Vehicle windscreen wiper control method and device, electronic equipment and storage medium

Abstract

The invention provides a vehicle windscreen wiper control method and device, electronic equipment and a storage medium, and relates to the technical field of vehicle control. Dividing a key image area from the image and calculating a first definition score; if the first definition score exceeds the preset evaluation range, the vehicle windscreen wiper is controlled based on the first definition score; if the first definition score does not exceed the preset evaluation range, multi-scale features are extracted for weighting, and a second definition score and the confidence of the second definition score are obtained; when the confidence degree reaches the standard, the vehicle windscreen wiper is controlled based on the second definition score; and when the confidence does not reach the standard, extracting image space features, determining a third definition score and a variance thereof in combination with the driving environment parameters, and controlling the vehicle windscreen wiper based on the third definition score and the variance thereof. By adopting the technical scheme provided by the invention, the vehicle windscreen wiper is dynamically adjusted, the control accuracy is improved, and meanwhile, the system energy consumption is effectively reduced.

Description

Technical Field

[0001] This application relates to the field of vehicle control technology, and in particular to a control method, device, electronic device and storage medium for a vehicle windshield wiper. Background Technology

[0002] With the continuous improvement of automotive intelligence, automatic windshield wiper systems have become one of the key technologies for enhancing driving safety and convenience. Existing wiper systems typically rely on rain sensors, which control the wiper's operation by detecting the physical signals of raindrops. Specifically, existing wiper systems consist of three parts: a rain sensor, a control unit, and a wiper motor. The rain sensor is mounted on the windshield and is responsible for collecting rainfall information in real time; the vehicle's control unit determines whether to activate the wipers based on set thresholds and adjusts their operating frequency; the wiper motor performs the wiping operation to clean the windshield.

[0003] However, in situations where rainfall is light but visibility is obscured by fogging or dirt on the windshield, the wipers may not activate promptly, compromising driving safety. Furthermore, rain sensors cannot adapt to complex and changing weather conditions and driving environments, such as uneven rain distribution at high speeds. Additionally, the fixed frequency of wiper operation often leads to unnecessary energy waste. Summary of the Invention

[0004] In view of this, embodiments of this application provide a method, device, electronic device, and storage medium for controlling a vehicle windshield wiper. By assessing the driver's visibility quality, the method enables dynamic adjustment of the vehicle windshield wiper, thereby improving the driver's visibility quality, enhancing the accuracy of the vehicle windshield wiper control, and effectively reducing system energy consumption.

[0005] This application mainly includes the following aspects: In a first aspect, embodiments of this application provide a method for controlling a vehicle windshield wiper, wherein a camera is installed at the rearview mirror of the vehicle; the camera is used to capture images of the scene within its field of view; the control method includes: Acquire images of the scene in front of the vehicle captured by the camera; The image is divided into at least one key image region related to vehicle driving according to a preset division rule; A first sharpness score for the image is determined based on the pixel variance of each key image region and the underlying weights corresponding to each key image region. If the first clarity score is not within the preset evaluation range, then based on the preset clarity score range in which the first clarity score is located, control the vehicle's windshield wipers to wipe the vehicle's windshield according to the corresponding wiping parameters. If the first sharpness score is within a preset evaluation range, then multiple scale features are extracted from the image, and the multiple scale features are weighted and fused to determine the second sharpness score of the image and the confidence level corresponding to the second sharpness score; If the confidence level is greater than or equal to the preset confidence threshold, then based on the preset clarity score range in which the second clarity score is located, the vehicle wipers are controlled to wipe the vehicle windshield according to the corresponding wiping parameters. If the confidence level is less than a preset confidence threshold, then image spatial features are extracted from the image, and based on the image spatial features and driving environment parameters, the third sharpness score of the image and the score variance corresponding to the third sharpness score are determined. Based on the preset clarity score range in which the third clarity score falls and the score variance, the vehicle's windshield wipers are controlled to wipe the windshield according to the corresponding wiping parameters.

[0006] Preferably, determining the first sharpness score of the image based on the pixel variance of each key image region and the basic weight corresponding to each key image region includes: The pixel variance of each key image region is determined using the Laplacian operator; Based on the vehicle driving environment parameters, determine whether to adjust the basic weights corresponding to at least one key image region; If it is determined that the basic weights corresponding to at least one key image region should be adjusted, the basic weights corresponding to at least one key image region should be adjusted according to a preset adjustment step size; the result of taking the logarithm of the pixel variance of each key image region should be weighted with the corresponding adjusted basic weights to obtain the first sharpness score of the image. If it is determined that the base weights corresponding to at least one key image region will not be adjusted, then the result of taking the logarithm of the pixel variance corresponding to each key image region is weighted with the corresponding base weight to obtain the first sharpness score of the image.

[0007] Preferably, the step of weighted fusion of the multiple scale features to determine the second sharpness score of the image and the confidence level corresponding to the second sharpness score includes: Align the features at multiple scales; For each scale feature, the aligned scale feature is convolved, and the result of the convolution is determined as the compressed feature of that layer. The compressed features of each layer are fused, and the fused result is determined as the initial fused feature; Based on the initial fusion features and the channel weights corresponding to the initial fusion features, a second sharpness score of the image and a confidence level corresponding to the second sharpness score are predicted.

[0008] Preferably, predicting the second sharpness score of the image and the confidence level corresponding to the second sharpness score based on the initial fusion features and the channel weights corresponding to the initial fusion features includes: The initial fused features are then subjected to global average pooling to obtain the first global feature vector; Based on the first global feature vector, determine the channel weights corresponding to the initial fusion features; Each layer of compressed features is weighted with the channel weights, and the weighted result is determined as the fusion feature; The fusion features are then reduced in dimensionality. The fused features after dimensionality reduction are then subjected to global average pooling to obtain a second global feature vector; Based on the second global feature vector, predict the second sharpness score of the image and the confidence level corresponding to the second sharpness score.

[0009] Preferably, determining the third sharpness score of the image and the corresponding score variance based on the image spatial features and driving environment parameters includes: The spatial features of the image are reduced in dimensionality to obtain the spatial dimensionality-reduced features of the image. Each parameter in the driving environment parameters is preprocessed, and the preprocessed parameters are concatenated. The concatenated result is determined as the context feature vector. The spatial dimensionality reduction features of the image are concatenated with the first sharpness score of the image to obtain the concatenated feature vector corresponding to the image. Temporal features that characterize the change in image sharpness over time are extracted from the stitched feature vector corresponding to the image and the stitched feature vector corresponding to historical images that precede the image by a preset number of frames. Based on the spatial dimensionality reduction features and temporal features of the image, determine the high-level features corresponding to the context feature vector; Based on the high-level features, at least one prediction is made for the image's sharpness score; The average of all predicted sharpness scores is taken as the third sharpness score of the image, and the variance of all predicted sharpness scores is taken as the score variance corresponding to the third sharpness score.

