features of different obstacles and the obstacle categories until the training is completed, so as to achieve efficient classification and recognition in practical applications.

[0052] After the training is completed, the feedback signal received in real time by the ultrasonic sensor array is input into the self-trained neural network model to identify and output the category of obstacles in front of the vehicle, and at the same time, the current environmental information, such as rain or fog, can be effectively judged.

[0053] In the above manner, the ultrasonic detection signal in the time domain is converted into a spectral signal by Fourier transform, and after preprocessing, it is input into the neural network to complete the identification and classification of obstacles. It should be noted that the specific neural network structure is not limited in this embodiment, and any existing neural network can be used to implement it.

[0054] As an implementation, this embodiment proposes an improved neural network model. Specifically, for ultrasonic signal features, a Physics-Guided Ultrasonic Network (PGU-Net) architecture oriented to ultrasonic perception is designed. The time-series ultrasonic detection signal is converted into a frequency-domain signal through Fourier transform and then input into the model after signal preprocessing. In this model, the cross-attention fusion mechanism of the Transformer is used to extract the features of the frequency-domain signal, and then a convolutional layer is used to perform adaptive classification and output on the feature signal.

[0055] Specifically, first, feature extraction is performed on the ultrasonic detection signal, including:

[0056] (1) Considering that the received signal is a lossy signal, this embodiment uses signal tomography to reconstruct the signal. According to the reflection signal attenuation rate: A = 20 * log10(V receive / V emit ), the received ultrasonic detection signal is reconstructed.

[0057] (2) When an object is detected ahead, the frequency of the signal generator is changed to emit different frequencies to detect the object ahead, so as to make the classification by the neural network model more accurate. Correspondingly, if there is no object ahead, the ultrasonic frequency will not be changed. On this basis, when there is an object ahead, the short-time Fourier transform (STFT) is performed on the received signals of different frequencies to generate a time-frequency spectrogram.

[0058] (3) Mel-scale filtering is performed on the generated time-frequency spectrogram to obtain the final time-frequency spectrogram.

[0059] Secondly, the network structures of lightweight MobileNetV3 and HRNet are used to construct a neural network model, and the extracted signal feature map is input into the model to output the final obstacle classification result.

[0060] In the above model, for the input feature signal, the carrier phase information is extracted through Hilbert transform, and the phase difference of the frequency-domain feature signal arrival is determined, so as to improve the material recognition ability; then, the PDE constraint layer is used to ensure that the network prediction conforms to the acoustic wave propagation law; the dynamic time warping (DTW) algorithm is used for waveform matching, and according to the vehicle speed range (0 - 120 km / h), Doppler effect simulation is added, that is, the Doppler frequency shift amount Δf = (2vcosθ) / λ is added according to the vehicle speed v to inject environmental noise: the noise spectra of pre-collected raindrop / sandstone impact, etc. are mixed, and the confidence fusion P final = αP 超声 +(1 - α)P CANPerform feature fusion, where α represents visibility; the fused features pass through a fully connected layer to output the final classification result.

[0061] Furthermore, the loss function of the above network model is:

[0062] Loss = αCE loss + βTriplet loss + γThysics loss ;

[0063] where α, β, and γ are weighting coefficients; CE loss is the cross-entropy loss and can be expressed as:

[0064]

[0065] In the above formula, x i represents the i-th input sample, y i represents the category of the i-th sample, and σ() is the activation function.

[0066] Triplet loss is the triplet loss and can be expressed as:

[0067] L = max(d(a,p) - d(a,n) + margin, 0);

[0068] In the above formula, the input is a triplet, including an anchor (a) example, a positive (p) example, and a negative (n) example, where p and a are samples of the same category, n and a are samples of different categories, and margin is a constant greater than 0. By optimizing the distance between the anchor example and the positive example to be less than the distance between the anchor example and the negative example, the similarity calculation between samples is achieved.

[0069] γThysics loss is the physical loss, including the physical information loss L PDE and the data error loss L data , and can be expressed as:

[0070] L = L PDE + τL data ;

[0071]

[0072] where the sample x and time t are input into the model, and the output prediction result represents the predicted category, and y represents the actual category.

