Self-cleaning method and system for windshield in front of ADAS camera

Abstract

The invention discloses a self-cleaning method and system for a windshield in front of an ADAS camera, and the method comprises the steps: recognizing the fuzzy degree of an image of the windshield in front of the ADAS camera based on a constructed fuzzy degree recognition model in a vehicle driving process; and according to the fuzzy degree of the identified image, a corresponding self-cleaning mode is adopted to perform self-cleaning on the windshield in front of the ADAS camera. The fuzzy degree of the windshield pollution is judged in real time by constructing the fuzzy degree recognition model, the cleaning modes with different intensities are automatically matched based on the recognition result, the technical problems that a traditional scheme depends on manual operation and is low in cleaning efficiency are effectively solved, intelligent graded cleaning control is achieved, and the cleaning efficiency is improved. And the operation reliability of the ADAS system is obviously improved.

Description

TECHNICAL FIELD

[0001] The application belongs to the technical field of intelligent driving assistance, and particularly relates to a self-cleaning method and system for a front windshield of an ADAS camera. BACKGROUND

[0002] Commercial vehicles face complex road environments during long-term operation, and the front windshield is easily affected by various pollutants such as rainwater, mud, dust, etc. These pollutants can cause the image quality collected by the ADAS camera to decrease, seriously affecting the function implementation of the advanced driving assistance system. Specifically, in rainy conditions, water droplets on the windshield can cause image blurring; when driving on muddy roads, splashed mud can form stains on the glass surface; and long-term accumulated dust can reduce image contrast. These situations can cause the accuracy of key functions such as front target recognition and lane line detection to decrease, thereby affecting the performance of systems such as automatic emergency braking (AEB), adaptive cruise control (ACC), lane keeping assistance (LKA), and lane centering control (LCC), and potentially posing a risk to driving safety.

[0003] Traditional solutions mainly use mechanical cleaning devices that require the driver to manually trigger the cleaning function. This type of solution has obvious limitations: first, the reliance on manual operation results in insufficient automation, and the driver may not be able to judge the camera cleaning needs in real time under complex road conditions; second, the mechanical structure design is complex, increasing system costs and maintenance difficulty. In addition, existing technologies lack intelligent judgment capabilities for pollution levels and are unable to adopt differentiated cleaning strategies based on actual pollution conditions, resulting in low cleaning efficiency or waste of cleaning resources.

[0004] In view of the above problems, existing technologies need to be improved. SUMMARY

[0005] The purpose of the present application is to solve the problems in the background art, and to provide a self-cleaning method and system for a front windshield of an ADAS camera, which has the advantages of improving cleaning efficiency and realizing intelligent graded cleaning.

[0006] The technical solution adopted by the present application is as follows: a self-cleaning method for a front windshield of an ADAS camera, During vehicle driving, the blur degree of the image of the front windshield of the ADAS camera is identified based on a constructed blur degree identification model. According to the identified blur degree of the image, a corresponding self-cleaning mode is adopted to perform self-cleaning on the front windshield of the ADAS camera.

[0007] Further, the construction process of the blur degree identification model is as follows: training sets and test sets of different blur degrees are obtained; select a pre-trained ResNet model as a base model; train the base model based on a training set, and test the trained model by using a test set to obtain an end-to-end blur degree recognition model.

[0008] Further, a plurality of pictures of a dirty front windshield of an ADAS camera are collected, the plurality of pictures are divided into a plurality of blur levels according to different blur degrees, the pictures divided into the blur levels are enhanced to form an image data set with clear and blur marks, and the image data set is divided into a training set and a test set according to a proportion.

[0009] Further, the plurality of blur levels include: a first level: the blur degree is less than A1; a second level: the blur degree is greater than or equal to A1 and less than A2; a third level: the blur degree is greater than or equal to A2 and less than A3; a fourth level: the blur degree is greater than A3; wherein A1, A2 and A3 are respectively the first, second and third blur degree percentages.

[0010] Further, the corresponding self-cleaning mode is adopted according to the recognized blur degree, including: if the recognized blur degree is less than A1, a first self-cleaning mode is adopted; if the recognized blur degree is greater than or equal to A1 and less than A2, a second self-cleaning mode is adopted; if the recognized blur degree is greater than or equal to A2 and less than A3, a third self-cleaning mode is adopted; if the recognized blur degree is greater than A3, a fourth self-cleaning mode is adopted; wherein A1, A2 and A3 are respectively the first, second and third blur degree percentages.

[0011] Further, the first self-cleaning mode is to exit the automatic self-cleaning mode, the second self-cleaning mode is that the wiper intermittently works at a set time interval; the third self-cleaning mode is that the wiper continuously works at a set first frequency; the fourth self-cleaning mode is that the wiper continuously works at a set second frequency; the second frequency is greater than the first frequency.

[0012] Further, when the wiper continuously works at the set second frequency, the spraying of cleaning liquid is controlled.

