Wiper blade life prediction method and vehicle

Abstract

The application relates to the technical field of vehicle electronic control, and provides a wiper blade life prediction method and a vehicle. The wiper blade life prediction method comprises the following steps: acquiring the running current of a driving motor of a wiper and the image of a glass wiped by a wiper blade of the wiper under the condition that the wiper in the vehicle is in a preset working condition; determining the wear degree of the wiper blade according to the running current of the driving motor and the image of the glass; and predicting the residual life of the wiper blade based on the wear degrees of the wiper blade determined for multiple times within a preset prediction period. The wiper blade life prediction method can improve the prediction accuracy of the residual life of the wiper blade by jointly predicting the residual life of the wiper blade through the running current of the driving motor and the image of the wiped glass.

Description

Technical Field

[0001] This application relates to the field of vehicle electronic control technology, specifically to a method for predicting the lifespan of windshield wipers and a vehicle. Background Technology

[0002] As a device used in vehicles to remove rain, snow, dust, mud and other stains from the windshield, the wiper blades are susceptible to wear and tear due to natural aging, environmental erosion and other factors.

[0003] Severely worn windshield wiper blades can lead to water leakage and residue buildup on the windshield, making it difficult to clean. If these issues aren't addressed and replaced promptly, they can cause blurred vision and obstructed visibility in rainy weather, significantly increasing driving safety hazards. Therefore, it's necessary to predict the remaining lifespan of windshield wiper blades. However, current technologies for predicting the remaining lifespan of wiper blades have low accuracy, hindering timely replacement and impacting the overall performance of the wiper system. Summary of the Invention

[0004] In view of this, this application aims to propose a method for predicting the lifespan of windshield wipers, so as to improve the accuracy of predicting the remaining lifespan of windshield wipers.

[0005] To achieve the above objectives, the technical solution of this application is implemented as follows: A method for predicting wiper blade life, applied to vehicles, includes: When the windshield wipers in the vehicle are in a preset working condition, the operating current of the drive motor of the windshield wipers and the image of the glass after being wiped by the wiper blades are acquired. The degree of wear of the wiper blade is determined based on the operating current of the drive motor and the glass image. The remaining lifespan of the wiper blade is predicted based on the wear level of the wiper blade determined multiple times within a preset prediction period.

[0006] Furthermore, based on the operating current of the drive motor and the glass image, the wear degree of the wiper blade is determined, including: Based on the operating current of the drive motor, the harmonic energy distribution parameters of the operating current of the drive motor are calculated, and the visual clarity index of the glass after the wiper blade has wiped is determined based on the glass image. The visual acuity index and the harmonic energy distribution parameters are fused to obtain the fused feature; The wear degree of the wiper blade is determined based on the fusion characteristics.

[0007] Furthermore, the step of calculating the harmonic energy distribution parameters of the operating current of the drive motor based on the operating current of the drive motor includes: The power frequency and fundamental frequency components in the operating current of the drive motor are filtered to obtain the load harmonic current signal. The load harmonic current signal is subjected to a fast Fourier transform to obtain the spectrum of the load harmonic current signal; Each frequency point within a preset high-frequency band is determined from the spectrum, and the total harmonic energy within the preset high-frequency band is calculated based on the amplitude corresponding to each determined frequency point. The harmonic energy distribution parameters are calculated based on the total harmonic energy and the amplitude corresponding to each frequency point.

[0008] Furthermore, the step of calculating the harmonic energy distribution parameters based on the total harmonic energy and the amplitude corresponding to each frequency point includes: Based on the amplitude corresponding to each frequency point, calculate the harmonic energy corresponding to each frequency point respectively; Determine the proportion of harmonic energy corresponding to each frequency point to the total harmonic energy; Based on the respective energy proportions, the harmonic entropy of the spectrum is calculated and used as the harmonic energy distribution parameter.

[0009] Furthermore, determining the visual clarity index of the glass after the wiper blade has wiped the glass based on the glass image includes: The glass image is preprocessed to obtain a feature image; Based on the feature image, determine the visual clarity index of the glass after the wiper blade has wiped it.

[0010] Furthermore, the preprocessing of the glass image to obtain a feature image includes: The glass image is subjected to grayscale conversion and contrast enhancement processing; Obtain a preset background image, and use the preset background image to perform background removal processing on the processed glass image to obtain the feature image; and / or, Determining the visual clarity index of the glass after the wiper blade has wiped the windshield based on the feature image includes: Identify the water ripple features present in the feature image; Based on the water ripple characteristics, the visual clarity index of the glass after being wiped by the wiper blade is determined.

[0011] Furthermore, the prediction of the remaining lifespan of the wiper blade based on the wear degree of the wiper blade determined multiple times within a preset prediction period includes: Based on the wear degree of the wiper blade determined in each of the preset prediction periods, a wear degree change curve is obtained by fitting. The remaining lifespan of the wiper blade is predicted based on the wear variation curve.

[0012] Furthermore, the method also includes: Based on the currently determined wear level of the wiper blade and the predicted remaining lifespan of the wiper blade, determine the current replacement warning level for the wiper blade; Based on the current replacement warning level corresponding to the wiper blade, a preset wiper blade replacement prompt message is output to the vehicle.

[0013] Furthermore, the preset operating conditions include the windshield wipers being in a preset gear and the vehicle speed being within a preset speed range.

[0014] Compared with related technologies, this application has at least the following advantages: (1) The wiper blade life prediction method described in this application first determines the wear degree of the wiper blade in the current state by using the operating current of the drive motor and the glass image. Then, based on the wear degree in multiple states determined within a preset prediction period, the remaining life of the wiper blade is predicted. This method of predicting the remaining life of the wiper blade does not use a single physical quantity analysis to determine the current state of the wiper blade, but rather uses a comprehensive analysis of two physical quantities: the operating current and the glass image. This can improve the accuracy of determining the current wear degree of the wiper blade. Therefore, when using the wear degree of the wiper blade to predict the remaining life of the wiper blade, the change in the wear degree of the wiper blade can be analyzed more accurately, thereby improving the accuracy of the remaining life of the wiper blade.

[0015] (2) By fusing the visual clarity index with the harmonic energy distribution parameter to obtain a fused feature, the wear degree of the wiper blade is determined using this fused feature. This allows for the simultaneous use of both electrical and visual features to determine the wear degree of the wiper blade, thereby improving the stability and accuracy of the wear degree determination. Furthermore, this application uses the harmonic energy distribution parameter to characterize the motor's operating characteristics for determining the wear degree of the wiper blade. This accurately captures the characteristics of motor current harmonic changes caused by wiper blade friction disturbance during the aging process of the wiper blade, further enhancing the ability to identify the true wear state of the wiper blade and effectively improving the accuracy of the wear degree determination.

[0016] (3) By first filtering out the power frequency and fundamental components in the operating current when calculating the harmonic energy distribution parameters, and then performing frequency domain conversion and extracting the harmonic energy distribution parameters, the current components (i.e., power frequency and fundamental components) that are not related to wiper blade wear can be filtered out, and only the effective components are retained, thereby improving the accuracy of the calculation of harmonic energy distribution parameters.

[0017] (4) The harmonic entropy of the spectrum is determined by the proportion of harmonic energy at each frequency point to the total harmonic energy in order to obtain the harmonic energy distribution parameter. The wear degree of the wiper blade can then be determined by the harmonic energy distribution parameter, so as to accurately predict the remaining life of the wiper blade.

[0018] (5) By preprocessing the glass image before extracting the visual acuity index, and then determining the visual acuity index of the glass based on the preprocessed image (i.e., the feature image), the influence of factors such as ambient light on the visual acuity index can be reduced, thereby improving the accuracy of the visual acuity index determination.

[0019] (6) The visual sharpness index of the glass after the wiper blade has wiped is determined by extracting water ripple features from the feature image. Since water ripple features reflect the residual water ripples on the glass after the wiper blade has wiped, the visual sharpness index determined by these water ripple features is more suitable for judging the wear degree of the wiper blade. That is, the wear degree of the wiper blade can be determined more accurately by using such a visual sharpness index, thereby improving the accuracy of the wear degree determination.

[0020] (7) By fitting the wear degree change curve of the wiper blade determined each time within the preset prediction period, the remaining life of the wiper blade can be predicted using the wear degree change curve. Since the wear degree change curve can reflect the change trend of the wear degree of the wiper blade after each wipe, the aging law of the wiper blade characterized by the wear degree at multiple historical moments can be used to predict the remaining life, thus avoiding the blind prediction based on the wear degree of a single instance, thereby improving the accuracy of life prediction.