[0073] Preferably, the ultrasonic time difference ratio (Δt ratio)As a network input feature, a hard constraint loss function is used to ensure that the prediction conforms to the wave equation.

[0074] As an implementation, after the training of the above network model is completed, model distillation technology is adopted, and heterogeneous model distillation is used in the aspect of lightweight implementation. Specifically, the above network model is used as the teacher model, which adopts the HRNet-W48 architecture and sets Transformer as the accuracy guidance. The student model adopts the MobileNetV3 architecture and introduces LSTM for deployment guidance, so as to perform model distillation. The distillation loss is KL(Teacher_feat||Student_feat)+MSE(Teacher_out,Student_out). After distillation, the number of parameters is compressed to less than 2M, and a student model with a smaller number of parameters is obtained, making it suitable for automotive-grade chip deployment.

[0075] Step S3: Integrate the position, distance, and category of the obstacle in front of the vehicle into the adaptive front lighting system, and automatically adjust the direction, intensity, and range of the headlight irradiation.

[0076] In this embodiment, the obstacle position, distance, and classification results obtained based on the ultrasonic detection signal are integrated into the adaptive headlight adjustment algorithm. The system automatically adjusts the irradiation direction, intensity, and range of the headlight according to the detection information to adapt to the current road conditions and improve driving safety.

[0077] Specifically, the AFS system automatically adjusts the irradiation direction, intensity, and range of the headlight according to the position, distance, and category of the obstacle in front of the vehicle. The adjustment process is as Figure 2 shown, including the following steps:

[0078] Step S3.1: The AFS system continuously obtains the obstacle detection signal, multi-sensor signal, and the current state signal of the vehicle during the driving process, and determines whether to start the headlight adaptive adjustment according to the obtained signals.

[0079] Among them, the obstacle detection signal includes the position, distance, and category information of the obstacle detected in front of the vehicle. The multi-sensor signal includes the steering wheel angle, acceleration, speed, body height, etc. The current state signal of the vehicle includes the AFS function state signal, reverse state signal, and low beam state signal.

[0080] Further, when an obstacle is detected in front of the vehicle (i.e., when an obstacle detection signal is obtained at the current moment), it is then determined whether there is an abnormality in the multi-sensor signals. If there is an abnormality, a fault warning is given. Conversely, if there is no abnormality, it is determined whether to activate the headlight adaptive adjustment according to the current state signal of the vehicle. When the AFS function status signal is ON (i.e., the enabled state), the reverse gear status signal is OFF (i.e., the disabled state), and the low beam status signal lamp is ON (i.e., the enabled state), it is determined to activate the headlight adaptive adjustment at this time.

[0081] Step S3.2: When it is determined to activate the headlight adaptive adjustment, determine the headlight adjustment mode to be executed by the AFS system according to the current vehicle speed. According to the current headlight adjustment mode, combine the position, distance, and category information of the obstacle detected in front of the vehicle to adjust the irradiation direction, intensity, and range of the headlights.

[0082] Among them, the AFS low beam is divided into 4 types of headlight adjustment modes: C-level basic low beam, V-level urban road, E-level highway, and W-level wet road surface. And each mode or level can be combined with the cornering lighting mode. When it is determined to activate the headlight adaptive adjustment at this time, determine the corresponding headlight adjustment mode according to the current vehicle speed. When the vehicle speed > 10 km / h, enter the C-level basic low beam mode at this time. Conversely, no headlight adjustment is performed; on the basis of the C-level basic low beam mode, when the vehicle speed < 60 km / h, enter the V-level urban road mode at this time; when 60 km / h < vehicle speed < 80 km / h, enter the W-level wet road surface mode at this time; when the vehicle speed > 80 km / h, enter the E-level highway mode at this time.