[0010] Preferably, determining the high-level features corresponding to the context feature vector based on the spatial dimensionality reduction features of the image and the temporal features includes: The temporal features are convolved, and the result of the convolution is determined as the attention weight corresponding to the temporal features. The temporal features are weighted with the corresponding attention weights, and the weighted result is determined as the spatial weighted features. Based on the temporal features and the spatial weighted features, the weights corresponding to the context feature vector are determined; The context feature vector is weighted by its corresponding weight; The weighted result is fused with the spatial weighted features, and the fused result is determined as the high-level feature corresponding to the context feature vector.

[0011] Preferably, the wiping parameters include: wiping speed and wiping cycle.

[0012] Secondly, embodiments of this application also provide a control device for a vehicle windshield wiper, wherein a camera is installed at the rearview mirror of the vehicle; the camera is used to capture images of the scene within its field of view; the control device includes: The acquisition module is used to acquire images of the scene in front of the vehicle captured by the camera; The segmentation module is used to divide the image into at least one key image region related to vehicle driving according to a preset segmentation rule; The first evaluation module is used to determine the first sharpness score of the image based on the pixel variance of each key image region and the basic weight corresponding to each key image region. The first control module is used to control the vehicle windshield wipers to wipe the vehicle windshield according to the corresponding wiping parameters based on the preset clarity score interval in which the first clarity score is located if the first clarity score is not within the preset evaluation range. The second evaluation module is used to extract multiple scale features from the image if the first sharpness score is within a preset evaluation range, and to perform weighted fusion of the multiple scale features to determine the second sharpness score of the image and the confidence level corresponding to the second sharpness score. The second control module is used to control the vehicle windshield wipers to wipe the vehicle windshield according to the corresponding wiping parameters based on the preset clarity score range in which the second clarity score is located if the confidence level is greater than or equal to the preset confidence threshold. The third evaluation module is used to extract image spatial features from the image if the confidence level is less than a preset confidence threshold, and to determine the third sharpness score of the image and the score variance corresponding to the third sharpness score based on the image spatial features and driving environment parameters. The third control module is used to control the vehicle's windshield wipers to wipe the windshield according to the corresponding wiping parameters based on the preset clarity score range in which the third clarity score is located and the score variance.

[0013] Thirdly, embodiments of this application also provide an electronic device, including: a processor, a memory, and a bus. The memory stores machine-readable instructions executable by the processor. When the electronic device is running, the processor communicates with the memory through the bus. The machine-readable instructions are executed by the processor to perform the steps of the windshield wiper control method described in the first aspect or any possible implementation of the first aspect.

[0014] Fourthly, embodiments of this application also provide a computer-readable storage medium storing a computer program, which, when executed by a processor, performs the steps of the windshield wiper control method described in the first aspect or any possible implementation of the first aspect.

[0015] This application provides a method, device, electronic device, and storage medium for controlling a vehicle windshield wiper. The method involves acquiring an image of the scene in front of the vehicle; dividing the image into key image regions and calculating a first sharpness score; if the first sharpness score exceeds a preset evaluation range, controlling the vehicle windshield wiper based on the first sharpness score; if the first sharpness score does not exceed the preset evaluation range, extracting multi-scale features and weighting them to obtain a second sharpness score and its confidence level; when the confidence level meets the standard, controlling the vehicle windshield wiper based on the second sharpness score; when the confidence level does not meet the standard, extracting image spatial features and combining them with driving environment parameters to determine a third sharpness score and its variance, and controlling the vehicle windshield wiper based on the third sharpness score and its variance.

[0016] This allows for dynamic adjustment of the vehicle's windshield wipers, improving control precision while effectively reducing system energy consumption.

[0017] To make the above-mentioned objectives, features and advantages of this application more apparent and understandable, preferred embodiments are described below in detail with reference to the accompanying drawings. Attached Figure Description

[0018] To more clearly illustrate the technical solutions of the embodiments of this application, the accompanying drawings used in the embodiments will be briefly introduced below. It should be understood that the following drawings only show some embodiments of this application and should not be regarded as a limitation of the scope. For those skilled in the art, other related drawings can be obtained based on these drawings without creative effort.

[0019] Figure 1 This invention provides a flowchart of one of the control methods for a vehicle windshield wiper according to an embodiment of this application. Figure 2 The flowchart of a vehicle windshield wiper control method provided in an embodiment of this application is shown; Figure 3 A second flowchart of a vehicle windshield wiper control method provided in an embodiment of this application is shown; Figure 4 A flowchart of a vehicle windshield wiper control method provided in this application embodiment is shown as third; Figure 5 A flowchart of a vehicle windshield wiper control method provided in this application embodiment is shown as fourth; Figure 6 The fifth flowchart illustrates a method for controlling a vehicle windshield wiper according to an embodiment of this application. Figure 7 A flowchart of a vehicle windshield wiper control method provided in this application embodiment is shown as sixth of the flowcharts. Figure 8 This paper shows a schematic diagram of the structure of a vehicle windshield wiper control device provided in an embodiment of this application; Figure 9 A schematic diagram of the structure of an electronic device provided in an embodiment of this application is shown. Detailed Implementation

[0020] To make the objectives, technical solutions, and advantages of the embodiments of this application clearer, the technical solutions of the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. It should be understood that the drawings in this application are for illustrative and descriptive purposes only and are not intended to limit the scope of protection of this application. Furthermore, it should be understood that the schematic drawings are not drawn to scale. The flowcharts used in this application illustrate operations implemented according to some embodiments of this application. It should be understood that the operations in the flowcharts may not be implemented in sequence, and steps without logical contextual relationships may be reversed or implemented simultaneously. In addition, those skilled in the art, guided by the content of this application, may add one or more other operations to the flowcharts, or remove one or more operations from the flowcharts.

[0021] Furthermore, the described embodiments are merely some, not all, of the embodiments of this application. The components of the embodiments of this application described and illustrated herein can typically be arranged and designed in various different configurations. Therefore, the following detailed description of the embodiments of this application provided in the accompanying drawings is not intended to limit the scope of the claimed application, but merely to illustrate selected embodiments of the application. All other embodiments obtained by those skilled in the art based on the embodiments of this application without inventive effort are within the scope of protection of this application.

[0022] The methods, apparatus, electronic devices, or computer-readable storage media described in this application can be applied to any scenario requiring windshield wiper control. This application does not limit specific application scenarios, and any scheme using the vehicle windshield wiper control method and apparatus provided in this application is within the protection scope of this application.

[0023] It is worth noting that with the continuous improvement of automotive intelligence, automatic windshield wiper systems have become one of the key technologies for enhancing driving safety and convenience. Existing windshield wiper systems typically rely on rain sensors, which control the wiper's operation by detecting the physical signals of raindrops. Specifically, existing wiper systems consist of three parts: a rain sensor, a control unit, and a wiper motor. The rain sensor is mounted on the windshield and is responsible for collecting rainfall information in real time; the vehicle's control unit determines whether to activate the wipers based on set thresholds and adjusts their operating frequency; the wiper motor performs the wiping operation to clean the windshield. However, in situations where rainfall is light but visibility is blurred due to water fog or smudges on the windshield, the wipers may not activate in time, affecting driving safety. Furthermore, rain sensors cannot adapt to complex and changing weather conditions and driving environments, such as uneven rain distribution at high speeds. In addition, because the wipers operate at a fixed frequency, unnecessary energy waste often occurs.