[0021] (8) Based on the currently determined wear level and predicted remaining life of the wiper blades, corresponding wiper blade replacement reminders are output to the vehicle in a tiered manner to provide graded warnings based on the actual wear status and remaining aging trend. This can avoid premature and unnecessary reminders that may cause user inconvenience, and also prevent the problem of delayed warnings leading to wiper failure and affecting driving visibility and safety.

[0022] (9) By collecting operating current and glass images when the wipers are in a preset setting and the vehicle speed is within a preset speed range, the wear level and remaining life can be determined. This allows the operating current and glass images to be collected when the wiper load is stable and the wind disturbance is minimal, rather than collecting them under any conditions. This ensures that the collected operating current and glass images can accurately reflect the friction load characteristics of the wiper blades and the residual characteristics of the wiping water patterns, thereby avoiding interference from the operating conditions and improving the accuracy of subsequent wear level determination and life prediction.

[0023] Another object of this application is to provide a vehicle whose controller includes: Memory, used to store computer programs; A processor is used to execute a computer program stored in the memory to implement the aforementioned wiper blade life prediction method.

[0024] The vehicle described in this application can predict the remaining life of the wiper blades by combining the operating current of the drive motor with the image of the wiped glass, rather than by directly predicting it using a single physical quantity. This can improve the accuracy of the prediction of the remaining life of the wiper blades. Attached Figure Description

[0025] The accompanying drawings, which form part of this application, are used to provide a further understanding of this application. The illustrative embodiments and descriptions of this application are used to explain this application and do not constitute an undue limitation of this application. In the drawings: Figure 1 This is a flowchart illustrating the wiper blade life prediction method described in the embodiments of this application; Figure 2 This is a schematic diagram of the process for determining the degree of wear in the wiper blade life prediction method described in the embodiments of this application; Figure 3 This is a schematic diagram of the process for calculating harmonic energy distribution parameters in the wiper blade life prediction method described in the embodiments of this application; Figure 4 This is a schematic diagram illustrating the process of predicting the remaining lifespan of a wiper blade in the wiper blade lifespan prediction method described in this application embodiment. Figure 5 This is a schematic diagram of the configuration of the vehicle controller described in the embodiments of this application.

[0026] Explanation of reference numerals in the attached figures: 510, processor; 520, memory. Detailed Implementation

[0027] To make the technical solution and advantages of this application clearer, the following detailed description is provided in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative and not intended to limit the scope of this application.

[0028] It should be noted that, unless otherwise specified, the embodiments and features described in this application can be combined with each other.

[0029] Furthermore, in the description of this application, the terms "first" and "second" are used for descriptive purposes only and should not be construed as indicating or implying relative importance.

[0030] In this application, the terms "one embodiment," "some embodiments," "example," "specific example," or "some examples," etc., refer to a specific feature, structure, material, or characteristic described in connection with that embodiment or example, which is included in at least one embodiment or example of this application. In this specification, the illustrative expressions of the above terms do not necessarily refer to the same embodiment or example. Moreover, the specific features, structures, materials, or characteristics described may be combined in any suitable manner in one or more embodiments or examples.

[0031] The present application will now be described in detail through exemplary embodiments. However, it should be understood that, without further description, elements, structures, and features in one embodiment may be advantageously incorporated into other embodiments.

[0032] An embodiment of the first aspect of this application provides a method for predicting the lifespan of a windshield wiper blade. This method is applied to a vehicle and determines the degree of wear of the wiper blade by combining the operating current of the wiper drive motor and the glass image. Then, based on the degree of wear of the wiper blade within a preset prediction period, the remaining lifespan of the wiper blade is predicted to improve the accuracy of the prediction of the remaining lifespan of the wiper blade.

[0033] In related technologies, windshield wipers (also known as wipers, windshield wipers, etc.) use their wiper blades to wipe away rain, snow, dust, mud and other dirt adhering to the windshield of a vehicle, thereby improving the driver's visibility and increasing driving safety.

[0034] Currently, the windshield wipers installed on vehicles generally consist of wiper blades (usually made of rubber strips, which are the parts that come into direct contact with the glass to remove dirt), linkage mechanisms, drive motors, return springs, and other structures. During long-term use, due to the combined effects of multiple factors such as natural aging, environmental erosion, mechanical wear, and ultraviolet radiation, the wiper blades are prone to hardening and cracking, edge deformation, and other wear, which can lead to problems such as water leakage and residue, or even incomplete cleaning of dirt when wiping the glass.

[0035] As a critical safety component of a vehicle, the windshield wipers directly determine the clarity of the driver's vision in rainy weather. If severely worn wiper blades are not detected and replaced in time, it will lead to blurred vision and obstructed visibility, significantly increasing driving safety hazards. Furthermore, wiper blade wear is a gradual process, and drivers and passengers cannot visually assess the actual wear to determine its remaining lifespan. This can easily result in premature replacement, leading to waste, or delayed replacement, causing safety risks.

[0036] Therefore, it is necessary to predict the remaining life of the wiper blades on the vehicle so that the driver and passengers can be reminded to replace the wiper blades in time when the wiper blades are severely worn and have insufficient remaining life.

[0037] In related technologies, the prediction of the remaining life of wiper blades is mostly based on the degree of dirt (or water) residue on the glass image after wiping, which determines the current wear state of the wiper blades and thus predicts their remaining life.

[0038] However, this method of predicting wiper blade life using a single physical quantity is prone to misjudging or missing the current wear condition of the wiper blades when judging their current state, thus affecting the accuracy of the wiper blade life prediction.

[0039] In view of this, in order to overcome the shortcomings of related technologies, the wiper blade life prediction method in this embodiment combines... Figure 1 In terms of overall design, it includes the following steps S110-S130.

[0040] Step S110: When the windshield wipers in the vehicle are in a preset working condition, acquire the operating current of the wiper drive motor and the image of the glass after being wiped by the wiper blades.

[0041] The preset operating condition refers to the condition in which the wipers are in a stable wiping state. When the wipers are in the preset operating condition, it means that the wipers are currently working. In order to predict the remaining life of the wiper blades, it is necessary to first determine the current wear level of the wiper blades.

[0042] Therefore, the operating current of the wiper drive motor and the image of the glass after being wiped by the wiper blade are first obtained, so that the subsequent step S120 can use the operating current and the glass image to determine the current wear level of the wiper blade.

[0043] Specifically, the drive motor is the motor that drives the windshield wipers to perform reciprocating wiping motions. The operating current of the drive motor refers to the current that the drive motor operates in real time during the reciprocating wiping process. It is worth noting that when the drive motor is a three-phase drive motor, the operating current refers to the phase current of the three-phase drive motor. Under the preset operating conditions of the windshield wipers, the three-phase currents Ia, Ib, and Ic of the three-phase drive motor are synchronously collected at a preset sampling rate (e.g., 20Hz) and maintained for at least two complete wiping cycles.

[0044] Each wiper action (starting from the initial state, completing one wipe, and returning to the initial state indicates the completion of one wipe action) represents one wiping cycle.

[0045] The image of the glass after the wiper blades have wiped the windshield can be obtained by controlling the camera to capture an image of the preset area of ​​interest after the wiper blades have moved upwards and passed through it. This preset area of ​​interest can be the driver's main field of vision, or it can be determined based on the camera's installation position. For example, images captured closer to the camera are less susceptible to external interference; therefore, the area directly in front of the camera can also be chosen as the preset area of ​​interest, without limitation. It is also worth noting that the camera can be installed outside the windshield, allowing it to directly capture images of the windshield.

[0046] Step S120: Determine the degree of wear of the wiper blades based on the operating current of the drive motor and the glass image.

[0047] Step S130: Based on the wear degree of the wiper blade determined multiple times within a preset prediction period, predict the remaining life of the wiper blade.

[0048] Specifically, in predicting the remaining lifespan of the wiper blade, the wear level of the wiper blade in the current state is determined in step S120 based on the operating current of the drive motor and the glass image. Then, in step S130, the wear level determined by multiple diagnoses within a preset prediction period is analyzed to analyze the changing trend of the wear level of the wiper blade, thereby predicting the future state of the wiper blade under the wear level and thus predicting the remaining lifespan of the wiper blade.

[0049] Through steps S110-S130, the wear level of the wiper blade in its current state is first determined by the operating current of the drive motor and the glass image. Then, based on the wear levels in multiple states determined within a preset prediction period, the remaining lifespan of the wiper blade is predicted. This method of predicting the remaining lifespan of the wiper blade does not use a single physical quantity to analyze and determine the current state of the wiper blade, but rather uses two physical quantities—operating current and glass image—to comprehensively analyze and determine the remaining lifespan. This improves the accuracy of determining the current wear level of the wiper blade. Therefore, when using the wear level of the wiper blade to predict its remaining lifespan, the changes in the wear level of the wiper blade can be analyzed more accurately, thereby improving the accuracy of the remaining lifespan of the wiper blade. This allows users to replace the wiper blades in a timely manner, thus improving the quality of the wiper system.