[0013] Further, in the self-cleaning process of the windshield in front of the ADAS camera, if the blur degree is identified to be reduced below the lower critical point of the corresponding blur level, the time is counted from the time when the blur degree is reduced to the critical value, and after the set time is counted, the self-cleaning mode is switched to a lower level to perform self-cleaning.

[0014] Further, the self-cleaning mode is switched to a lower level to perform self-cleaning if the blur degree is identified to be reduced below the lower critical point of the corresponding blur level, the time is counted from the time when the blur degree is reduced to the critical value, and after the set time is counted. If the blur degree identified at the current time corresponds to the blur level i, the self-cleaning mode adopted is the jth self-cleaning mode. At a certain time after self-cleaning for a period of time, if the identified blur degree is lower than the lower critical point of the blur degree of the ith level, the time is counted, and after the set time is counted, the self-cleaning mode is switched from the jth self-cleaning mode to the j-1th self-cleaning mode. The i≥2, j≥2.

[0015] An ADAS camera windshield self-cleaning system in front of the windshield, comprising An image acquisition module for acquiring images of the windshield in front of the ADAS camera in real time during vehicle driving; A blur degree identification module for identifying the acquired images based on the constructed blur degree identification model to determine the image blur degree; A self-cleaning module for self-cleaning the windshield in front of the ADAS camera according to the image blur degree and using the corresponding self-cleaning mode.

[0016] The beneficial effects of the present application are: The present application can ensure that the dirt, water marks, etc. on the windshield are automatically cleaned, and the image of the ADAS camera is clear, effectively solving the technical problems of the traditional scheme relying on manual operation and low cleaning efficiency, realizing intelligent hierarchical cleaning control, and significantly improving the operation reliability of the ADAS system. BRIEF DESCRIPTION OF DRAWINGS

[0017] Figure 1 The flowchart of the present application. DETAILED DESCRIPTION

[0018] The specific embodiments of the present application will be further described below with reference to the drawings. It should be noted that the description of these embodiments is used to help understand the present application, but does not constitute a limitation on the present application. In addition, the technical features involved in the various embodiments of the present application described below can be combined with each other as long as they do not conflict with each other.

[0019] In the prior art, the commercial vehicle has a long operating mileage and complex working conditions. In actual driving, the front windshield is easily affected by rain, mud and other environmental factors, resulting in blurred images obtained by the ADAS camera, affecting target recognition and lane line detection, and further causing abnormal functions such as AEB / ACC. The traditional cleaning scheme relies on the driver manually triggering the mechanical cleaning device, which has the problems of response lag, complex structure and low automation, and cannot guarantee real-time clear camera view.

[0020] To solve the above problems, research has found that it is difficult for the driver to monitor the cleaning state of the camera front view in real time, and the existing mechanical cleaning device cannot dynamically adjust the cleaning intensity according to the dirt level. Through analysis, it is found that if the blurring degree of the camera view can be quantitatively identified and a corresponding relationship between the blurring degree and the cleaning mode is established, the automation and precision of the cleaning operation can be realized. Further, by using an image recognition model to replace manual judgment and combining a hierarchical cleaning strategy, the problems of relying on manual triggering and low cleaning efficiency of the traditional method can be solved.

[0021] Therefore, the present application provides a self-cleaning method for the front windshield of an ADAS camera, as shown in Figure 1 During vehicle driving, the blurring degree of the image of the front windshield of the ADAS camera is identified based on the constructed blurring degree identification model, and the windshield is self-cleaned according to the corresponding self-cleaning mode according to the identification result.

[0022] Among them, the blurring degree identification model refers to an end-to-end image analysis model established through a training data set. Specifically, a pre-trained ResNet model can be used for transfer learning, and the model can output a blur level classification result by inputting real-time images collected by the camera. This model can quantitatively evaluate the influence of glass surface dirt on image quality. The self-cleaning mode refers to a cleaning intensity strategy matched according to the blur level, for example, by controlling the working frequency of the wiper or the spraying amount of cleaning liquid to realize different levels of cleaning operation. This mode balances resource optimization and cleaning efficiency by dynamically adjusting the cleaning intensity.

[0023] Specifically, during vehicle driving, the camera continuously captures images of the front windshield, and the blur level recognition model analyzes the images in real time and outputs the blur level. For example, when the model determines that the blur level is light, intermittent wiper operation is triggered; if severe blur is detected, high-frequency wiper operation is started in combination with cleaning fluid spraying. The blur level is continuously monitored during cleaning, and when the blur level decreases below the threshold, the cleaning mode is automatically switched to low intensity. This closed-loop control mechanism ensures that the cleaning operation is always matched to the current dirt level.

[0024] The present application realizes intelligent cleaning operation through image recognition and hierarchical control, without the need for driver intervention to maintain clear camera vision, while simplifying mechanical structure design. The present application can real-time perceive the dirt level of the windshield during vehicle driving, automatically trigger adaptive cleaning operation, effectively avoid ADAS function abnormality caused by image blur, and improve system reliability. In addition, the hierarchical cleaning strategy ensures cleaning effect while reducing wiper wear and cleaning fluid consumption, prolonging the service life of the components.