[0024] To address the aforementioned issues, this application proposes a method, device, electronic device, and storage medium for controlling vehicle windshield wipers. By assessing the driver's visibility, the method enables dynamic adjustment of the vehicle windshield wipers, improving the driver's visibility, enhancing the accuracy of windshield wiper control, and effectively reducing system energy consumption.

[0025] To facilitate understanding of this application, the technical solutions provided in this application will be described in detail below with reference to specific embodiments.

[0026] The core flaw of traditional rain detection technology lies in the fact that it detects physical signals related to rainfall, rather than the quality of vision as perceived by the driver. Therefore, in certain scenarios, existing technology cannot accurately trigger the windshield wipers, causing obstructed vision and affecting driving safety. For example, when rainfall is light but the windshield is partially covered by water, obscuring visibility, automatic wipers based on traditional rainfall thresholds may still not activate, failing to clear the water from the windshield in time.

[0027] Please see Figure 1 , Figure 1 This is one of the flowcharts for a vehicle windshield wiper control method provided in an embodiment of this application.

[0028] like Figure 1 As shown in the figure, the vehicle windshield wiper control method provided in this application embodiment includes the following steps: Step S101: Obtain an image of the scene in front of the vehicle captured by the camera.

[0029] Here, the camera is positioned at the rearview mirror of the vehicle to capture images of the scene within the camera's field of view, simulating the actual quality of the driver's vision.

[0030] In the embodiments of this application, such as Figure 2 As shown, to ensure the real-time performance and accuracy of image sharpness scoring detection, a three-layer detection method is employed, including a fast detection layer, a CNN (Convolutional Neural Network) enhancement layer, and a multimodal fusion layer. The wiper action is triggered progressively based on the output of each layer. The three-layer detection method will be explained in detail below.

[0031] Step S102: Divide the image into at least one key image region related to vehicle driving according to a preset division rule.

[0032] Here, the key image region is defined based on the visual attention distribution characteristics of the driver during actual driving. For example, the key image region may include, but is not limited to, roads, traffic signs, and the area of ​​vehicles ahead. The preset division rules are image region division criteria established based on the spatial distribution characteristics of the key image regions in the image. For example, the road area is set as the bottom 60% of the image, the traffic sign area is set as the top 30% of the image and centered horizontally at 80%, and the area of ​​vehicles ahead is set as the middle area between 30% and 70% of the image's vertical height.

[0033] Step S103: Determine the first sharpness score of the image based on the pixel variance of each key image region and the basic weight corresponding to each key image region.

[0034] Here, the base weights are initial weights assigned based on the importance of key image regions. The sum of the base weights for all key image regions is 1. For example, the base weight for the road region is 0.5, the base weight for the traffic sign region is 0.3, and the base weight for the vehicle ahead region is 0.2. The sharpness score reflects the current actual visual clarity of the windshield; the higher the score, the clearer the image.

[0035] The following is combined with Figure 3 This section will explain in detail how to determine the first sharpness score of the image based on the pixel variance of each key image region and the basic weight corresponding to each key image region.

[0036] Please see Figure 3 , Figure 3This is a second flowchart of a vehicle windshield wiper control method provided in an embodiment of this application.

[0037] like Figure 3 As shown, regarding step S103, in a specific implementation, as an example, the following steps may be included: Step S1031: Use the Laplacian operator to determine the pixel variance of each key image region.

[0038] Here, as an example, the pixel variance of each key image region is calculated using a 7×7 Laplacian operator.

[0039] It should be noted that the Laplacian operator can effectively enhance edge features in image processing. The basic principle is to perform a Laplacian convolution operation on the image to generate an edge enhancement map, and then calculate the pixel variance of the image. In image quality assessment, sharp images, due to their rich edge information, will have larger variance values, while blurry images will exhibit smaller variance values. However, traditional global variance evaluation methods have limitations: on the one hand, a single overall evaluation ignores the differences in local image regions, leading to distorted evaluation results; on the other hand, this method lacks adaptability to complex scenes (such as situations where multiple regions and multiple scales of blur coexist), making it difficult to accurately reflect the actual image quality. To address these issues, an improved multi-region Laplacian variance algorithm is used to perform weighted calculations on key image regions in each frame.

[0040] Step S1032: Based on the vehicle driving environment parameters, determine whether to adjust the basic weights corresponding to at least one key image region.

[0041] Here, the vehicle driving environment may include, but is not limited to, vehicle speed and windshield wiper status. The purpose of the basic weight adjustment is to dynamically adjust the weight allocation of each key image region according to the vehicle driving environment, thereby improving the accuracy of the assessment of actual visibility quality.

[0042] Step S1033: If it is determined that the basic weights corresponding to at least one key image region need to be adjusted, then the basic weights corresponding to at least one key image region need to be adjusted according to the preset adjustment step size.

[0043] Here, as an example, assuming the vehicle speed exceeds 60 km / h, given the importance of traffic sign recognition at high speeds, the base weight corresponding to the traffic sign area is increased by 0.1; assuming the windshield wipers are on, and rainwater is partially cleared, the base weight corresponding to the road area is decreased by 0.05. It should be noted that during the adjustment of base weights, the sum of the adjusted base weights for all key image areas must be 1.

[0044] Step S1034: The result of taking the logarithm of the pixel variance of each key image region is weighted with the corresponding adjusted base weight to obtain the first sharpness score of the image.

[0045] Here, taking the logarithm of the pixel variance for each key image region avoids the influence of extreme variance values ​​on the results. This step, through a dynamic weighting mechanism based on the importance of key image regions, effectively captures the low-frequency blurring features caused by rain in each key image region, and obtains the image sharpness score through weighted calculation.

[0046] Step S1035: If it is determined that the basic weights corresponding to at least one key image region will not be adjusted, then the result of taking the logarithm of the pixel variance corresponding to each key image region is weighted with the corresponding basic weight to obtain the first sharpness score of the image.

[0047] See again Figure 1 In step S104, if the first clarity score is not within the preset evaluation range, then based on the preset clarity score range in which the first clarity score is located, control the vehicle wipers to wipe the vehicle windshield according to the corresponding wiping parameters.