[0050] In some exemplary embodiments, the preset working condition in step S110 may specifically include: the windshield wipers are in a preset gear and the vehicle speed is within a preset speed range.

[0051] The preset setting refers to a low-speed working setting that can form a continuous and uniform thin water film on the glass surface. For example, the preset setting can be an intermittent setting (meaning that the wiper blade completes a single reciprocating wiping at a fixed time interval) or a low-speed continuous setting (meaning that the wiper blade wips back and forth at a constant low speed).

[0052] The preset speed range refers to the low-speed range that can reduce the interference of wind noise on the wiper blades and windshield washer fluid film. For example, the preset speed range can be: below a preset vehicle speed threshold, which is determined based on actual vehicle wind noise test calibration, for example, it can be set to 60km / h.

[0053] Therefore, when the windshield wipers are in a preset setting and the vehicle speed is within a preset speed range, the windshield wipers are determined to be in that preset operating condition. Then, step S110 is executed to collect the operating current and glass image. This allows for the collection of operating current and glass images only when a natural, uniform thin water film forms on the glass, the wiper load is stable, and wind disturbance is minimal, rather than collecting them under all conditions. This ensures that the collected operating current and glass image accurately reflect the wiper blade friction load characteristics and the residual water ripples, thereby avoiding interference from operating conditions and improving the accuracy of subsequent wear assessment and lifespan prediction.

[0054] It is worth noting that if the windshield wipers are not in the preset gear or the vehicle speed is not in the preset speed range, it means that the windshield wipers are not in the preset working condition and the above steps S110-S130 will not be executed.

[0055] It is worth noting that, considering the significant load disturbance caused by insufficient glass lubrication during dry wiping, which can interfere with the determination of wear level by affecting the characteristics of the drive motor's operating current, in this embodiment, the preset working condition not only includes the wipers being in a preset gear and the vehicle speed being within a preset speed range, but also requires the user to trigger the windshield washer fluid spraying command simultaneously. That is, in this embodiment, the wiper blades must operate with windshield washer fluid assistance lubrication, the wipers must be in a preset gear, and the vehicle speed must be within a preset speed range before executing steps S110-S130.

[0056] Continuing from the above Figure 1 and combined Figure 2 As shown, in some exemplary embodiments, the determination of the wear degree of the wiper blade in step S120 based on the operating current of the drive motor and the glass image may specifically include steps S121-S123 below.

[0057] Step S121: Calculate the harmonic energy distribution parameters of the driving motor's operating current based on the driving motor's operating current, and determine the visual clarity index of the glass after the wiper blade has wiped the glass based on the glass image.

[0058] Specifically, as wiper blades wear down, the dynamic characteristics of the friction coefficient between the wiper blade and the glass change. Since the friction state directly determines the load on the drive motor, this change directly translates into load torque pulsation in the wiper drive motor. This torque pulsation modulates the motor's operating current, thereby generating harmonic components in a specific frequency band within the drive motor's operating current. These harmonic components refer to the high-frequency current fluctuations superimposed on the motor's fundamental operating current, induced by wiper blade friction impact and torque pulsation (in this embodiment, high frequency refers to frequencies in the 500 to 2000 Hz range). The more severe the wiper blade wear and the more intense the friction impact between the wiper blade and the glass, the stronger the high-frequency harmonic energy, resulting in a more dispersed spectral energy distribution.

[0059] Therefore, in step S120, when determining the wear level of the wiper blade, the harmonic energy distribution parameters in the operating current of the drive motor can be identified (quantifying the energy dispersion of the harmonics of the operating current at each frequency point, especially the energy dispersion of high-frequency harmonics), so as to identify the wear level of the wiper blade through the harmonic energy distribution parameters.

[0060] In addition, when the wiper blades wear down, the contact pressure between the wiper blades and the glass becomes uneven. As a result, after the wiper blades wipe, they cannot form a uniform water film on the glass surface. Instead, they leave behind strip-shaped or arc-shaped water patterns or tiny water droplets in the wiping area.

[0061] Therefore, in step S120, when determining the wear level of the wiper blade, the visual clarity index of the glass after the wiper blade has wiped can be determined based on the glass image, so as to identify and determine the wear level of the wiper blade through the visual clarity index.

[0062] Step S122: Fuse the visual clarity index with the harmonic energy distribution parameters to obtain the fused features.

[0063] Step S123: Determine the wear degree of the wiper blades based on the fusion characteristics.

[0064] Specifically, determining the wear level of wiper blades solely based on the visual sharpness index extracted from the glass image is easily affected by changes in lighting, local water stains, and slight camera shake, resulting in insufficient robustness of a single visual feature. On the other hand, determining the wear level of wiper blades solely based on the harmonic energy distribution parameters extracted from the operating current is easily affected by mechanical assembly clearances and slight disturbances due to individual differences in the motor itself, resulting in occasional deviations in electrical characteristics.

[0065] Therefore, to improve the accuracy of wear degree and thus the accuracy of lifespan prediction, in steps S122 and S123 of this embodiment, the visual clarity index and harmonic energy distribution parameters are fused to obtain a fused feature. The wear degree of the wiper blade is then determined based on this fused feature. This allows for the simultaneous use of both electrical and visual features to determine the wiper blade wear degree, thereby improving the stability and accuracy of wear degree determination. Furthermore, this application uses harmonic energy distribution parameters to characterize the motor's operating characteristics for determining the wiper blade wear degree. This accurately captures the motor current harmonic changes caused by wiper blade friction disturbance during the wiper blade wear and aging process, further enhancing the ability to identify the true wear state of the wiper blade and effectively improving the accuracy of wear degree determination.

[0066] Continuing from the above Figures 1 to 2 and combined Figure 3 As shown, in some exemplary embodiments, in step S121 above, the harmonic energy distribution parameters of the operating current of the drive motor are calculated based on the operating current of the drive motor, which may specifically include the following steps S1211-S1214.

[0067] Step S1211: Filter the power frequency and fundamental frequency components in the operating current of the drive motor to obtain the load harmonic current signal.

[0068] Specifically, the high-frequency current component caused by load torque fluctuations in the operating current of the drive motor is the effective component representing the degree of wiper blade wear. However, the operating current includes not only the current component caused by the load torque but also the power frequency and fundamental frequency components. These components are mainly composed of the power grid frequency and the inherent current of the motor during steady-state uniform operation, and are unrelated to wiper blade wear, thus belonging to invalid background components. Therefore, when calculating the harmonic energy distribution parameters in step S121, the power frequency and fundamental frequency components in the operating current are first filtered out in step S1211.

[0069] Among them, the power frequency and fundamental frequency components refer to the 50Hz power frequency component of civil power supply and the low-frequency fundamental frequency current component corresponding to the steady-state rotation of the motor itself.

[0070] In step S1211, filtering the power frequency and fundamental frequency components can specifically involve inputting the operating current to a bandpass filter to filter out current components that are not within the passband frequency range of the bandpass filter. By setting the passband frequency range of the bandpass filter, the power frequency and fundamental frequency components in the operating current can be filtered out. Considering that the power frequency is 50Hz, the passband frequency range of the bandpass filter can be set, for example, to 50Hz-2500Hz.

[0071] Step S1212: Perform a fast Fourier transform on the load harmonic current signal to obtain the spectrum of the load harmonic current signal.

[0072] Specifically, the original signal of the motor operating current is a time-domain signal, which can only observe the change in current magnitude over time and cannot intuitively distinguish the harmonic energy distribution characteristics at different frequencies. However, the torque pulsation disturbance induced by wiper blade wear is reflected in the energy difference of a specific high-frequency band. Therefore, it is necessary to use Fast Fourier Transform to convert the time-domain operating current into a frequency-domain spectrum signal in order to decompose the harmonic amplitude and energy corresponding to different frequencies, and provide a frequency domain data foundation for subsequent extraction of high-frequency band characteristics and calculation of total harmonic energy and distribution parameters.

[0073] Among them, the Fast Fourier Transform (FFT) is a mathematical transformation method that rapidly converts discrete-time signals into frequency-domain signals, enabling the decomposition and analysis of current from the time dimension to the frequency dimension.

[0074] In step S1212, the load harmonic current signal obtained in step S1211 is subjected to a fast Fourier transform to obtain the spectrum of the load harmonic current signal. Then, steps S1213 and S1214 can be executed to determine the harmonic energy distribution parameters using the spectrum.