[0025] The present application further proposes a construction process of the blur level recognition model: obtaining a training set and a test set of different blur levels; selecting a pre-trained ResNet model as a base model; training the base model based on the training set, and testing the trained model using the test set to obtain an end-to-end blur level recognition model.

[0026] Among them, the training set and the test set of different blur levels refer to a data set containing various windshield dirt states, which can be realized by pre-collecting blurred images captured by the camera during actual vehicle operation and classifying and labeling, to cover image degradation caused by different interference factors such as raindrops, stains, and oil films. The pre-trained ResNet model refers to a deep residual network pre-trained based on a large-scale image data set, which can be realized by using a pre-trained ResNet-50 model of the ImageNet data set, using its learned general image feature extraction ability as a basis. The end-to-end blur level recognition model refers to a complete processing flow that directly outputs the blur level classification result from the original image input, which can be realized by freezing the front-end convolutional layer of the ResNet model and fine-tuning the fully connected layer, effectively reducing the computational resource consumption of model training.

[0027] Specifically, in the model construction process, first, the windshield images under different working conditions are collected by the vehicle-mounted camera, and the classification data set is formed by manual labeling according to the blur degree. Then, the weight parameters of the pre-trained ResNet model are loaded, the bottom feature extraction capability is retained, and only the top classifier layer is redesigned to adapt to the blur level classification task. In the training stage, the cross-entropy loss function is used to optimize the model parameters, and the network weight is adjusted through the back propagation algorithm. In the test stage, the independent verification set is used to evaluate the classification accuracy of the model, and when the preset threshold (such as more than 95%) is reached, the model deployment is completed. This construction method enables the model to automatically extract the blur features in the image, without the need for manual feature extraction algorithm design, and effectively solves the problem of insufficient data in practical scenarios using transfer learning technology.

[0028] The accurate identification of the blur degree is the key to ensuring the clarity of the camera image and realizing self-cleaning. The present application uses a direct method to identify the blur degree. The image captured by the camera will have a blurred part, and the degree of blur will vary with the degree of dirt. An end-to-end large model is trained to detect the degree of dirt under various working conditions.

[0029] The present application uses a deep learning model to automatically learn the mapping relationship between blur features and cleaning needs, significantly improving the accuracy of judgment. The pre-trained model fine-tuning method is used to shorten the model development cycle while ensuring classification accuracy, so that the system can quickly adapt to the camera installation position and windshield characteristics of different vehicle models.

[0030] The present application further proposes to collect pictures of the front windshield with dirt in front of the ADAS camera, divide the pictures into multiple blur levels according to the different blur degrees, enhance the pictures divided by blur levels, and form an image data set with clear and blurred markers. The image data set is divided into a training set and a test set according to the proportion.

[0031] Among them, the blur level division refers to quantitatively classifying according to the stain coverage area or texture distortion degree in the image, which can be realized by edge sharpness detection or gray variance calculation, and different cleaning needs are set by setting different threshold intervals.

[0032] Among them, the enhancement processing refers to the operation of rotating, scaling or adding noise to the original image, which can be realized by random brightness adjustment or Gaussian blur superposition, and the model robustness is improved by simulating different environmental interference factors.

[0033] Among them, the image data set division refers to allocating samples to training and testing links according to a predetermined proportion, which can be realized by stratified sampling or cross-validation, ensuring that the samples of different blur levels are evenly distributed in the training and testing links.

[0034] Specifically, first, the real scene pictures of the windshield with raindrops and mud in actual driving are collected by the vehicle-mounted camera to avoid feature deviation caused by artificially synthesized data. Then, the blur degree of each picture is quantitatively evaluated by using an image processing algorithm, for example, the local contrast or high-frequency component energy value is calculated, and the pictures are classified into different blur level intervals according to the evaluation results. In order to increase the diversity of data, the classified pictures are geometrically transformed or random noise is added, for example, the image exposure is adjusted or the simulated water stain texture is superimposed. Finally, the processed pictures are randomly allocated to the training set and the test set according to a fixed ratio, for example, a ratio of seven to three, to ensure that the model can learn the features sufficiently and avoid overfitting during the training process.

[0035] The present application can reflect the light changes and stain shape differences in the actual environment by collecting dirty pictures in real scenes and introducing data enhancement technology, thereby improving the recognition accuracy of the model in dynamic scenes.

[0036] The present application can accurately distinguish different cleaning needs by constructing a multi-level blur dataset in a real scene, providing a reliable basis for subsequent grading control, and enhancing the generalization ability of the model to interference factors to avoid misjudgment or omission problems caused by overfitting.

[0037] The present application further proposes a plurality of blur levels including a first level, a second level, a third level, and a fourth level, wherein the blur degree of the first level is less than A1, the blur degree of the second level is greater than or equal to A1 and less than A2, the blur degree of the third level is greater than or equal to A2 and less than A3, and the blur degree of the fourth level is greater than A3, i.e., four levels of "clear", "slightly blurred", "moderately blurred", and "severely blurred" are divided. Wherein, A1, A2, A3 are the first, second, and third blur degree percentages respectively.