[0048] Here, as an example, the preset evaluation range is [300, 1200]. The preset clarity scoring intervals are based on the driver's line-of-sight status and include: a first scoring interval, a second scoring interval, a third scoring interval, and a fourth scoring interval. The lower limit of the first scoring interval is greater than the upper limit of the second scoring interval, the lower limit of the second scoring interval is greater than the upper limit of the third scoring interval, and the lower limit of the third scoring interval is greater than the upper limit of the fourth scoring interval. For example, the first scoring interval is greater than 1200, the second scoring interval is [800, 1200], the third scoring interval is [300, 800], and the fourth scoring interval is less than 300. The driver's line-of-sight status in the first scoring interval is excellent, with sharp image edges and clear texture details. The driver's line-of-sight status in the second scoring interval is good, with slight blurring but key image areas are still identifiable and do not affect driving. The driver's line-of-sight status in the third scoring interval needs improvement, with significantly blurred image features and lost details. The driver's line-of-sight status in the fourth scoring interval is severely blurred, with key image areas being invisible. The preset assessment range includes: the third scoring range and the fourth scoring range.

[0049] In this embodiment, the swiping parameters may include, but are not limited to, swiping speed and swiping cycle. When the first sharpness score is within the first score range, both the swiping speed and swiping cycle are 0, and swiping stops to save energy. When the first sharpness score is within the fourth score range, high-speed continuous swiping is performed, with the swiping speed being the first swiping speed and the swiping cycle being the first swiping cycle. For example, the first swiping speed can be 45-60 times / minute, and the first swiping cycle can be 1.0-1.3 seconds / time.

[0050] Step S105: If the first sharpness score is within a preset evaluation range, then extract multiple scale features from the image and perform weighted fusion of the multiple scale features to determine the second sharpness score of the image and the confidence level corresponding to the second sharpness score.

[0051] It should be noted that steps S102-S103 constitute the first layer of detection. When the first sharpness score output by the first layer of detection is within the preset evaluation range, it indicates that it cannot accurately distinguish between slight blur and obvious blur. Therefore, the second CNN enhancement layer needs to be triggered for a more refined sharpness evaluation. The second CNN enhancement layer uses a lightweight neural network to evaluate sharpness.

[0052] In another possible implementation, when the difference between the first sharpness scores of three consecutive frames exceeds a preset dynamic range, i.e. the results of the three consecutive frames are unstable, a second CNN enhancement layer can be triggered to perform a more refined sharpness evaluation.

[0053] In step S105, the second layer employs a multi-stage image preprocessing method to improve the sharpness assessment accuracy: First, the Y channel pixel values ​​are linearly scaled from [0, 255] to [0, 1] through normalization to unify the numerical range of each input, adapting to the standardization requirements of the neural network for input data distribution to unify the input distribution; then, for backlight or strong light, the multi-scale Retinex algorithm is used to separate the illumination component and reflection component in key image regions, suppressing overexposure and enhancing shadow details, avoiding edge blurring caused by global processing; subsequently, the CLAHE (Contrast Limited Adaptive Histogram Equalization) algorithm with a block size of 8×8 and a threshold of 2.0 is used to enhance contrast, enhancing the brightness contrast between blurred and sharp areas, making detail differences more obvious (e.g., the high-frequency edge features of small raindrops are more prominent), while effectively suppressing the noise amplification problem caused by over-enhancement; finally, BM3D (Block-Matching and 3D) is used... The filtering (3D block matching filter) algorithm removes nighttime noise interference while preserving the characteristics of raindrops / snowflakes. Finally, Z-score (standard score) standardization is implemented based on the statistical mean and standard deviation of the training set to stabilize the data distribution and improve model convergence.

[0054] The second layer preprocesses the image and inputs it into a model with MobileNetV2 as the backbone network, extracting multi-scale feature maps from different depths. Shallow features are used to extract high-frequency features, such as subtle blurring and edge jaggedness caused by light raindrops; mid-level features are used to extract mid-frequency features, such as local blurring caused by moderate rain and fog; and deep features are used to extract low-frequency features, such as global blurring caused by heavy rain (affecting all areas of the image) and the overall blurring trend caused by vehicle movement (blurring direction along the movement direction). Features at different depths represent blur information at different scales in the image.

[0055] The following is combined with Figure 4 This section will specifically explain how to perform weighted fusion of the multiple scale features to determine the second sharpness score of the image and the confidence level corresponding to the second sharpness score.

[0056] Please see Figure 4 , Figure 4 This is the third flowchart of a vehicle windshield wiper control method provided in the embodiments of this application.

[0057] like Figure 4 As shown, regarding step S105, in a specific implementation, as an example, the following steps may be included: Step S1051: Align the multiple scale features.

[0058] Here, the mid-level and deep features are upsampled to the same size as the shallow features. For example, assuming the shallow feature size is 64×64×32, the mid-level feature size is 32×32×96, and the deep feature size is 7×7×320, after aligning the mid-level and deep features with the shallow features, the aligned mid-level features will be 64×64×96, and the aligned deep features will be 64×64×320.

[0059] Step S1052: For each scale feature, perform convolution on the aligned scale feature, and determine the result of the convolution as the compressed feature of that layer.

[0060] Here, as an example, following the example above, the aligned seed-scale features are convolved with 1×1 to compress the channels to 64.

[0061] Step S1053: Fuse the compressed features of each layer and determine the fused result as the initial fused feature.

[0062] Here, assuming the shallow feature is F1, the middle feature is F2, and the deep feature is F3, the initial fused feature is F1+F2+F3.

[0063] Step S1054: Based on the initial fusion features and the channel weights corresponding to the initial fusion features, predict the second sharpness score of the image and the confidence level corresponding to the second sharpness score.

[0064] The following is combined with Figure 5 This section will specifically explain how to predict the second sharpness score of the image and the confidence level corresponding to the second sharpness score based on the initial fusion features and the channel weights corresponding to the initial fusion features.

[0065] Please see Figure 5 , Figure 5 This is the fourth flowchart of a vehicle windshield wiper control method provided in the embodiments of this application.

[0066] like Figure 5 As shown, regarding step S1054, in a specific implementation, as an example, the following steps may be included: Step S10541: Perform global average pooling on the initial fused features to obtain the first global feature vector.

[0067] Here, continuing from the example above, global average pooling of the initial fused features yields a 64-dimensional first global feature vector.

[0068] Step S10542: Determine the channel weights corresponding to the initial fusion features based on the first global feature vector.

[0069] Here, the first global feature vector is input into a two-layer fully connected layer, undergoes a two-layer fully connected nonlinear transformation, and is then processed by a Sigmoid activation function to generate channel weights. Continuing with the example above, the first global feature vector is sequentially passed through a 64-16-64 dimension fully connected layer, and a Sigmoid activation function is applied at the output to generate 64-dimensional channel weights.

[0070] Step S10543: Weight each layer of compressed features with the channel weights, and determine the weighted result as the fusion feature.

[0071] Here, feature fusion can highlight feature channels that contribute significantly to sharpness assessment.

[0072] Step S10544: Dimensionality reduction of the fused features.

[0073] Continuing with the example above, the fused features can be reduced from 64×64×64 to 32×32×32 through 3×3 convolution.

[0074] Step S10545: Perform global average pooling on the dimensionality-reduced fused features to obtain the second global feature vector.