[0075] Step S1213: Determine each frequency point in the preset high frequency band from the spectrum, and calculate the total harmonic energy in the preset high frequency band based on the amplitude corresponding to each determined frequency point.

[0076] The preset high-frequency band is a pre-defined high-frequency band that corresponds to the frequency band of torque pulsation characteristics induced by wiper blade friction impact and contact vibration. The harmonic amplitude and energy of each frequency point within the preset high-frequency band are directly related to the degree of wiper blade wear. Therefore, by analyzing the energy and distribution of each frequency point (referred to as high-frequency point in this embodiment) within the preset high-frequency band, the degree of wiper blade wear can be determined. In some embodiments, the preset high-frequency band can be 500 to 2000 Hz.

[0077] The total harmonic energy within the preset high-frequency band refers to the sum of the harmonic energies at all frequency points within the preset high-frequency band, representing the overall strength level of the disturbance.

[0078] Specifically, the degree of wear is reflected in the dispersion of energy distribution among various high-frequency points. The total energy alone can only reflect the strength of the disturbance and cannot characterize the regularity of the spectral distribution. Therefore, it is necessary to first calculate the total harmonic energy of the high-frequency points within the preset high-frequency band, and then statistically obtain the harmonic energy distribution parameters that can characterize the wear characteristics by using the proportion of energy at each frequency point to the total harmonic energy of all high-frequency points. Therefore, in this embodiment, the total harmonic energy within the preset high-frequency band is calculated first in step S1213.

[0079] Specifically, the spectrum obtained by the Fast Fourier Transform (FFT) consists of multiple frequency points. These frequency points within the preset high-frequency band are the frequency points determined in this embodiment, also known as the aforementioned high-frequency points, which are used to calculate the harmonic energy distribution parameters. In step S1213, the FFT result corresponding to each high-frequency point is obtained. This FFT result is a complex number; the modulus of this complex number is calculated to obtain the amplitude corresponding to that high-frequency point. Then, based on the amplitudes corresponding to each high-frequency point, the total harmonic energy corresponding to the preset high-frequency band is calculated.

[0080] Wherein, the total harmonic energy Etotal=Σ(A i ²) = A1² + A2² + ... + A N ².

[0081] Where i takes values ​​from 1 to N, A i Let N be the amplitude of the i-th high-frequency point, and N be the number of high-frequency points.

[0082] Step S1214: Calculate the harmonic energy distribution parameters based on the total harmonic energy and the amplitude corresponding to each frequency point.

[0083] Specifically, the frequency points in step S1214 are the frequency points located in the preset high-frequency band determined in step S1212, which are the aforementioned high-frequency points. After calculating the total harmonic energy in step S1213, the harmonic energy distribution parameters can be calculated in step S1214 based on the total harmonic energy and the amplitude corresponding to each high-frequency point.

[0084] Therefore, through the above steps S1211-S1214, the power frequency and fundamental components of the driving motor's operating current are filtered out, time-frequency domain conversion is performed, and high-frequency band energy statistics and distribution characteristics are quantified, thereby accurately extracting the harmonic energy distribution characteristics strongly correlated with wiper blade wear, which can improve the accuracy of subsequent wear degree judgment.

[0085] Continuing from the above Figures 1 to 3 As shown, in step S1214 above, the harmonic energy distribution parameters are calculated based on the total harmonic energy and the amplitude corresponding to each frequency point. Specifically, this may include: calculating the harmonic energy corresponding to each frequency point based on the amplitude corresponding to each frequency point; determining the energy proportion of the harmonic energy corresponding to each frequency point in the total harmonic energy; and calculating the harmonic entropy of the spectrum based on each energy proportion, which serves as the harmonic energy distribution parameter.

[0086] Specifically, in step S1214 above, the amplitude A of each frequency point within the preset high-frequency band is first determined. i Calculate the harmonic energy A corresponding to each high-frequency point. i2 Each frequency point corresponds to a harmonic energy.

[0087] Then calculate the harmonic energy A corresponding to each high-frequency point. i 2 The proportion of the total harmonic energy Etotal, that is, the energy proportion P corresponding to each high-frequency point. i Specifically, P i =A i 2 / Etotal. Each high-frequency point corresponds to a specific energy percentage.

[0088] Then, based on the energy percentage P corresponding to each high-frequency point i Calculate the harmonic entropy Hcurrent. This harmonic entropy Hcurrent is the energy distribution parameter of the harmonic. Wherein, the harmonic entropy Hcurrent = -Σ(P i ×log2(P i )).

[0089] The higher the wear of the wiper blade, the more irregular the friction between the wiper blade and the glass becomes, which makes the high-frequency harmonic energy distribution more scattered, thus increasing the harmonic entropy. Therefore, the wear level of the wiper blade can be determined by the harmonic entropy (harmonic energy distribution parameter).

[0090] Therefore, the harmonic entropy of the spectrum is determined by the proportion of harmonic energy at each frequency point to the total harmonic energy, so as to obtain the harmonic energy distribution parameter. This harmonic energy distribution parameter can then be used to determine the wear degree of the wiper blade, so as to accurately predict the remaining life of the wiper blade.

[0091] Continue by Figures 1 to 3 As shown, in some exemplary embodiments, after determining the harmonic energy distribution parameters using the operating current, it is also necessary to determine the visual acuity index using the glass image before the harmonic energy distribution parameters and the visual acuity index can be fused to obtain a fused feature for identifying the degree of wear.

[0092] Specifically, in step S121 above, determining the visual sharpness index of the glass after the wiper blade has wiped the windshield based on the glass image may include: preprocessing the glass image to obtain a feature image; and determining the visual sharpness index of the glass after the wiper blade has wiped the windshield based on the feature image.

[0093] Specifically, when determining the visual sharpness index of the glass after the wiper blade has wiped the glass, it is important to consider that directly using the glass image as a feature image for visual sharpness index recognition may result in errors due to factors such as ambient light.

[0094] Therefore, in this embodiment, the glass image is first preprocessed, and then the visual acuity index of the glass is determined based on the preprocessed image (i.e., the feature image). This reduces the influence of factors such as ambient light on the visual acuity index, thereby improving the accuracy of the visual acuity index determination.

[0095] Continue by Figures 1 to 3 As shown, in some exemplary embodiments, the preprocessing of the glass image to obtain a feature image may specifically include: performing grayscale conversion and contrast enhancement processing on the glass image; obtaining a preset background image; and using the preset background image to perform background removal processing on the processed glass image to obtain the feature image.

[0096] Specifically, the grayscale conversion and contrast enhancement processing of the glass image includes: first, converting the acquired colored glass image to grayscale to remove color redundancy interference and reduce computational complexity. Then, histogram equalization and contrast enhancement processing is applied to the grayscale image to enhance the grayscale differences between the water ripples and watermarks on the glass surface and the background area, while mitigating interference from gradual changes in ambient light.

[0097] After grayscale conversion and contrast enhancement, the processed glass image can be directly used as a feature image for subsequent processing, or the processed glass image can be first processed to remove the background, and then the background-removed image can be used as a feature image for subsequent processing.

[0098] Specifically, the background removal process for the processed glass image can be performed as follows: A clean image without watermarks, under the same lighting conditions as the glass image, is acquired and used as a preset background image. Then, pixel grayscale difference operations are performed between the preset background image and the processed glass image to remove the background.

[0099] By subtracting pixels, the same and fixed background components (such as the inherent texture of the glass, the overall ambient lighting, and fixed reflective areas) in the two images can be canceled out, while the differential details that only exist in the image after wiping (such as residual water ripples, water film boundaries, and uneven wiping texture) are preserved and highlighted, resulting in a feature image that eliminates background interference.

[0100] It is worth noting that when the user presses the glass washer switch (triggering the spraying of glass washer fluid and the start of wiping by the wipers), there is a brief delay in the spraying of the washer fluid. During the instant before the washer fluid is sprayed and the wipers begin wiping, an image of the glass in the current environment is captured as a preset background image. This preset background image is free of washer fluid runoff marks and wiping residue. Then, after the wipers have cleaned the glass, step S110 is triggered again to capture an image of the glass after it has been wiped by the wiper blades.

[0101] Furthermore, after obtaining the feature image, determining the visual sharpness index of the glass after the wiper blade has wiped it can be achieved in the following way: Identify the water ripple features present in the feature image. Based on the water ripple features, determine the visual sharpness index of the glass after the wiper blade has wiped it.

[0102] Specifically, the water ripple feature includes texture features and gradient features. The texture feature includes contrast features and homogeneity features.