[0038] Wherein, the blur degree percentage refers to the ratio of the area of the windshield region covered by the image processing algorithm to the total area of the field of view, which can be realized by pixel gray value comparison or edge detection algorithm, and is used for quantitative evaluation of the coverage degree of dirt. Wherein, the first, second, and third blur degree percentages are threshold parameters determined by experimental tests, which can be realized by statistical calibration of image samples under different working conditions, for example, A1 can be 10%, A2 can be 25%, and A3 can be 50%, which is used to establish the mapping relationship between the blur degree and the cleaning demand.

[0039] Specifically, by dividing the blur degree into four progressive intervals, a hierarchical judgment standard from low to high is constructed. When the system detects that the blur degree is in the first level, it indicates that there is only slight dirt on the windshield, and at this time, high-intensity cleaning does not need to be triggered; when the blur degree enters the second level, it indicates that the dirt has an identifiable impact on the camera's field of view, and the basic cleaning mode needs to be started; when the blur degree reaches the third level, it indicates that the dirt coverage range has expanded, and higher frequency cleaning actions need to be used; when it enters the fourth level, it indicates that there is a serious risk of obstruction, and the highest intensity cleaning strategy needs to be activated. The division of the four levels covers the whole scene from slight to serious, providing accurate trigger basis for dynamic adjustment of the cleaning mode.

[0040] By multi-level threshold division, the application can match the corresponding cleaning intensity according to the actual dirt degree, avoiding resource waste and equipment loss. The application can accurately identify the dirt coverage degree of the windshield and dynamically adjust the cleaning strategy according to different levels of blur state. For example, unnecessary cleaning actions are avoided in low-level blur state, reducing the wear of the wiper; high-intensity cleaning is started in time in high-level blur state to prevent the camera's field of view from being continuously obstructed. This hierarchical control mechanism effectively solves the problem of mismatch between cleaning mode and dirt degree in the prior art, improving the cleaning efficiency and system reliability.

[0041] The application further proposes a method for adopting corresponding self-cleaning modes according to the identified blur degree of the image, comprising: if the identified blur degree is less than A1, adopting a first self-cleaning mode; if the identified blur degree is greater than or equal to A1 and less than A2, adopting a second self-cleaning mode; if the identified blur degree is greater than or equal to A2 and less than A3, adopting a third self-cleaning mode; if the identified blur degree is greater than A3, adopting a fourth self-cleaning mode.

[0042] Wherein, A1, A2, A3 refer to the calibrated blur degree percentage threshold, which can be realized by calibrating the critical value corresponding to different cleaning needs through experiment, and is used to quantify the blur state of the windshield. The first self-cleaning mode refers to the no cleaning action state, which can be realized by closing the wiper drive signal, and is used to avoid equipment idling when there is no pollution. The second self-cleaning mode refers to the intermittent cleaning mode, which can be realized by setting the periodic start-stop time parameter to control the wiper, and is used to deal with slight dirt. The third self-cleaning mode refers to the low-frequency continuous cleaning mode, which can be realized by setting the fixed low-frequency parameter to control the wiper working period, and is used to handle moderate pollution. The fourth self-cleaning mode refers to the high-frequency continuous cleaning mode, which can be realized by setting a higher frequency parameter and linking the cleaning liquid spraying device, and is used to eliminate serious stains.

[0043] Specifically, the visual blur state is divided into four intervals by a pre-set blur percentage threshold, each interval corresponding to a differentiated cleaning intensity. When the visual detection module identifies that the current blur level is in the A1-A2 interval, an intermittent wiper action is triggered, which maintains the basic cleaning ability while reducing equipment wear and tear. When the blur level exceeds the A3 threshold, the system automatically switches to a high-frequency cleaning mode and simultaneously starts spraying cleaning liquid, forming a composite cleaning mechanism. This hierarchical control strategy dynamically matches the cleaning intensity with the pollution level, optimizing resource utilization while ensuring cleaning effectiveness.

[0044] The present application realizes automatic hierarchical response by establishing a mapping relationship between blur level and cleaning mode, and triggers different working modes through multiple threshold values, avoiding over-cleaning while ensuring processing capacity in case of severe pollution. In rainy weather scenarios, the system can automatically identify the windshield water film thickness and match the corresponding wiper frequency to maintain a clear view for ADAS cameras. When driving on muddy roads, the system can quickly respond to sudden heavy pollution, restore visibility in time through high-frequency cleaning mode combined with cleaning liquid spraying, and ensure the normal operation of the driving assistance system.

[0045] The present application further proposes that the first self-cleaning mode is to exit the automatic self-cleaning mode, the second self-cleaning mode is that the wiper intermittently works at a set time interval, the third self-cleaning mode is that the wiper continuously works at a set first frequency, and the fourth self-cleaning mode is that the wiper continuously works at a set second frequency, the second frequency being greater than the first frequency.