[0083] Further, according to the current headlight adjustment mode, combine the position, distance, and category information of the obstacle detected in front of the vehicle to adjust the irradiation direction, intensity, and range of the headlights, including:

[0084] (1) Adjustment of the headlight irradiation direction, including horizontal and vertical adjustments, that is:

[0085] First, according to the position of the obstacle in front of the vehicle, that is, the left / right offset position of the obstacle relative to the vehicle, determine the left and right headlights to be adjusted;

[0086] Then, according to the distance of the obstacle in front of the vehicle, dynamically adjust the horizontal deflection angle of the headlights; at the same time, according to the change amount of the vehicle body height, dynamically adjust the vertical deflection angle of the headlights. Specifically, taking the detection of an obstacle in the left front as an example, it is determined at this time that the left headlight needs to be adjusted for the deflection angle. In the horizontal direction, the adjustment formula for the left headlight to deflect to the left by an angle is:

[0087]

[0088] Among them, Δx is the lateral offset (which can be determined according to the type of obstacle. For example, the offset corresponding to a small obstacle is A, and the offset corresponding to a large obstacle is B), and d is the distance to the obstacle.

[0089] In the vertical direction, the compensation of the vertical deflection angle of the vehicle lamp is determined based on the change in vehicle body height and the installation position of the vehicle lamp. The calculation formula for the vertical deflection angle of the vehicle lamp is as follows:

[0090]

[0091] Among them, a represents the dynamic correction coefficient, L represents the horizontal distance from the installation position of the vehicle lamp to the pitch center of the vehicle, and Δh represents the change in vehicle body height.

[0092] (2) Adjustment of the vehicle lamp illumination intensity, that is:

[0093] According to the obstacle distance d and the classification result, the dynamic adjustment of the vehicle lamp illumination intensity is carried out, and its formula is:

[0094] ΔI = I max ·e -kd ;

[0095] Among them, I max represents the maximum illumination intensity, d is the obstacle distance; k represents the adjustment coefficient, which is determined according to the type of obstacle. For example, when the obstacle type is a pedestrian, k = 0.2 is taken, and at this time, it can quickly decay to a safe intensity at a short distance; when the obstacle type is a vehicle, k = 0.1 is taken, and at this time, it can gradually decay at a medium distance to avoid high beam interference.

[0096] (3) Adjustment of the vehicle lamp illumination range, that is:

[0097] According to the current vehicle lamp adjustment mode, determine the adjustment method of the vehicle lamp illumination range. When in the V-level urban road and W-level wet road surface mode, expand the lateral coverage range of the light to 120° to enhance the recognition of road surface details; when in the E-level highway mode, focus the light into a narrow beam to improve the long-distance vision.

[0098] Through the above methods, the irradiation direction, intensity and range of the vehicle lamp can be adjusted to realize the intelligent control of the vehicle lamp and improve the driving safety.

[0099] As another implementation, the results of ultrasonic detection and image recognition can also be integrated by calculating the weighted sum to obtain the final information on the position, distance, and category of obstacles in front of the vehicle. Preferably, in adverse weather or environments such as rain and fog, the ultrasonic data weights can be further automatically adjusted according to the actual situation to ensure the robustness of the system. For example, the position and distance of the obstacle obtained by ultrasonic detection are weighted and fused with the recognition results of the vision sensor. In adverse weather (such as rain and fog), the weight α of the ultrasonic data is set to 0.8, which is higher than the weight β of the vision data, which is 0.2, to improve the robustness of the system. It can be expressed by the following fusion formula:

[0100] P final =αP 超声 +βP 视觉 ;

[0101] where P 超声 and P 视觉 are the probability distributions of obstacles output by the ultrasonic and vision sensors respectively.

[0102] The above solution proposed in this embodiment can accurately identify various obstacles in front. Especially in the case of poor vision conditions, compared with traditional methods, this embodiment shows significant performance advantages in adjusting the headlight strategy, and can effectively improve driving safety and comfort at night and in adverse weather conditions.

[0103] Embodiment 2

[0104] This embodiment provides an adaptive headlight control system assisted by ultrasonic detection, including:

[0105] An environmental information acquisition module for real-time acquisition of external environmental images during vehicle driving;

[0106] An obstacle recognition module for judging the current driving environment according to the external environmental image. If the current driving environment is abnormal, ultrasonic detection in front of the vehicle is started, and the obstacles in front of the vehicle, their positions, distances, and categories are recognized according to the real-time obtained ultrasonic detection signals; otherwise, recognition is performed according to the external environmental image. Among them, to recognize the obstacles in front of the vehicle according to the ultrasonic detection signal, the position and distance of the obstacles in front of the vehicle relative to the vehicle are calculated, and then the signal is input into a classification model based on deep learning to recognize the category of the obstacles in front of the vehicle;

[0107] An adaptive headlight regulation module for integrating the position, distance, and category of the obstacles in front of the vehicle into the adaptive front lighting system to automatically adjust the direction, intensity, and range of headlight irradiation.