[0103] The gradient feature extraction process includes: First, using gradient operators (most commonly the Sobel operator) to calculate the gradient values ​​in the horizontal direction (x direction) and vertical direction (y direction) at each pixel (x,y) in the feature image, respectively, to obtain the horizontal gradient Gx(x,y) and the vertical gradient Gy(x,y).

[0104] The horizontal gradient Gx(x,y) reflects the intensity change of the feature image in the left-right direction, while the vertical gradient Gy(x,y) reflects the intensity change of the image in the up-down direction.

[0105] Next, the gradient magnitude M(x,y) of each pixel is calculated. Specifically, for each pixel in the image, its gradient magnitude M(x,y) is calculated from the horizontal gradient Gx(x,y) and vertical gradient Gy(x,y) of that pixel. Specifically, the formula for calculating the gradient magnitude M(x,y) is: M(x,y) = The gradient magnitude M(x,y) represents the total intensity of change at that pixel.

[0106] Next, the average gradient magnitude of the feature image is calculated. Specifically, the entire feature image (or a specified region of interest within the feature image) is traversed, the gradient magnitudes M(x,y) of all pixels are summed, and then divided by the total number of pixels K to obtain the average gradient magnitude Mean_Gradient. That is, Mean_Gradient = [ΣM(x,y)] / K. This average gradient magnitude is the gradient feature.

[0107] In addition, striped water ripples (which appear on the glass after being wiped by worn wiper blades) typically exhibit high contrast and low homogeneity. Therefore, in this embodiment, the water ripple feature also includes texture features, which include contrast features and homogeneity features.

[0108] Specifically, the process of determining the contrast feature and the homogeneity feature includes:

[0109] First, a gray-level co-occurrence matrix (GLCM) is generated based on the preprocessed feature image. This GLCM is a square matrix that describes the probability of a pair of pixels with specific gray values ​​appearing simultaneously in the feature image at a specific direction and distance.

[0110] The distance is usually set to d=1 (i.e., adjacent pixels). The direction is usually calculated in four directions (0°, 45°, 90°, 135°), and then the average value is taken to obtain rotation invariance.

[0111] The generation process of the Gray-Level Co-occurrence Matrix (GLCM) includes: assuming that the gray levels of the feature image are quantized to L levels (e.g., 0-255 is quantized to 16 levels, then L=16). First, create an L×L zero matrix P as the initial GLCM.

[0112] Then, iterate through each pixel (i,j) in the feature image, finding neighboring pixels (i+Δi,j+Δj) at a specified direction (e.g., horizontal 0°) and distance (d=1). Let the grayscale value of the current pixel be g1, and the grayscale value of the neighboring pixel be g2. Then, increment the element value in the g1-th row and g2-th column of matrix P by 1 to count the number of occurrences of the grayscale pair (g1,g2).

[0113] After traversal, all elements of matrix P are divided by the sum of all gray-level pairs to normalize it into a probability matrix. At this point, P(i,j) represents the probability that gray-level values ​​i and j co-occur under a specific spatial relationship, which is the gray-level co-occurrence matrix GLCM.

[0114] After obtaining the gray-level co-occurrence matrix GLCM (i.e., matrix P(i,j)), the eigenvalues ​​of contrast and homogeneity features can be calculated. Specifically, let the dimension of the matrix be L, and i and j represent gray-level indices.

[0115] For contrast features: Contrast is calculated using the formula Contrast=ΣΣ|ij|²×P(i,j), which is a double summation of i from 0 to L-1 and j from 0 to L-1, to measure the intensity of local changes and the sharpness of texture in the feature image. It reflects the weight of elements in the matrix that are far from the main diagonal.

[0116] For an ideal, uniform water film, the grayscale difference between pixels is small, the |ij| value is small, and the feature value of the contrast feature is low.

[0117] If there are striped water ripples, a large number of pixel pairs with large gray-level differences will be generated at the edge of the water ripples. The |ij| value is large, and the square operation will amplify this difference. Therefore, the feature value of the contrast feature will be significantly increased.

[0118] For homogeneous features: Homogeneity is calculated using the formula: Homogeneity=ΣΣP(i,j) / (1+|ij|), which is a double summation of i from 0 to L-1 and j from 0 to L-1.

[0119] The homogeneity feature measures the uniformity or local consistency of image texture. It assigns higher weights to elements near the main diagonal (i.e., pixel pairs with similar gray values).

[0120] It is worth noting that the water ripple feature can also include the image energy proportion feature. Considering that fine water stains will appear as specific components in an image, a two-dimensional fast Fourier transform can be performed on the feature image to analyze the energy proportion in the feature image, thereby obtaining the image energy proportion feature.

[0121] After obtaining the above water ripple characteristics, each water ripple characteristic is normalized and then weighted and summed to finally obtain the visual clarity index. The higher the visual clarity index, the clearer the glass is after wiping and the less wear the wiper blades have.

[0122] Normalization can be achieved by dividing each water ripple feature by its corresponding preset feature benchmark value.

[0123] When performing a weighted summation of the normalized water ripple features, the weight of each feature can be set by the staff based on its impact on the visual acuity index (and the sum of the weights of all features should be 1). For example, if it is found in actual testing that the homogeneous feature is more typical when the wiper blade is worn, then the homogeneous feature can be given a higher weight.

[0124] Therefore, by extracting water ripple features from the feature image, the visual sharpness index of the glass after being wiped by the wiper blade can be determined. Since the water ripple feature reflects the residual water patterns on the glass after the wiper blade has wiped it, this visual sharpness index determined by the water ripple feature is more suitable for judging the degree of wear on the wiper blade. In other words, using such a visual sharpness index can more accurately determine the degree of wear on the wiper blade, thus improving the accuracy of the wear determination.

[0125] It is worth noting that in the above embodiments, after extracting the visual clarity index from the glass image and the harmonic energy distribution parameter from the operating current, in the above steps S122 and S123, the visual clarity index and the harmonic energy distribution parameter are fused, and the wear degree of the wiper blade is determined based on the fused features. Specifically, this can be achieved in the following ways.

[0126] First, determine the electrical feature vector V_electric of the driving motor's operating current, and determine the visual feature vector V_vision of the image captured by the camera.

[0127] The electrical characteristic vector V_electric contains the aforementioned harmonic energy distribution parameters and their statistics.

[0128] Specifically, to make the electrical feature vector more robust and trend-oriented, in this embodiment, the statistics in the electrical feature vector V_electric can be designed from two dimensions: multiple brushing cycles within the same diagnostic period and historical diagnostic sequences. Specifically, these can include: the mean of harmonic energy distribution parameters H_mean, the standard deviation of harmonic energy distribution parameters H_std, the rate of change of harmonic energy distribution parameters H_trend, and the total rate of change of harmonic energy distribution parameters H_ratio.

[0129] That is, the electrical feature vector V_electric is: V_electric=[H_current,H_mean,H_std,H_trend,H_ratio], which is a 5-dimensional vector containing information about the current state, volatility, and trend of change.

[0130] Specifically, the calculation process for each statistic is as follows: First, in this embodiment, the operating current is collected once during two or more wiping cycles within each diagnostic cycle (after the user triggers the wipers to turn on, the wipers will perform multiple wiping cycles, and the diagnostic cycle refers to the cycle from when the wipers are in the above-mentioned preset working condition to when the wipers completely stop moving, including multiple wiping cycles). The harmonic energy distribution parameter H_current corresponding to each wiping cycle is calculated respectively.

[0131] Then, the average value of the harmonic energy distribution parameter H_current within this diagnostic cycle is calculated to obtain the mean value of the harmonic energy distribution parameter H_mean.

[0132] The standard deviation of the harmonic energy distribution parameter H_current during this diagnostic cycle is calculated to obtain the standard deviation H_std. An increase in the standard deviation indicates instability in the scraping process, a sign of wear.

[0133] Calculate the linear regression slope of the harmonic energy distribution parameter H_current for the most recent N diagnoses (e.g., 5 times; if there are fewer than 5 times in the current diagnostic cycle, it can be obtained from relevant historical diagnostic data), and obtain the rate of change of the harmonic energy distribution parameter H_trend. This rate of change of the harmonic energy distribution parameter H_trend is used to reflect the trend of accelerated wear.

[0134] Calculate the ratio of the current harmonic energy distribution parameter H_current to the baseline value H_current_0 of the harmonic energy distribution parameter when the wiper blade is brand new, and obtain the total rate of change of the harmonic energy distribution parameter H_ratio.