[0046] The exit of the automatic self-cleaning mode means that the system stops executing the pre-set cleaning program, which can be realized by turning off the wiper drive signal. This mode is suitable for light blur state to avoid energy waste. The time interval intermittent work means that the wiper performs a single wiping action at a fixed period, for example, triggering a wiping action every 30 seconds, which is used to maintain the basic cleaning effect. The first frequency continuous work means that the wiper continuously wipes at a frequency of 10-15 times per minute, which is suitable for moderate dirt removal. The second frequency continuous work means that the wiper continuously wipes at a frequency of 20-30 times per minute, which is higher than the first frequency and is used to deal with severe dirt or liquid residue.

[0047] Specifically, when the image blur degree is in the first level, the system terminates the automatic cleaning process to reduce energy consumption. When the second level of blur is detected, the wiper blade removes surface attachments with periodic intermittent action, such as performing a single sweep every 30 seconds. In the third level state, the wiper blade switches to a continuous mode of 12 times per minute to enhance cleaning strength. When reaching the fourth level, the wiper blade is raised to a high-frequency action of 25 times per minute, while cooperating with the cleaning liquid spraying device. The frequency difference between each mode is realized by the controller preset parameters, for example, the first frequency is set to 12 times per minute, and the second frequency is set to 25 times per minute, forming a step-by-step cleaning strength.

[0048] The present application sets up the exit mode, the intermittent mode and the double-frequency continuous mode, and constructs a step-by-step cleaning control strategy that accurately matches the dirt degree, effectively reduces the wiper blade wear and cleaning liquid consumption under the premise of ensuring the cleaning effect. The combination of high-frequency continuous operation mode and intermittent mode can quickly remove stubborn stains while avoiding equipment wear caused by long-term high-frequency operation.

[0049] The present application further proposes a technical solution for controlling the spraying of cleaning liquid when the wiper blade is continuously working at the set second frequency.

[0050] Among them, the second frequency refers to the wiper blade working frequency higher than the first frequency, which can be realized by adjusting the duty cycle of the motor driving signal to increase the number of physical scraping strength.

[0051] Among them, the spraying of cleaning liquid refers to spraying cleaning liquid to the windshield surface through the vehicle-mounted liquid tank, which can be realized by controlling the opening and closing time of the nozzle with the electromagnetic valve, and using chemical solvents to dissolve stubborn stains attached to the glass surface.

[0052] Specifically, when the system detects that the fog degree of the windshield reaches the fourth level, the wiper blade continuously works at the second frequency, and the cleaning liquid spraying device is triggered. At this time, the cleaning liquid uniformly covers the glass surface through the nozzle, dissolves the contaminants such as oil film and shellac that are difficult to remove, and then the high-frequency moving wiper blade quickly removes the dissolved stains. The process coordinates the working time sequence of the wiper motor and the cleaning liquid pump through the electronic control unit, and the synergistic effect of physical scraping and chemical cleaning can be realized without adding additional mechanical structure.

[0053] The present application can automatically perform efficient cleaning in a severe fog state, and through the cooperation of cleaning liquid dissolution and high-frequency scraping, it can eliminate contaminants that are difficult to remove by conventional physical cleaning, while maintaining the structural simplicity of the vehicle-mounted device, reducing system complexity and manufacturing cost.

[0054] The application further proposes that, in the self-cleaning process of the windshield in front of the ADAS camera, if the blur degree is identified to be reduced to below the lower critical point of the corresponding blur level, timing is started from the time when the blur degree is reduced to the critical value, and after the timing setting time, the self-cleaning mode is switched to a lower level to perform self-cleaning.

[0055] The lower critical point of the corresponding blur level refers to the lowest blur degree threshold corresponding to the current blur level, which can be realized by using a pre-calibrated percentage parameter, for example, the lower critical point of the third level is set as A2, which is used to judge whether the current cleaning effect meets the downgrade condition.

[0056] The timing setting time refers to the delay time before triggering mode switching, which can be realized by using a fixed or dynamically adjusted time parameter, for example, between 5 seconds and 30 seconds, preferably 10 seconds or 15 seconds or 20 seconds, which is used to ensure that the reduction of the blur degree is stable and continuous.

[0057] The lower level self-cleaning mode refers to a lower action combination than the current mode cleaning intensity, which can be realized by reducing the wiper working frequency or stopping the spraying of cleaning liquid, for example, switching from continuous working at the second frequency to continuous working at the first frequency.

[0058] Specifically, when the self-cleaning mode is at a higher intensity, if the blur degree is detected to be below the lower critical point of the current level, the system will not immediately switch the mode, but start a timer. For example, when in the third self-cleaning mode, if the blur degree is reduced from A2 or above to below A2, timing is started, and if the blur degree does not rise to the current level within the set time, the second self-cleaning mode is switched to. This mechanism avoids false switching caused by temporary environmental interference or cleaning action fluctuations by delaying the judgment, while ensuring that the current cleaning mode fully functions, so that the surface state of the windshield tends to be stable before reducing the cleaning intensity.