[0108] Embodiment 3

[0109] This embodiment provides a vehicle that executes the adaptive headlight control method based on ultrasonic detection proposed in Embodiment 1, or includes the adaptive headlight control system based on ultrasonic detection proposed in Embodiment 2.

[0110] The steps involved in the above Embodiments 2 and 3 correspond to those in Embodiment 1 of the method. For specific implementation manners, reference may be made to the relevant description part of Embodiment 1.

[0111] Those skilled in the art should understand that the above-mentioned modules or steps of the present invention can be implemented by a general-purpose computer device. Optionally, they can be implemented by program codes executable by the computing device, so that they can be stored in the storage device and executed by the computing device, or they can be separately fabricated into individual integrated circuit modules, or multiple modules or steps among them can be fabricated into a single integrated circuit module for implementation. The present invention is not limited to any specific combination of hardware and software.

[0112] The above are only the preferred embodiments of the present invention. Although the specific implementation manners of the present invention are described in conjunction with the accompanying drawings, it is not a limitation to the protection scope of the present invention. Those skilled in the art should understand that various modifications or deformations that can be made without creative labor on the basis of the technical solution of the present invention are still within the protection scope of the present invention.

Claims

1. An adaptive headlight control method based on ultrasonic detection assistance, characterized in that, Including: Obtaining external environment images during the vehicle's driving in real time; Judging the current driving environment according to the external environment images. When the current driving environment is abnormal, ultrasonic detection in front of the vehicle is started, and the obstacles in front of the vehicle, their positions, distances and categories are identified according to the ultrasonic detection signals obtained in real time; otherwise, identification is performed according to the external environment images; Among them, to identify the obstacles in front of the vehicle according to the ultrasonic detection signals, calculate the position and distance of the obstacles in front of the vehicle relative to the vehicle, and then input the signals into a classification model based on deep learning to identify the categories of the obstacles in front of the vehicle; Integrate the position, distance and category of the obstacles in front of the vehicle into the adaptive front lighting system to automatically adjust the direction, intensity and range of the headlight irradiation.

2. The adaptive headlight control method based on ultrasonic detection assistance according to claim 1, wherein Identifying the obstacles in front of the vehicle and their positions and distances according to the ultrasonic detection signals obtained in real time includes: Sending and receiving ultrasonic detection signals in real time, and dynamically correcting the sound speed value of signal propagation according to the current environmental temperature and humidity; Judging the position of the obstacles in front of the vehicle relative to the vehicle according to the time difference ratio of the ultrasonic detection signal sending and receiving; Based on the position of the obstacle relative to the vehicle, select the corresponding ultrasonic detection signal, and combine the corrected sound speed value to calculate the distance of the obstacle in front of the vehicle.

3. The adaptive headlight control method based on ultrasonic detection assistance according to claim 2, characterized in that The formula for dynamic correction of the sound speed is: v = 331.4 + 0.6 * T + 0.0124 * RH; Among them, the basic sound speed is 331.4 m / s, that is, the ultrasonic speed at 0 °C; the temperature coefficient is 0.6 m / (s·°C), that is, when the temperature rises by 1 °C, the sound speed increases by 0.6 m / s; T is the environmental temperature, in degrees Celsius; RH represents the humidity, in percentage.

4. The adaptive headlight control method based on ultrasonic detection assistance according to claim 2, wherein The formula for calculating the distance of the obstacle in front of the vehicle is: d = (ct) / 2; In the above formula, d represents the distance, c represents the sound speed, and t represents the round-trip time of the ultrasonic signal.