[0135] The visual feature vector V_vision includes the aforementioned visual sharpness index and its components (including the various water ripple features mentioned above, such as gradient features, contrast features, homogeneity features, and image energy proportion features). That is, the visual feature vector V_vision is: V_vision = [Mean_Gradient, Contrast, Homogeneity, HighFreqEnergy]. This visual feature vector V_vision is a 4-dimensional vector that represents the edge sharpness, local contrast, uniformity, and subtle texture of the water film residue, providing rich and interpretable raw visual information for subsequent two-stream fusion models.

[0136] The electrical feature vector and the visual feature vector are then input into a pre-trained attention-based dual-stream fusion model. The attention-based dual-stream fusion model is used to fuse the electrical feature vector and the visual feature vector to obtain fused features. Based on the fused features, the current wear level of the wiper blade is predicted and output.

[0137] Specifically, the attention-based dual-stream fusion model is a lightweight network designed specifically for multimodal feature fusion. It outputs a wear index based on the input electrical and visual feature vectors, which represents the degree of wear.

[0138] More specifically, the architecture of this dual-stream fusion model includes a dual-stream feature embedding layer, an attention fusion layer, and regression and classification heads.

[0139] This two-stream feature embedding layer serves as two independent input branches, used for high-order abstraction and dimensionality reduction of the electrical and visual feature vectors. Specifically, this two-stream feature embedding layer typically consists of several fully connected layers, each followed by an activation function (such as ReLU) and a batch normalization layer (optional).

[0140] The dual-flow feature embedding layer specifically includes an electrical flow sub-network and a visual flow sub-network. The electrical flow sub-network receives the electrical feature vector and outputs the high-level abstract feature Fe corresponding to the electrical feature based on the electrical feature vector. The visual flow sub-network receives the visual feature vector and outputs the high-level abstract feature Fv corresponding to the visual feature based on the visual feature vector.

[0141] Each sub-network is typically a simple multilayer perceptron. Taking the electrical flow sub-network as an example, its architecture includes an input layer, two hidden layers, and an output layer. The input layer receives the original electrical feature vector V_electric (assuming it is 5-dimensional). The first hidden layer consists of a fully connected layer, followed by an activation function (such as ReLU) and a batch normalization layer, performing the operation: H1 = ReLU(BN(W1 × V_electric + b1)).

[0142] Here, W1 and b1 are learnable weights and biases. Batch Normalization (BN) stabilizes training, and ReLU introduces non-linearity. This layer can map dimensions from 5 to 16.

[0143] The structure of the second hidden layer is similar to that of the first hidden layer, and it can map the dimensions from 16 to 8.

[0144] The output layer is a fully connected layer (usually without an activation function or using linear activation) used to map the dimension to the high-level feature dimension of the target, for example, mapping 8 dimensions to 4 dimensions, the output being Fe. The operation performed is Fe = W² × H² + b².

[0145] Specifically, the input layer of this electric current subnetwork is: x = V_electric ∈ R 5 (Assume 5 dimensions). First layer (hidden layer): h1 = ReLU(W1·x + b1). Where W1 ∈ R (8×5) b1∈R 8 Output h1∈R 8 The second layer (output layer / feature embedding layer): Fe = ReLU(W2·h1+b2). ​​Where W2∈R (4×8) b2∈R4, output Fe∈R 4 .

[0146] That is, the final expression for Fe is: Fe = ReLU(W2·ReLU(W1·V_electric+b1)+b2).

[0147] It is worth noting that the visual flow subnetwork has the same architecture as the electrical flow subnetwork. Through this visual flow subnetwork, V_vision is output as a high-level abstract feature Fv. Specifically, the structure of the visual flow subnetwork is symmetrical to that of the electrical flow subnetwork, and its function is to map the multidimensional visual feature vector V_vision to the high-level abstract feature Fv. Referring to the aforementioned description of the electrical flow subnetwork, specifically, the input of this visual flow subnetwork is x = V_vision ∈ R. 4 First layer: h1 = ReLU(W1_v·x + b1_v), W1_v ∈ R (6×4) ,b1_v∈R 6Output h1∈R 6 Second layer: F_v = ReLU(W2_v·h1 + b2_v), W2_v ∈ R^(4×6), b2_v ∈ R 4 Output Fv∈R 4 Therefore, the expression for this high-level abstract feature Fv is: Fv = ReLU(W2_v·ReLU(W1_v·V_vision+b1_v)+b2_v).

[0148] It is worth noting that the weights and biases W1, b1, W2, b2 mentioned above are not set manually, but are automatically learned and optimized using sample data during the model training phase through the backpropagation algorithm with the goal of minimizing the final wear index prediction error and wear pattern classification error.

[0149] The learning objective of the entire dual-stream feature embedding layer is to transform the original features into a form (Fe, Fv), enabling subsequent attention fusion layers and regression / classification heads to make judgments most easily and accurately. If, during training, a certain original feature is found to be useless for the task (i.e., determining the wear index and classifying wear patterns), its weight will be learned to be close to zero after training, thus "forgotten" in high-level features.

[0150] The high-level abstract features Fe and Fv output by the dual-stream feature embedding layer will then be fed into an attention fusion layer, which will perform a weighted fusion of these two high-level abstract features to obtain the fused feature F. fused .

[0151] Specifically, the attention fusion layer first calculates the attention weights of the high-level abstract features Fe and Fv. In particular, it performs interactive calculations on the high-level abstract features Fe and Fv, for example, by concatenating or adding Fe and Fv. Then, through a small fully connected network (followed by a Sigmoid function), it generates an attention weight scalar α between 0 and 1.

[0152] Then, the attention weight scalar α is used to weight and combine the two high-level abstract features Fe and Fv. For example, a simple form is: F fused =α×Fe+(1-α)×Fv.

[0153] It is worth noting that environmental factors (such as temperature and vehicle speed) can also be used as input to determine the attention weight. For example, when the current environment is judged to be more reliable by electrical features (such as an oily film on the glass interfering with vision), the generated attention weight α is close to 1; when the visual features are more reliable (such as heavy rain causing large fluctuations in current load), α is close to 0.

[0154] Then, the regression and classification heads use the fused features F output by this attention fusion layer. fused Outputs the degree of wear and wear pattern (e.g., uniform aging, unilateral wear). The regression and classification headers include a regression header (for outputting the degree of wear) and a classification header (for outputting the wear pattern).

[0155] Specifically, the regression head typically consists of 1-2 fully connected layers, outputting a neuron that is constrained to a wear index between 0 and 1 (corresponding to 0-100%) using the Sigmoid activation function. This wear index, Wear_Index, represents the wear level mentioned above.

[0156] The classification head is parallel to the regression head and is usually composed of a fully connected layer connected to a Softmax activation function. It outputs the probability distribution of wear patterns (such as uniform aging or unilateral wear), and then determines the current wear pattern of the wiper blade based on the probability distribution.

[0157] It is worth noting that this attention-based dual-stream fusion model is trained using a large number of historical feature vector samples (each sample includes an electrical feature vector V_electric and a visual feature vector V_vision). Each historical feature vector sample corresponds to two labels (wear index and wear pattern).

[0158] The loss function of this attention-based two-stream fusion model can be the total loss obtained by weighted summing of regression loss and classification loss. The regression loss typically uses mean squared error to measure the difference between the predicted wear index and the true value. The classification loss typically uses cross-entropy loss to measure the difference between the predicted wear pattern probability distribution and the true wear pattern label.

[0159] The specific training process includes: dividing the prepared sample data (including labels) into training, validation, and test sets. An optimizer (such as Adam) is used to minimize the total loss through backpropagation, continuously updating all parameters in the network (including the weights of the two-stream feature embedding layer, the attention fusion layer, and the two output heads (i.e., the classification head and the regression head)). During training, the attention fusion layer automatically learns how to allocate the weights of the two high-level abstract features under different features to achieve optimal prediction performance.

[0160] In the above embodiments, once the wear level of the wiper blade in the current state is determined, step S130 is executed to use the wear level in the current state to predict the remaining life of the wiper blade.

[0161] Continue by Figures 1 to 3 and combined Figure 4As shown, in some exemplary embodiments, the remaining lifespan of the wiper blade in step S130 is predicted based on the wear level of the wiper blade determined multiple times within a preset prediction period, which may specifically include the following steps S131-S132.

[0162] Step S131: Based on the wear degree of the wiper blade determined in each preset prediction period, fit the wear degree change curve.

[0163] The preset prediction period can be the period from the last wiper blade replacement to the current moment. Within this period, a wear level diagnosis will be performed each time the wipers are in the preset operating condition. (It is worth noting that if the wipers are in the preset operating condition multiple times within a short period, only one wear level diagnosis may be performed. For example, after completing one wiper blade wear level diagnosis, if the wipers are in the preset operating condition again after 1 minute, no further wear level diagnosis will be performed. Instead, the wear level can be determined again after more than 1 day when the wipers are in the preset operating condition, to avoid invalid repeated sampling under repeated operating conditions in a short period of time.) In other words, the preset prediction period includes multiple diagnosis periods, and each diagnosis period includes multiple wiping periods.