[0059] The application reduces the number of invalid actions by introducing a delay judgment and threshold constraint, so that mode switching only occurs when the blur degree is stably below the critical value. The application solves the problem of frequent switching of cleaning modes caused by temporary cleaning effect or environmental interference, realizes the stability of the transition of the self-cleaning mode, reduces the invalid loss of the wiper and the cleaning liquid while ensuring the cleaning effect, and prolongs the service life of the related components.

[0060] The application further proposes a method for dynamically adjusting the cleaning mode during the self-cleaning process, if the blur degree identified at the current time corresponds to the blur level i, the self-cleaning mode used is the jth self-cleaning mode. At a certain moment after self-cleaning for a period of time, if the identified blur degree is lower than the lower critical point of the blur degree of the i-th level, timing is started, and after the set time, the self-cleaning mode is switched from the j-th self-cleaning mode to the j-1-th self-cleaning mode; The value of i is 2 or 3 or 4, that is, corresponding to the second, third and fourth levels described above, and the value of j is 2 or 3 or 4, that is, corresponding to the second, third and fourth self-cleaning modes described above.

[0061] Among them, the blur level i refers to the pollution degree classification according to the image blur degree, that is, the first to fourth levels described above. Specifically, an image processing algorithm can be used to calculate the edge sharpness of the windshield area, and a plurality of threshold intervals are set to realize level classification, which is used to quantify the intensity standard of different cleaning needs.

[0062] Among them, the self-cleaning mode j refers to the combination of cleaning actions corresponding to the blur level i, that is, the first to fourth self-cleaning modes described above. Specifically, the working frequency of the wiper, the spraying period of the cleaning liquid and other parameter combinations can be used to form a ladder type cleaning intensity to ensure that different pollution levels match the corresponding cleaning resource investment.

[0063] Among them, the lower critical point refers to the blur degree threshold value that triggers mode degradation. Specifically, the lowest blur index corresponding to the current blur level can be used to start the degradation judgment mechanism when the detection value is continuously lower than the index.

[0064] Among them, the set time refers to the delay period for maintaining the current cleaning mode. Specifically, an adjustable parameter in the range of 5-30 seconds can be used to record the duration of the pollution level being stable at a low level by a timer module.

[0065] Specifically, when the vehicle is in the third blur level, the system activates the third self-cleaning mode. During the continuous operation of the wiper at the first frequency, the image processing module continuously monitors the sharpness of the windshield. When the blur degree is detected to be lower than the lower critical point A2 of the third level, the timer immediately starts. If the blur degree does not rise above A2 within the set 10 seconds, the system automatically switches the cleaning mode to the second self-cleaning mode, and changes to intermittent wiper action. This delay switching mechanism effectively avoids frequent mode switching caused by temporary cleaning effect fluctuations, such as the instantaneous sharpness improvement phenomenon that may occur after a single wiping of the wiper. At the same time, by limiting the conditions of i≥2 and j≥2, it is ensured that the mode degradation operation is only performed in the working condition that requires high cleaning intensity, preventing unnecessary mode switching in the low pollution state.

[0066] The present application further proposes that the set time is 5-30 seconds.

[0067] The set time refers to a delay time period before the switching of the self-cleaning mode, and can be realized by using a timer module or a clock function built in a microcontroller, and the time period is used to ensure the stability of the change trend of the blur degree. The 5-second lower limit refers to the minimum duration for the system to maintain the current mode, and can be realized by using a hardware timing circuit or a software loop counting method, and the parameter is used to prevent false switching caused by sensor noise or instantaneous cleaning effect. The 30-second upper limit refers to the maximum waiting time for allowing the current cleaning mode to be maintained, and can be realized by using a time relay function in a programmable logic controller, and the parameter is used to limit the continuous consumption of cleaning resources in an invalid state.

[0068] Specifically, when the blur degree of the windshield is detected to be lower than the lower critical point of the current blur level, the system starts the timing module to accumulate the duration. For example, during the running of the self-cleaning mode corresponding to the third blur level, if the blur degree is detected to suddenly drop to the range of the second level, the system does not immediately downgrade the cleaning mode, but waits for any value (such as 10 seconds) in the set time interval before detecting the blur state again. If the blur degree does not rise to the original level range during the waiting period, the cleaning mode is switched to the next lower level. This mechanism filters out the instantaneous clear phenomenon caused by the temporary wiping of the wiper or the splashing of water drops, and avoids the excessive consumption of cleaning liquid caused by continuous waiting.

[0069] The delay judgment mechanism introduced in the application balances the shortest effective action time required by the physical cleaning process and the maximum allowed waiting time, which not only ensures the number of effective wiping times of the wiper blade, but also avoids the pressure fluctuation of the cleaning liquid pipeline caused by short-time multiple triggering. The setting of the time window makes the working rhythm of the cleaning system match the actual stain attachment characteristics, reduces the invalid working time of the wiper motor by about 40%, and reduces the single use amount of cleaning liquid by about 25%.

[0070] In some embodiments, during the driving of the vehicle, the camera captures images in real time, and the blur degree percentage is obtained by the perception algorithm.

[0071] When the blur degree is less than 10%, the identification of the target and the lane line is not affected, and the automatic driving is not affected.