[0075] Continuing with the example above, the 32×32×32 fused features after dimensionality reduction can be obtained by global average pooling to obtain a 32-dimensional second global feature vector.

[0076] Step S10546: Based on the second global feature vector, predict the second sharpness score of the image and the confidence level corresponding to the second sharpness score.

[0077] In this application, a dual-branch structure is employed, used for regressing sharpness scoring and confidence levels respectively. The sharpness scoring branch takes the second global feature vector as input, passes through two fully connected layers, and then outputs a normalized sharpness score between 0 and 1 via a Sigmoid activation function. This score undergoes denormalization to map to the physical range corresponding to the first layer score; for example, the physical range is 0-1500. The confidence level branch takes the second global feature vector as input, passes through two fully connected layers, and then outputs a normalized confidence level between 0 and 1 via a Sigmoid activation function.

[0078] When training the two branches, the loss function for the sharpness scoring branch uses mean squared error (MSE), and the formula for mean squared error is: ,in, The loss function for the sharpness scoring branch, Score the sharpness of the prediction. The sharpness is the true value. The loss function for the confidence branch uses a custom loss function based on the prediction error. The formula for the custom loss function is: , The loss function for the confidence branch. represents the confidence level. During training, the total loss function for both branches is weighted and summed to maintain both the accuracy and reliability of the prediction results in terms of sharpness prediction. The specific formula is as follows: ,in, and It's a hyperparameter.

[0079] See again Figure 1 In step S106, if the confidence level is greater than or equal to a preset confidence threshold, then based on the preset clarity score range in which the second clarity score is located, control the vehicle wipers to wipe the vehicle windshield according to the corresponding wiping parameters.

[0080] Here, when the second sharpness score falls within the first score range, both the wiping speed and wiping cycle are 0. When the first sharpness score falls within the second score range, slow, intermittent wiping is performed at the second wiping speed and the second wiping cycle. For example, the second wiping speed could be 6-12 times / minute, and the second wiping cycle could be 5.0-10.0 seconds / wipe. When the second sharpness score falls within the third score range, low-speed continuous wiping is performed at the third wiping speed and the third wiping cycle. For example, the third wiping speed could be 30-45 times / minute, and the third wiping cycle could be 1.3-2.0 seconds / wipe. When the second sharpness score falls within the fourth score range, high-speed continuous wiping is performed at the first wiping speed and the first wiping cycle.

[0081] Step S107: If the confidence level is less than a preset confidence threshold, then extract image spatial features from the image, and determine the third sharpness score of the image and the score variance corresponding to the third sharpness score based on the image spatial features and driving environment parameters.

[0082] Here, image spatial features are obtained by inputting the image into a ResNet50 pre-trained model (with the classification head removed) and extracting the image spatial features output by the ResNet50 (Residual Network 50, a 50-layer residual neural network) pre-trained model. To effectively prevent overfitting and better adapt the model to rain and fog blur features, the following adjustments were made to the pre-trained model: the parameters of the first three layers were frozen, the fourth and fifth layers could be fine-tuned during training, and a Dropout mechanism was introduced to improve the model's robustness against interference.

[0083] As an example, the preset confidence threshold is 0.8. It should be noted that step S105 is the second-layer detection. When the confidence level of the second sharpness score output by the second-layer detection is less than the preset confidence threshold, it indicates that the first two layers cannot accurately assess edge scenes. Therefore, the third-layer multimodal fusion layer needs to be triggered for a more refined sharpness assessment. Step S107 is the third-layer detection.

[0084] In another possible implementation, the third-layer multimodal fusion layer can also be triggered for sharpness assessment when a sensor malfunction occurs or when the user performs calibration. Sensor malfunctions include, but are not limited to, situations such as camera obstruction, CAN bus data interruption, and ambient light sensor value fluctuation exceeding 50% / second. The user calibration trigger condition is that the system automatically performs a recalibration within 30 seconds after the driver manually adjusts the windshield wipers.

[0085] In step S107, the third layer preprocesses the multi-source input data respectively: for images from the camera, adaptive brightness equalization (to cope with backlight and low light conditions) and standardization methods are used, which are consistent with the second layer preprocessing method; for time-series data from the same front-view camera, timestamp alignment and time-series normalization are performed (subtracting the mean of the last 10 frames).

[0086] The following is combined with Figure 6 This section will specifically explain how to determine the third sharpness score of the image and the corresponding score variance based on the image spatial features and driving environment parameters.

[0087] Please see Figure 6 , Figure 6 The fifth flowchart is a method for controlling a vehicle windshield wiper provided in the embodiments of this application.

[0088] like Figure 6 As shown, regarding step S107, in a specific implementation, as an example, the following steps may be included: Step S1071: Dimensionality reduction of the spatial features of the image is performed to obtain the spatial dimensionality reduction features of the image.

[0089] Step S1072: Preprocess each parameter in the driving environment parameters, and concatenate the preprocessed parameters to determine the concatenated result as the context feature vector.

[0090] Here, the normalized vehicle speed, the logarithmic ambient light intensity, and the current wiper state are concatenated to obtain the context feature vector.

[0091] Step S1073: The spatial dimensionality reduction features of the image are concatenated with the first sharpness score of the image to obtain the concatenated feature vector corresponding to the image.

[0092] Step S1074: Extract temporal features that characterize the change of image sharpness over time from the stitching feature vector corresponding to the image and the stitching feature vector corresponding to historical images that precede the image by a preset number of frames.

[0093] Here, as an example, the preset frame count is 9 frames. Specifically, the stitched feature vector corresponding to the image and the stitched feature vector corresponding to the historical images of the preset frame count are input into Bi-LSTM (Bidirectional Long Short-Term Memory) to extract temporal features.

[0094] Step S1075: Based on the spatial dimensionality reduction features of the image and the temporal features, determine the high-level features corresponding to the context feature vector.

[0095] The following is combined with Figure 7 This section will specifically explain how to determine the high-level features corresponding to the context feature vector based on the spatial dimensionality reduction features and the temporal features of the image.

[0096] Please see Figure 7 , Figure 7 This is the sixth flowchart of a vehicle windshield wiper control method provided in the embodiments of this application.

[0097] like Figure 7 As shown, regarding step S1075, in a specific implementation, as an example, the following steps may be included: Step S10751: Convolve the temporal features and determine the result of the convolution as the attention weight corresponding to the temporal features.

[0098] Here, by compressing the channels to 1 through a 1×1 convolution and then activating them with a Sigmoid function, we can obtain the attention weights corresponding to the temporal features.

[0099] Step S10752: The temporal features are weighted with the corresponding attention weights, and the weighted result is determined as the spatial weighted features.

[0100] Here, the weight of roads, vehicles ahead, and traffic signs can be increased.

[0101] Step S10753: Based on the temporal features and the spatial weighted features, determine the weights corresponding to the context feature vector.

[0102] Step S10754: Weight the context feature vector with the corresponding weights.