[0164] In step S131, based on the wear degree of the wiper blades determined in each of the preset prediction cycles, a wear degree sequence is formed in chronological order. After forming the wear degree sequence, a wear degree change curve is obtained by fitting the wear degree change trend of the wiper blades in each cycle with the number of wiping actions (which is also equivalent to the number of wear degree diagnosis times).

[0165] It is worth noting that in some embodiments, after forming a wear degree sequence, a wear degree change curve reflecting the change of wear degree over time can be fitted based on the wear degree sequence, which is not limited here. In this embodiment, the wear degree change curve is used as an example to illustrate the trend of the wear degree of the wiper blade with the number of times.

[0166] Step S132: Predict the remaining life of the wiper blades based on the wear degree change curve.

[0167] The remaining lifespan refers to the duration from the current moment until the wear level of the wiper blade reaches a preset wear level threshold.

[0168] Specifically, after fitting the wear degree change curve in step S131, step S132 can predict the future wear degree of the wiper blade based on the wear degree change curve to determine when the wear degree of the wiper blade will reach the preset wear degree threshold in the future, thereby predicting the remaining life of the wiper blade. For example, based on the trend of the wear degree change curve, it is calculated that it will only take 10 days from the current moment for the wear degree to reach the preset wear degree threshold, then the predicted remaining life of the wiper blade is 10 days.

[0169] More specifically, in steps S131 and S132, when predicting the remaining lifespan of the wiper blade, the wear level collected each time can be fitted using an exponentially smoothed state-space model to obtain the wear level change curve. This exponentially smoothed state-space model is a time-series prediction model that integrates exponential smoothing and state-space modeling. It can decompose the wiper blade wear level data into two dimensions: wear level state and wear trend state. By weighted iterative updates of the wear level obtained from each diagnosis, it assigns higher weight to recent wear data and exponentially decays weight to long-term data. It automatically learns and outputs the current baseline value of the wiper blade wear level (wear level state) and the average wear growth rate (wear trend state), which characterizes the rate of aging. Simultaneously, it can smooth out random errors and operating condition disturbances from a single diagnosis, accurately reproducing the gradual aging pattern of the wiper blade. Then, based on the steady-state wear trend, it performs multi-step extrapolation prediction to achieve a quantitative estimate of the remaining lifespan of the wiper blade.

[0170] In other words, this exponentially smoothed state-space model can automatically learn the current wear level baseline value and the average increase in wear level after each wear level diagnosis (i.e., the wear rate). The current wear level baseline value is obtained by taking the wiper blade wear level obtained from each actual diagnosis as the observation input, combining it with the previous wear level baseline value and historical wear growth rates through weighted fusion. This exponential weighting method weakens the influence of random errors in single detections and operational disturbances, preserving the steady-state level of gradual aging of the wiper blades, and iteratively updating to obtain the smoothed wear level baseline value for the current moment.

[0171] For example, with a smoothing coefficient of 0.2, a previous wear level baseline value of 60, and an average wear rate of 0.5, if the current wear level diagnosis is 63, then the smoothed wear level baseline value is 0.2 × 63 + 0.8 × (60 + 0.5) = 61.0. That is, the current wear level is 63 (which may have errors due to slight changes in lighting or minor fluctuations in motor load), and this exponentially smoothed state-space model can filter out these errors to obtain a more realistic wear level (i.e., the wear level baseline value).

[0172] In the prediction, the exponential smoothing state-space model assumes that the increasing trend of wear will continue for a period of time in the future. Then it predicts when the wear will reach a preset wear threshold in the future, or predicts the wear of the wiper blade after N wipes / M days in the future, thus predicting the remaining life of the wiper blade.

[0173] For example, if the degree of wear after N future wipings is to be predicted, the prediction step size h = N, and the predicted degree of wear = current wear baseline value + prediction step size × wear growth trend, thus obtaining the degree of wear after N future wipings.

[0174] To predict the wear level of the wiper blades M days from now, first calculate the historical average diagnostic interval R (days / time). For example, if the wipers are used and a diagnostic is performed every 5 days on average, then the historical average diagnostic interval R = 5. Next, calculate the equivalent prediction step size h = M / R. Then, substitute the predicted wear level = current wear level baseline + equivalent prediction compensation × wear level growth trend to obtain the wiper blade wear level M days from now.

[0175] Therefore, through steps S131 and S132, by using the wear degree of the wiper blade determined each time within a preset prediction period, a wear degree change curve is fitted to predict the remaining life of the wiper blade. Since the wear degree change curve can reflect the changing trend of the wear degree of the wiper blade after each wipe, predicting the remaining life by using the aging pattern of the wiper blade characterized by the wear degree at multiple historical moments can avoid blindly predicting based on the wear degree of a single instance, thereby improving the accuracy of life prediction.

[0176] Continuing from the above Figures 1 to 4 As shown, in some exemplary embodiments, the wiper blade life prediction method may further include: determining the current wiper blade replacement warning level based on the currently determined wear level of the wiper blade and the predicted remaining life of the wiper blade; and outputting a preset wiper blade replacement reminder message in the vehicle based on the current wiper blade replacement warning level.

[0177] Specifically, if the current wear level of the wiper blades is between a first preset threshold and a second preset threshold (inclusive of the second preset threshold, excluding the first preset threshold), or if the remaining lifespan of the wiper blades is predicted to meet the first preset lifespan condition, it indicates that the current wear level of the wiper blades is relatively high, but they can still be used for a certain period of time. In this case, the user can be notified in advance so that they can prepare to replace the wiper blades. Therefore, the current corresponding replacement warning level is determined as the preset warning level, and the preset wiper blade replacement prompt information corresponding to this preset warning level could be: displaying "Wiper blade performance moderately degraded, attention recommended" on the dashboard.

[0178] If the current wear level of the wiper blades falls between the second and third preset thresholds (excluding the second threshold, but including the third), or if the remaining lifespan of the wiper blades is predicted to meet the second preset lifespan condition, it indicates that the wiper blades are currently highly worn, or that they will reach the end of their lifespan in the near future. In this case, it is necessary to clearly prompt the user to replace the wiper blades and also assist the user in finding the nearest wiper blade replacement service center. Therefore, in this situation, the current corresponding replacement warning level is determined as the preset warning level, and under this preset warning level, the preset wiper blade replacement prompt message could be: a continuous icon warning and text prompt on the dashboard until the wiper blades are replaced, with the text prompt message being "Wipe blade replacement recommended." Additionally, a "one-click search for nearby wiper blade replacement service centers" option can be provided in conjunction with the navigation.

[0179] When the current wiper blade wear reaches the third preset threshold, or when the current wear mode is output as "unilateral wear" in the aforementioned attention-based dual-flow fusion model, it indicates a high safety risk associated with continuing to use the current wiper blades, requiring the user's attention. Therefore, in this situation, the corresponding replacement warning level is set to a preset severe warning level. Under this preset severe warning level, the preset wiper blade replacement prompt message can be: a strong audible and visual alarm, and the text message "Wiper blades are ineffective, replace immediately" displayed on the dashboard. Furthermore, the corresponding wiper blade replacement prompt message can be triggered again the next time the wipers are used.

[0180] The third preset threshold is greater than the second preset threshold, which in turn is greater than the first preset threshold. For example, the first preset threshold is a wear threshold used to indicate a decline in wiper blade performance and to advise the user to pay attention to the wiper blade's condition; for example, it can be set to 60. The second preset threshold is a moderate wear threshold representing a significant decrease in wiper blade wiping performance, indicating that the wiper blade is nearing the end of its service life; for example, it can be set to 85. The third preset threshold is a heavy wear threshold representing a near-complete failure of the wiper blade's wiping function, indicating that the wiper blade has reached its service life limit; for example, it can be set to 95.

[0181] The first preset lifespan condition may include a prediction that the wear level of the wiper blade will exceed a second preset threshold within a first preset duration in the future. The first preset duration can be set according to the vehicle's normal usage frequency and the average aging cycle of the wipers, for example, it can be set to 30 days.

[0182] The second preset lifespan condition may include: predicting that the wear level of the wiper blade will exceed the aforementioned third preset threshold within a second preset duration in the future. The second preset duration can be set according to the critical cycle of rapid aging of the wiper blade under high-frequency use of the wiper, for example, it can be set to 7 days.