[0072] When the blur degree is between 10% and 25%, the automatic driving is affected to a certain extent, and the self-cleaning mode needs to be started. The camera sends a signal to the BCM (body control module), the BCM controls the wiper device, and the front windshield self-cleaning mode 2 is started. During the continuous cleaning process, when the blur degree identified at a certain moment is lower than 10% and the duration reaches 10s, the self-cleaning mode 1 is switched, that is, the automatic self-cleaning mode is exited.

[0073] When the blur degree is between 25% and 50%, the vehicle is in a moderate blur state, the influence on automatic driving is increased, and the front windshield self-cleaning mode 3 needs to be started; during the continuous cleaning process, when the blur degree is identified to be lower than 25% at a certain moment and the duration reaches 10s, the self-cleaning mode 2 is switched; during the continuous cleaning process, when the blur degree is identified to be lower than 10% at a certain moment and the duration reaches 10s, the self-cleaning mode 1 is switched.

[0074] When the blur degree is greater than 50%, the vehicle is in a serious blur state, and the automatic driving is greatly influenced, and the front windshield self-cleaning mode 4 needs to be started; during the continuous cleaning process, when the blur degree is identified to be lower than 50% at a certain moment and the duration reaches 10s, the self-cleaning mode 3 is switched; during the continuous cleaning process, when the blur degree is identified to be lower than 25% at a certain moment and the duration reaches 10s, the self-cleaning mode 2 is switched; during the continuous cleaning process, when the blur degree is identified to be lower than 10% at a certain moment and the duration reaches 10s, the self-cleaning mode 1 is switched.

[0075] When the vehicle is driving on the highway at night, small flying insects on the road collide and adhere to the windshield of the vehicle, which does not affect the driver's view, but has a certain influence on the ADAS camera, causing the image to be unclear, at this time, the degree of image blur needs to be identified, and the self-cleaning mode is started according to the blur degree, if the blur degree is more than 50%, the self-cleaning mode 4 needs to be started, water is sprayed and the surface dirt is wiped off.

[0076] When the vehicle is driving in the process of light rain, the rain is not large, which does not affect the driver's view, at this time, the driver does not turn on the wiper, but the rain flows down and produces water marks on the glass, causing the image obtained by the camera to be unclear, affecting the stability of target recognition and lane line recognition, at this time, it is necessary to identify that the image has been slightly blurred in real time, and the self-cleaning mode 2 needs to be started, and the intermittent wiper is used to wipe off the rainwater on the front windshield.

[0077] The application also provides an ADAS camera self-cleaning system, comprising An image acquisition module is used to acquire images of the front windshield in front of the ADAS camera in real time during the driving process of the vehicle; the function realized by the image acquisition module can be directly realized by the ADAS camera, or realized by another arranged camera; A blur degree identification module is used to identify the acquired images based on the constructed blur degree identification model, and determine the image blur degree; A self-cleaning module is used to adopt a corresponding self-cleaning mode to clean the front windshield in front of the ADAS camera according to the image blur degree.

[0078] The functions implemented by the above-mentioned modules have been described in detail in the aforementioned analysis method, and will not be repeated here.

[0079] It should be understood that the particular order or hierarchy of steps in the processes disclosed is an example that can be re-arranged as desired. Based on the disclosure and teachings provided herein, a person of ordinary skill in the art can contemplate other steps and / or related steps not mentioned in this disclosure while still staying within the scope of the present disclosure. The various steps of the appended method claims can be presented in an example order of operation, but the ordering of certain steps can be rearranged while remaining within the scope of the present disclosure.

[0080] The above-described embodiments of the present disclosure have been described in order to explain the present disclosure and not in order to limit it. Numerous modifications and changes can be made by persons of ordinary skill in the art without departing from the scope of the present disclosure, the scope of which is defined by the appended claims. For example, while the embodiments of the present disclosure have been described with reference to specific examples, which are intended to be illustrative only and not to be limiting of the scope of the application, a variety of specific modifications and adaptations of the specific examples can be used to implement the embodiments of the present disclosure without departing from the scope of the present disclosure.

[0081] Those of skill would further appreciate that the various illustrative logical blocks, modules, circuits, and steps described in connection with the embodiments disclosed herein can be implemented as electronic hardware, computer software, or combinations of both. To clearly illustrate this interchangeability of hardware and software, various illustrative components, blocks, modules, circuits, and steps have been described above generally in terms of their functionality. Whether such functionality is implemented as hardware or software depends upon the particular application and design constraints imposed on the overall system. Skilled artisans can implement the described functionality in varying ways for each particular application, but such implementation decisions should not be interpreted as causing a departure from the scope of the present embodiments.

[0082] The various illustrative logical blocks, modules, and circuits described in connection with the embodiments disclosed herein can be implemented or performed with a general purpose processor, a digital signal processor, an application specific integrated circuit (ASIC), a field programmable gate array (FPGA) or other programmable logic device, discrete gate or transistor logic, discrete hardware components, or any combination thereof designed to perform the functions described herein. A general purpose processor can be a microprocessor, but in the alternative, the general purpose processor can be any conventional processor, controller, microcontroller, or state machine. A processor can also be implemented as a combination of computing devices, e.g., a combination of a digital signal processor and a microprocessor, a plurality of microprocessors, one or more microprocessors in conjunction with a digital signal processor core, or any other such configuration.