[0103] Step S10755: The weighted result is fused with the spatial weighted feature, and the fused result is determined as the high-level feature corresponding to the context feature vector.

[0104] Here, the weighted result is fused with the spatial weighted features, that is, stability is enhanced through residual connections.

[0105] See again Figure 6 Step S1076: Based on the high-level features, make at least one prediction of the image's sharpness score.

[0106] Here, a regression head is used for sharpness prediction, and the Monte Carlo Dropout method is used to quantify uncertainty. The specific process is as follows: the fused features are first input into a fully connected layer, then through a Dropout layer, then through a fully connected output layer, and finally the sharpness score between 0 and 1 is output using the Sigmoid function, and this sharpness score is mapped to the physical range corresponding to the score of the first layer.

[0107] In an embodiment of this application, as an example, this application performs 10 predictions on the sharpness score of the image, that is, performs 10 forward propagations on the high-level features to obtain 10 predicted sharpness scores.

[0108] Step S1077: The average of all predicted sharpness scores is taken as the third sharpness score of the image, and the variance of all predicted sharpness scores is taken as the score variance corresponding to the third sharpness score.

[0109] Here, the third sharpness score is the average of all predicted sharpness scores.

[0110] The variance of the score is used as a supplementary measure of model confidence; the larger the variance, the lower the certainty of the third-level sharpness score. In the embodiments of this application, the sharpness score classifies the visual sharpness of the windshield into levels, and the uncertainty quantification result provides a risk decision reference for the wiper control layer. For example, even if the sharpness score is in the third-level range and needs improvement, if the uncertainty is high, i.e., the variance is low, the control layer can adopt a more conservative wiping strategy (e.g., using fast intermittent wiping instead of low-speed continuous wiping), or actively trigger a fault safety detection mechanism.

[0111] Step S108: Based on the preset clarity score range in which the third clarity score is located and the score variance, control the vehicle's windshield wipers to wipe the vehicle's windshield according to the corresponding wiping parameters.

[0112] Here, when the third sharpness score is within the first scoring interval, and the variance score is within either the first or second variance interval, both the wiping speed and wiping cycle are 0. For example, the first variance interval is less than 0.005, and the second variance interval is 0.005-0.02. When the third sharpness score is within the first scoring interval and the variance score is within the third variance interval, the wiping speed is the second wiping speed, and the wiping cycle is the second wiping cycle. When the third sharpness score is within the second scoring interval, slow intermittent wiping is performed, with the wiping speed and wiping cycle being the second wiping speed and the wiping cycle being the second wiping cycle. When the third sharpness score is within the second scoring interval, and the variance score is within either the first or second variance interval, fast intermittent wiping is performed, with the wiping speed being the fourth wiping speed and the wiping cycle being the fourth wiping cycle. For example, the fourth wiping speed can be 20-30 times / minute, and the fourth wiping cycle can be 2.0-3.0 seconds / time. When the third clarity score is in the third score range and the variance score is in the first variance range, perform low-speed continuous wiping at the third wiping speed and for the third wiping cycle. When the third clarity score is in the third score range and the variance score is in the second variance range, perform rapid intermittent wiping at the fourth wiping speed and for the fourth wiping cycle. When the third clarity score is in the third score range and the variance score is in the third variance range, or when the third clarity score is in the fourth score range and the variance score is in the first variance range, trigger a fault safety check. When the third clarity score is in the fourth score range and the variance score is in the first or second variance range, perform high-speed continuous wiping at the first wiping speed and for the first wiping cycle. When the third clarity score is in the fourth score range and the variance score is in the third range, perform high-speed continuous wiping and trigger a fault safety check.

[0113] It should be noted that a variance score in the first variance interval indicates low uncertainty, meaning the prediction is stable and the confidence level is high; a variance score in the second variance interval indicates moderate confidence; and a variance score in the third variance interval indicates high uncertainty and low confidence.

[0114] In another possible implementation, an anti-shake and state stabilization mechanism can be employed, specifically comprising the following two parts: First, a hysteresis range of ±50 points is set for the critical value of the clarity rating (e.g., 800 points, 1200 points) to prevent frequent switching of wiper control modes near the critical state. For example, when the score rises from 790 points to 810 points, the state to be improved must be maintained for 3 consecutive frames before switching the wiper wiping parameters. Second, if a sensor malfunction is detected or the clarity assessment results are inconsistent for 5 consecutive frames, the system automatically switches to a fail-safe mode, in which wiping is performed at fixed 5-second intervals, while simultaneously prompting the driver for manual intervention through the user feedback system.

[0115] In this embodiment, a three-layer decision-making mechanism consisting of a traditional signal processing layer (fast response) and a deep learning analysis layer is adopted. The traditional signal processing layer can handle normal operating conditions, while the deep learning layer can handle edge scenarios, such as those with light rainfall but severe visibility obstruction. This three-layer decision-making mechanism improves the system's real-time performance and the accuracy of its judgment of clarity. By fusing multi-dimensional parameters such as vehicle speed, ambient light intensity, and image clarity, a dynamic model for quantitatively evaluating the driver's actual visibility quality is constructed. This transforms the system's detection target from traditional physical rainfall to more direct visual perception quality, thereby solving the misjudgment problem caused by traditional systems relying solely on rainfall thresholds. A dynamic allocation mechanism for computing resources is introduced into the deep learning layer. This mechanism can adaptively adjust the computational depth of the neural network based on the complexity of real-time operating conditions. Furthermore, compared to traditional wiper control methods based on fixed frequencies, this application reduces the overall energy consumption of the system by dynamically adjusting the wiper operation strategy.

[0116] This application provides a method for controlling a vehicle windshield wiper. This method enables dynamic adjustment of the vehicle windshield wiper, improves the driver's visibility, enhances the accuracy of windshield wiper control, and effectively reduces system energy consumption.

[0117] Based on the same application concept, this application also provides a vehicle wiper control device corresponding to the vehicle wiper control method provided in the above embodiments. Since the principle of the device in this application to solve the problem is similar to the vehicle wiper control method in the above embodiments of this application, the implementation of the device can refer to the implementation of the method, and the repeated parts will not be described again.

[0118] Please see Figure 8 , Figure 8 This is a schematic diagram of the structure of a vehicle windshield wiper control device provided in an embodiment of this application.