[0183] It is worth noting that if the current wear level of the wiper blade is less than the first preset threshold, and the predicted remaining lifespan does not meet the above-mentioned first preset lifespan condition and second preset lifespan condition, it means that the current wiper blade performance is good and the remaining lifespan is long, and no wiper blade replacement prompt information is output to remind the user to replace the wiper blade.

[0184] Therefore, based on the currently determined wear level and predicted remaining lifespan of the wiper blades, corresponding wiper blade replacement reminders are output to the vehicle in a tiered manner, enabling graded warnings based on actual wear status and remaining aging trends. This avoids premature and unnecessary reminders that may inconvenience users, and also prevents delayed warnings that could lead to wiper failure and impair driving visibility.

[0185] It is worth noting that in this embodiment, it can also be connected to the weather forecast network. For example, if the current wiper blade wear is high (e.g., greater than a second preset threshold) before heavy rain / storm weather is predicted, the user can be prompted to replace the wiper blades Q days (e.g., two days) before the heavy rain / storm weather arrives. This can be done by outputting the corresponding wiper blade replacement prompt information in the vehicle or by pushing the corresponding wiper blade replacement prompt information to the vehicle application installed on the owner's terminal device. This is to remind the user to replace the wiper blades in advance before the heavy rain / storm weather arrives, so as to prevent the user from having blurred vision due to failure to replace the wiper blades before the heavy rain / storm weather arrives, which would affect driving safety.

[0186] It is worth noting that, regarding the wiper blade life prediction method of this embodiment, based on the above exemplary implementations, in specific implementation, as a preferred embodiment, it is still based on... Figures 1 to 4As shown, this could include, for example, triggering a wiper blade wear diagnosis and wiper blade life prediction operation when the vehicle's windshield wipers are in a preset setting, the vehicle's rain sensor outputs a preset detection signal, and the vehicle speed is within a preset speed range. Specifically, it first acquires the operating current of the wiper drive motor and an image of the glass after being wiped by the wiper blades. Then, based on the operating current of the drive motor, it calculates the harmonic energy distribution parameters of the drive motor's operating current, and determines the visual clarity index of the glass after being wiped by the wiper blades based on the glass image.

[0187] The current wear level of the wiper blades is then determined using the visual acuity index, harmonic energy distribution parameters, and a pre-trained attention-based dual-stream fusion model. Based on the wiper blade wear levels determined multiple times within a preset prediction period, a wear level change curve is fitted to predict the remaining lifespan of the wiper blades.

[0188] Furthermore, based on the current wear level and remaining lifespan of the wiper blades, a wiper blade replacement warning level is determined to remind car owners to replace the wiper blades in a tiered manner.

[0189] In the preferred embodiment of the above wiper blade life prediction method, the specific implementation means of each step can still be referred to the descriptions in the above exemplary embodiments, and the beneficial effects brought about by the design of each step in this preferred embodiment can also be referred to the descriptions in the above exemplary embodiments.

[0190] The wiper blade life prediction method in this embodiment adopts the design described above. It first determines the wear level of the wiper blade in its current state by using the operating current of the drive motor and the glass image. Then, based on the wear levels in multiple states determined within a preset prediction period, it predicts the remaining lifespan of the wiper blade. In this method of predicting the remaining lifespan of the wiper blade, the current state of the wiper blade is not determined by a single physical quantity, but by a comprehensive analysis of two physical quantities: the operating current and the glass image. This improves the accuracy of determining the current wear level of the wiper blade. Consequently, when using the wear level of the wiper blade to predict its remaining lifespan, it can more accurately analyze the changes in the wear level of the wiper blade, thereby improving the accuracy of the remaining lifespan prediction.

[0191] An embodiment of the second aspect of this application provides a vehicle, such as Figure 5 As shown, the vehicle's controller includes processor 510 and processor 520. The memory 520 stores computer programs. The processor 510 executes the computer programs stored in the memory to implement the wiper blade life prediction method described in the embodiments of the first aspect above.

[0192] It is worth noting that, Figure 5 The vehicle controller shown is merely an example and should not impose any limitations on the functionality and scope of use of the embodiments of this application.

[0193] In this embodiment, the vehicle, by executing the wiper blade life prediction method in the above method embodiment, can predict the remaining life of the wiper blade by combining the operating current of the drive motor and the image of the wiped glass, rather than by directly predicting a single physical quantity. This can improve the accuracy of predicting the remaining life of the wiper blade.

[0194] The above descriptions are merely some embodiments of this application and are not intended to limit this application. The technical features or structures in the foregoing different embodiments can be arbitrarily combined to form other specific technical solutions as needed. For those skilled in the art, this application can have various modifications and variations. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of this application should be included within the scope of protection of the claims of this application.

Claims

1. A method for predicting the lifespan of windshield wipers, applied to vehicles, characterized in that, The method includes: When the windshield wipers in the vehicle are in a preset working condition, the operating current of the drive motor of the windshield wipers and the image of the glass after being wiped by the wiper blades are acquired. The degree of wear of the wiper blade is determined based on the operating current of the drive motor and the glass image. The remaining lifespan of the wiper blade is predicted based on the wear level of the wiper blade determined multiple times within a preset prediction period.

2. The method for predicting wiper blade life according to claim 1, characterized in that, Determining the wear level of the wiper blade based on the operating current of the drive motor and the glass image includes: Based on the operating current of the drive motor, the harmonic energy distribution parameters of the operating current of the drive motor are calculated, and the visual clarity index of the glass after the wiper blade has wiped is determined based on the glass image. The visual acuity index and the harmonic energy distribution parameters are fused to obtain the fused feature; The wear degree of the wiper blade is determined based on the fusion characteristics.

3. The method for predicting wiper blade life according to claim 2, characterized in that, The step of calculating the harmonic energy distribution parameters of the operating current of the drive motor based on the operating current of the drive motor includes: The power frequency and fundamental frequency components in the operating current of the drive motor are filtered to obtain the load harmonic current signal. The load harmonic current signal is subjected to a fast Fourier transform to obtain the spectrum of the load harmonic current signal; Each frequency point within a preset high-frequency band is determined from the spectrum, and the total harmonic energy within the preset high-frequency band is calculated based on the amplitude corresponding to each determined frequency point. The harmonic energy distribution parameters are calculated based on the total harmonic energy and the amplitude corresponding to each frequency point.

4. The method for predicting wiper blade life according to claim 3, characterized in that, The step of calculating the harmonic energy distribution parameters based on the total harmonic energy and the amplitude corresponding to each frequency point includes: Based on the amplitude corresponding to each frequency point, calculate the harmonic energy corresponding to each frequency point respectively; Determine the proportion of harmonic energy corresponding to each frequency point to the total harmonic energy; Based on the respective energy proportions, the harmonic entropy of the spectrum is calculated and used as the harmonic energy distribution parameter.

5. The method for predicting wiper blade life according to claim 2, characterized in that, The step of determining the visual clarity index of the glass after the wiper blade has wiped the glass based on the glass image includes: The glass image is preprocessed to obtain a feature image; Based on the feature image, determine the visual clarity index of the glass after the wiper blade has wiped it.

6. The method for predicting wiper blade life according to claim 5, characterized in that, The preprocessing of the glass image to obtain a feature image includes: The glass image is subjected to grayscale conversion and contrast enhancement processing; Obtain a preset background image, and use the preset background image to perform background removal processing on the processed glass image to obtain the feature image; and / or, Determining the visual clarity index of the glass after the wiper blade has wiped the windshield based on the feature image includes: Identify the water ripple features present in the feature image; Based on the water ripple characteristics, the visual clarity index of the glass after being wiped by the wiper blade is determined.

7. The method for predicting wiper blade life according to claim 1, characterized in that, The method of predicting the remaining lifespan of the wiper blade based on the wear level determined multiple times within a preset prediction period includes: Based on the wear degree of the wiper blade determined in each of the preset prediction periods, a wear degree change curve is obtained by fitting. The remaining lifespan of the wiper blade is predicted based on the wear variation curve.

8. The method for predicting wiper blade life according to claim 1, characterized in that, The method also includes: Based on the currently determined wear level of the wiper blade and the predicted remaining lifespan of the wiper blade, determine the current replacement warning level for the wiper blade; Based on the current replacement warning level corresponding to the wiper blade, a preset wiper blade replacement prompt message is output to the vehicle.

9. The method for predicting wiper blade life according to claim 1, characterized in that: The preset operating conditions include the windshield wipers being in a preset gear and the vehicle speed being within a preset speed range.

10. A vehicle, characterized in that, The vehicle's controller includes: Memory, used to store computer programs; A processor is configured to execute a computer program stored in the memory to implement the wiper blade life prediction method according to any one of claims 1-9.

Images