5. The adaptive headlight control method based on ultrasonic detection assistance according to claim 1, characterized in that, Identifying the category of the obstacle in front of the vehicle according to the ultrasonic detection signal obtained in real time includes: inputting the received ultrasonic detection signal into a classification model based on deep learning to identify the category of the obstacle in front of the vehicle; The training of the classification model based on deep learning includes: Collecting a number of ultrasonic detection signals reflected under different conditions, labeling the obstacle category labels of each signal, and using the signals with labeled labels as samples to construct a sample training set; among them, different conditions include setting different types of obstacles on the road under different bad weather; Constructing a neural network model, using the sample training set to supervise and train the model, and continuously iteratively learning the potential relationship between the reflection signal characteristics of different obstacles and the obstacle categories until the training is completed.

6. The adaptive headlight control method based on ultrasonic detection assistance according to claim 1, characterized in that Integrating the position, distance and category of the obstacles in front of the vehicle into the adaptive front lighting system to automatically adjust the direction, intensity and range of the headlight irradiation includes: The adaptive front lighting system obtains in real time the obstacle detection signal, multi-sensor signals, and the current vehicle state signal during vehicle driving, and determines whether to start the adaptive adjustment of vehicle lights according to the obtained signals. Among them, the obstacle detection signal includes the position, distance, and category information of the obstacle in front of the vehicle detected, the multi-sensor signals include the steering wheel angle, acceleration, speed, and vehicle body height, and the current vehicle state signal includes the AFS function state signal, reverse state signal, and low beam state signal lamp. When it is determined to start the adaptive adjustment of vehicle lights, the headlight adjustment mode to be executed by the AFS system is determined according to the current vehicle speed. According to the current headlight adjustment mode, combined with the position, distance, and category information of the obstacle in front of the vehicle detected, the irradiation direction, intensity, and range of the vehicle lights are adjusted.

7. An adaptive vehicle headlight control system assisted by ultrasonic detection, characterized in that, It includes: An environmental information acquisition module for acquiring in real time the external environment image during vehicle driving. An obstacle recognition module for judging the current driving environment according to the external environment image. If the current driving environment is abnormal, ultrasonic detection in front of the vehicle is started, and the obstacle in front of the vehicle and its position, distance, and category are recognized according to the ultrasonic detection signal acquired in real time. Otherwise, recognition is performed according to the external environment image. Among them, according to the ultrasonic detection signal, the obstacle in front of the vehicle is recognized, the position and distance of the obstacle in front of the vehicle relative to the vehicle are calculated, and then the signal is input into the classification model based on deep learning to recognize the category of the obstacle in front of the vehicle. An adaptive headlight control module for integrating the position, distance, and category of the obstacle in front of the vehicle into the adaptive front lighting system and automatically adjusting the irradiation direction, intensity, and range of the vehicle lights.

8. The adaptive headlight control system based on ultrasonic detection assistance according to claim 7, characterized in that, Recognizing the obstacle in front of the vehicle and its position and distance according to the ultrasonic detection signal acquired in real time includes: Sending and receiving ultrasonic detection signals in real time, and dynamically correcting the sound speed value of signal propagation according to the current environmental temperature and humidity. Judging the position of the obstacle in front of the vehicle relative to the vehicle according to the time difference ratio of the ultrasonic detection signal sending and receiving. Based on the position of the obstacle relative to the vehicle, the corresponding ultrasonic detection signal is selected, and combined with the corrected sound speed value, the distance of the obstacle in front of the vehicle is calculated.

9. The adaptive headlight control system based on ultrasonic detection assistance according to claim 7, characterized in that, Recognizing the category of the obstacle in front of the vehicle according to the ultrasonic detection signal acquired in real time includes: inputting the received ultrasonic detection signal into the classification model based on deep learning to recognize the category of the obstacle in front of the vehicle. The training of the classification model based on deep learning includes: Collecting a number of ultrasonic detection signals reflected under different conditions, labeling the obstacle category label of each signal, and using the signals with labeled labels as samples to construct a sample training set. Among them, different conditions include setting different categories of obstacles on the road under different bad weather conditions. Constructing a neural network model, using the sample training set to supervise and train the model, and continuously iteratively learning the potential relationship between the reflected signal characteristics of different obstacles and the obstacle category until the training is completed.

10. A vehicle, characterized in that, Execute the adaptive headlight control method assisted by ultrasonic detection described in any one of claims 1-6, or include the adaptive headlight control system assisted by ultrasonic detection described in any one of claims 7-9.

Images