[0119] like Figure 8 As shown in the figure, the vehicle windshield wiper control device 810 provided in this application embodiment includes: The acquisition module 811 is used to acquire images of the scene in front of the vehicle captured by the camera; The segmentation module 812 is used to segment the image into at least one key image region related to vehicle driving according to a preset segmentation rule. The first evaluation module 813 is used to determine the first sharpness score of the image based on the pixel variance of each key image region and the basic weight corresponding to each key image region. The first control module 814 is used to control the vehicle windshield wipers to wipe the vehicle windshield according to the corresponding wiping parameters based on the preset clarity score interval in which the first clarity score is located if the first clarity score is not within the preset evaluation range. The second evaluation module 815 is used to extract multiple scale features from the image if the first sharpness score is within a preset evaluation range, and to perform weighted fusion of the multiple scale features to determine the second sharpness score of the image and the confidence level corresponding to the second sharpness score. The second control module 816 is used to control the vehicle windshield wipers to wipe the vehicle windshield according to the corresponding wiping parameters based on the preset clarity score range in which the second clarity score is located, if the confidence level is not less than a preset confidence threshold. The third evaluation module 817 is used to extract image spatial features from the image if the confidence level is less than a preset confidence threshold, and to determine the third sharpness score of the image and the score variance corresponding to the third sharpness score based on the image spatial features and driving environment parameters. The third control module 818 is used to control the vehicle windshield wipers to wipe the vehicle windshield according to the corresponding wiping parameters based on the preset clarity score range in which the third clarity score is located and the score variance.

[0120] Furthermore, the first evaluation module 813 is specifically used for: The pixel variance of each key image region is determined using the Laplacian operator; Based on the vehicle driving environment parameters, determine whether to adjust the basic weights corresponding to at least one key image region; If it is determined that the basic weights corresponding to at least one key image region should be adjusted, the basic weights corresponding to at least one key image region should be adjusted according to a preset adjustment step size; the result of taking the logarithm of the pixel variance of each key image region should be weighted with the corresponding adjusted basic weights to obtain the first sharpness score of the image. If it is determined that the base weights corresponding to at least one key image region will not be adjusted, then the result of taking the logarithm of the pixel variance corresponding to each key image region is weighted with the corresponding base weight to obtain the first sharpness score of the image.

[0121] Furthermore, when the second evaluation module 815 performs weighted fusion of the multiple scale features to determine the second sharpness score of the image and the confidence level corresponding to the second sharpness score, it is also specifically used for: Align the features at multiple scales; For each scale feature, the aligned scale feature is convolved, and the result of the convolution is determined as the compressed feature of that layer. The compressed features of each layer are fused, and the fused result is determined as the initial fused feature; Based on the initial fusion features and the channel weights corresponding to the initial fusion features, a second sharpness score of the image and a confidence level corresponding to the second sharpness score are predicted.

[0122] Furthermore, when the second evaluation module 815 predicts the second sharpness score of the image and the confidence level corresponding to the second sharpness score based on the initial fusion features and the channel weights corresponding to the initial fusion features, it is also specifically used for: The initial fused features are then subjected to global average pooling to obtain the first global feature vector; Based on the first global feature vector, determine the channel weights corresponding to the initial fusion features; Each layer of compressed features is weighted with the channel weights, and the weighted result is determined as the fusion feature; The fusion features are then reduced in dimensionality. The fused features after dimensionality reduction are then subjected to global average pooling to obtain a second global feature vector; Based on the second global feature vector, predict the second sharpness score of the image and the confidence level corresponding to the second sharpness score.

[0123] Furthermore, when the third evaluation module 817 is used to determine the third sharpness score of the image and the corresponding score variance based on the image spatial features and driving environment parameters, it is also specifically used for: The spatial features of the image are reduced in dimensionality to obtain the spatial dimensionality-reduced features of the image. Each parameter in the driving environment parameters is preprocessed, and the preprocessed parameters are concatenated. The concatenated result is determined as the context feature vector. The spatial dimensionality reduction features of the image are concatenated with the first sharpness score of the image to obtain the concatenated feature vector corresponding to the image. Temporal features that characterize the change in image sharpness over time are extracted from the stitched feature vector corresponding to the image and the stitched feature vector corresponding to historical images that precede the image by a preset number of frames. Based on the spatial dimensionality reduction features and temporal features of the image, determine the high-level features corresponding to the context feature vector; Based on the high-level features, at least one prediction is made for the image's sharpness score; The average of all predicted sharpness scores is taken as the third sharpness score of the image, and the variance of all predicted sharpness scores is taken as the score variance corresponding to the third sharpness score.

[0124] Furthermore, when the third evaluation module 817 determines the high-level features corresponding to the context feature vector based on the spatial dimensionality reduction features and the temporal features of the image, it is also specifically used for: The temporal features are convolved, and the result of the convolution is determined as the attention weight corresponding to the temporal features. The temporal features are weighted with the corresponding attention weights, and the weighted result is determined as the spatial weighted features. Based on the temporal features and the spatial weighted features, the weights corresponding to the context feature vector are determined; The context feature vector is weighted by its corresponding weight; The weighted result is fused with the spatial weighted features, and the fused result is determined as the high-level feature corresponding to the context feature vector.

[0125] Furthermore, the wiping parameters include: wiping speed and wiping cycle.

[0126] This application provides a vehicle windshield wiper control device that enables dynamic adjustment of the vehicle windshield wipers, improves the driver's visibility, enhances the accuracy of windshield wiper control, and effectively reduces system energy consumption.

[0127] Please see Figure 9 , Figure 9 This is a schematic diagram of the structure of an electronic device provided in an embodiment of this application.

[0128] like Figure 9 As shown, the electronic device 900 includes a processor 910, a memory 920, and a bus 930.

[0129] The memory 920 stores machine-readable instructions executable by the processor 910. When the electronic device 900 is running, the processor 910 and the memory 920 communicate via the bus 930. When the machine-readable instructions are executed by the processor 910, they can perform the operations described above. Figure 1 , Figure 2 , Figure 3 , Figure 4 , Figure 5 , Figure 6 and Figure 7 The steps of the vehicle windshield wiper control method in the illustrated method embodiment can be found in the method embodiment for specific implementation, and will not be repeated here.

[0130] This application also provides a computer-readable storage medium storing a computer program, which, when executed by a processor, can perform the above-described actions. Figure 1 , Figure 2 , Figure 3 , Figure 4 , Figure 5 , Figure 6 and Figure 7 The steps of the vehicle windshield wiper control method in the illustrated method embodiment can be found in the method embodiment for specific implementation, and will not be repeated here.

[0131] Those skilled in the art will clearly understand that, for the sake of convenience and brevity, the specific working processes of the systems and devices described above can be referred to the corresponding processes in the foregoing method embodiments, and will not be repeated here. In the several embodiments provided in this application, it should be understood that the disclosed systems, devices, and methods can be implemented in other ways. The device embodiments described above are merely illustrative. For example, the division of units is only a logical functional division; in actual implementation, there may be other division methods. Furthermore, multiple units or components may be combined or integrated into another system, or some features may be ignored or not executed. Another point is that the displayed or discussed mutual coupling or direct coupling or communication connection may be through some communication interfaces; the indirect coupling or communication connection of devices or units may be electrical, mechanical, or other forms.

[0132] The units described as separate components may or may not be physically separate. The components shown as units may or may not be physical units; that is, they may be located in one place or distributed across multiple network units. Some or all of the units can be selected to achieve the purpose of this embodiment according to actual needs.

[0133] In addition, the functional units in the various embodiments of this application can be integrated into one processing unit, or each unit can exist physically separately, or two or more units can be integrated into one unit.