[0083] The above are merely specific embodiments of the present invention, but the scope of protection of the present invention is not limited thereto. Any variations or substitutions that can be easily conceived by those skilled in the art within the technical scope disclosed in the present invention should be included within the scope of protection of the present invention. Contents not described in detail in this specification belong to prior art known to those skilled in the art.

Claims

1. A self-cleaning method for the windshield in front of an ADAS camera, characterized in that: During vehicle operation, based on the constructed fuzziness recognition model, the fuzziness level of the collected images of the windshield from the ADAS camera is identified. Based on the degree of blurriness of the identified image, the corresponding self-cleaning mode is used to self-clean the windshield in front of the ADAS camera.

2. The self-cleaning method for the windshield in front of an ADAS camera according to claim 1, characterized in that, The construction process of the fuzziness recognition model is as follows: Obtain training and test sets with different levels of fuzziness; Choose a pre-trained ResNet model as the base model; The basic model is trained using the training set, and the trained model is tested using the test set to obtain an end-to-end fuzziness recognition model.

3. The self-cleaning method for the windshield in front of an ADAS camera according to claim 2, characterized in that: Several images of dirty windshields in front of ADAS cameras were collected. These images were then divided into multiple blur levels based on their degree of blur. The images were then enhanced to form an image dataset with clear and blurred labels. The image dataset was then divided into training and testing sets according to a set ratio.

4. The self-cleaning method for the windshield in front of an ADAS camera according to claim 3, characterized in that: The multiple fuzziness levels include: First level: The degree of ambiguity is less than A1; Second level: The degree of ambiguity is greater than or equal to A1 and less than A2; Third level: The degree of ambiguity is greater than or equal to A2 and less than A3; Fourth level: The degree of ambiguity is greater than A3; A1, A2, and A3 represent the percentages of the first, second, and third degrees of ambiguity, respectively.

5. The self-cleaning method for the windshield in front of an ADAS camera according to claim 1, characterized in that: The step of employing a corresponding self-cleaning mode based on the degree of blurriness of the identified image includes: If the detected ambiguity is less than A1, the first self-cleaning mode is used; If the degree of ambiguity identified is greater than or equal to A1 and less than A2, then the second self-cleaning mode is adopted; If the detected level of ambiguity is greater than or equal to A2 and less than A3, then the third self-cleaning mode is used; If the detected blur level is greater than A3, then the fourth self-cleaning mode is used; A1, A2, and A3 represent the percentages of the first, second, and third degrees of ambiguity, respectively.

6. The self-cleaning method for the windshield in front of an ADAS camera according to claim 5, characterized in that: The first self-cleaning mode is to exit the automatic self-cleaning mode. The second self-cleaning mode is that the wipers work intermittently at set time intervals; The third self-cleaning mode is that the wipers operate continuously at a set first frequency. The fourth self-cleaning mode is that the wipers operate continuously at a set second frequency; The second frequency is greater than the first frequency.

7. The self-cleaning method for the windshield in front of an ADAS camera according to claim 6, characterized in that: When the windshield wipers operate continuously at a set second frequency, the spraying of cleaning fluid is controlled.

8. The self-cleaning method for the windshield in front of an ADAS camera according to claim 1, characterized in that: During the self-cleaning process of the windshield in front of the ADAS camera, if the blur level is detected to decrease to below the lower threshold of the corresponding blur level, the timer will start from the point when the blur level drops to the threshold. After the timer is set, the self-cleaning mode will switch to a lower level for self-cleaning.

9. The self-cleaning method for the windshield in front of an ADAS camera according to claim 8, characterized in that: If the ambiguity level is detected to decrease below the lower threshold of the corresponding ambiguity level, a timer is started from the point when the ambiguity level drops to the threshold. After the timer has been set, the system switches to a lower-level self-cleaning mode for self-cleaning, including: If the fuzziness level corresponding to the fuzziness level identified at the current moment is i, then the self-cleaning mode adopted is the j-th self-cleaning mode; If, at a certain point after a period of self-cleaning, the detected level of ambiguity is lower than the threshold of the ambiguity level of the i-th level, then a timer is started. After the timer is set, the self-cleaning mode is switched from the j-th self-cleaning mode to the (j-1)-th self-cleaning mode. Where i≥2, j≥2.

10. A self-cleaning system for the windshield in front of an ADAS camera, characterized in that: include The image acquisition module is used to acquire images of the windshield in front of the ADAS camera in real time while the vehicle is in motion; The blur level recognition module is used to identify the acquired images based on the constructed blur level recognition model and determine the blur level of the images; The self-cleaning module is used to clean the windshield in front of the ADAS camera by adopting the corresponding self-cleaning mode according to the degree of image blur.

Images