[0134] If the aforementioned functions are implemented as software functional units and sold or used as independent products, they can be stored in a processor-executable, non-volatile, computer-readable storage medium. Based on this understanding, the technical solution of this application, in essence, or the part that contributes to the prior art, or a part of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute all or part of the steps of the methods described in the various embodiments of this application. The aforementioned storage medium includes various media capable of storing program code, such as USB flash drives, portable hard drives, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical disks.

[0135] The above are merely specific embodiments of this application, but the scope of protection of this application 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 this application should be included within the scope of protection of this application. Therefore, the scope of protection of this application should be determined by the scope of the claims.

Claims

1. A method for controlling a vehicle windshield wiper, characterized in that, A camera is installed in the rearview mirror of the vehicle. The camera is used to acquire images of the scene within its field of view; the control method includes: Acquire images of the scene in front of the vehicle captured by the camera; The image is divided into at least one key image region related to vehicle driving according to a preset division rule; A first sharpness score for the image is determined based on the pixel variance of each key image region and the underlying weights corresponding to each key image region. If the first clarity score is not within the preset evaluation range, then based on the preset clarity score range in which the first clarity score is located, control the vehicle's windshield wipers to wipe the vehicle's windshield according to the corresponding wiping parameters. If the first sharpness score is within a preset evaluation range, then multiple scale features are extracted from the image, and the multiple scale features are weighted and fused to determine the second sharpness score of the image and the confidence level corresponding to the second sharpness score; If the confidence level is greater than or equal to the preset confidence threshold, then based on the preset clarity score range in which the second clarity score is located, the vehicle wipers are controlled to wipe the vehicle windshield according to the corresponding wiping parameters. If the confidence level is less than a preset confidence threshold, then image spatial features are extracted from the image, and based on the image spatial features and driving environment parameters, the third sharpness score of the image and the score variance corresponding to the third sharpness score are determined. Based on the preset clarity score range in which the third clarity score falls and the score variance, the vehicle's windshield wipers are controlled to wipe the windshield according to the corresponding wiping parameters.

2. The vehicle windshield wiper control method according to claim 1, characterized in that, The determination of the first sharpness score of the image based on the pixel variance of each key image region and the basic weight corresponding to each key image region includes: The pixel variance of each key image region is determined using the Laplacian operator; Based on the vehicle driving environment parameters, determine whether to adjust the basic weights corresponding to at least one key image region; If it is determined that the basic weights corresponding to at least one key image region should be adjusted, the basic weights corresponding to at least one key image region should be adjusted according to a preset adjustment step size; the result of taking the logarithm of the pixel variance of each key image region should be weighted with the corresponding adjusted basic weights to obtain the first sharpness score of the image. If it is determined that the base weights corresponding to at least one key image region will not be adjusted, then the result of taking the logarithm of the pixel variance corresponding to each key image region is weighted with the corresponding base weight to obtain the first sharpness score of the image.

3. The method for controlling a vehicle windshield wiper according to claim 1, characterized in that, The step of weighted fusion of the multiple scale features to determine the second sharpness score of the image and the confidence level corresponding to the second sharpness score includes: Align the features at multiple scales; For each scale feature, the aligned scale feature is convolved, and the result of the convolution is determined as the compressed feature of that layer. The compressed features of each layer are fused, and the fused result is determined as the initial fused feature; Based on the initial fusion features and the channel weights corresponding to the initial fusion features, a second sharpness score of the image and a confidence level corresponding to the second sharpness score are predicted.

4. The vehicle windshield wiper control method according to claim 3, characterized in that, The step of predicting the second sharpness score of the image and the confidence level corresponding to the second sharpness score based on the initial fusion features and the channel weights corresponding to the initial fusion features includes: The initial fused features are then subjected to global average pooling to obtain the first global feature vector; Based on the first global feature vector, determine the channel weights corresponding to the initial fusion features; Each layer of compressed features is weighted with the channel weights, and the weighted result is determined as the fusion feature; The fusion features are then reduced in dimensionality. The fused features after dimensionality reduction are then subjected to global average pooling to obtain a second global feature vector; Based on the second global feature vector, predict the second sharpness score of the image and the confidence level corresponding to the second sharpness score.

5. The method for controlling a vehicle windshield wiper according to claim 1, characterized in that, The step of determining the third sharpness score of the image and the corresponding score variance based on the image spatial features and driving environment parameters includes: The spatial features of the image are reduced in dimensionality to obtain the spatial dimensionality-reduced features of the image. Each parameter in the driving environment parameters is preprocessed, and the preprocessed parameters are concatenated. The concatenated result is determined as the context feature vector. The spatial dimensionality reduction features of the image are concatenated with the first sharpness score of the image to obtain the concatenated feature vector corresponding to the image. Temporal features that characterize the change in image sharpness over time are extracted from the stitched feature vector corresponding to the image and the stitched feature vector corresponding to historical images that precede the image by a preset number of frames. Based on the spatial dimensionality reduction features and temporal features of the image, determine the high-level features corresponding to the context feature vector; Based on the high-level features, at least one prediction is made for the image's sharpness score; The average of all predicted sharpness scores is taken as the third sharpness score of the image, and the variance of all predicted sharpness scores is taken as the score variance corresponding to the third sharpness score.

6. The method for controlling a vehicle windshield wiper according to claim 5, characterized in that, The determination of the high-level features corresponding to the context feature vector based on the spatial dimensionality reduction features of the image and the temporal features includes: The temporal features are convolved, and the result of the convolution is determined as the attention weight corresponding to the temporal features. The temporal features are weighted with the corresponding attention weights, and the weighted result is determined as the spatial weighted features. Based on the temporal features and the spatial weighted features, the weights corresponding to the context feature vector are determined; The context feature vector is weighted by its corresponding weight; The weighted result is fused with the spatial weighted features, and the fused result is determined as the high-level feature corresponding to the context feature vector.

7. The method for controlling a vehicle windshield wiper according to claim 1, characterized in that, The wiping parameters include: wiping speed and wiping cycle.

8. A control device for a vehicle windshield wiper, characterized in that, A camera is installed in the rearview mirror of the vehicle. The camera is used to capture images of the scene within its field of view; The control device for the vehicle windshield wipers includes: The acquisition module is used to acquire images of the scene in front of the vehicle captured by the camera; The segmentation module is used to divide the image into at least one key image region related to vehicle driving according to a preset segmentation rule; The first evaluation module is used to determine the first sharpness score of the image based on the pixel variance of each key image region and the basic weight corresponding to each key image region. The first control module is used to control the vehicle windshield wipers to wipe the vehicle windshield according to the corresponding wiping parameters based on the preset clarity score interval in which the first clarity score is located if the first clarity score is not within the preset evaluation range. The second evaluation module is used to extract multiple scale features from the image if the first sharpness score is within a preset evaluation range, and to perform weighted fusion of the multiple scale features to determine the second sharpness score of the image and the confidence level corresponding to the second sharpness score. The second control module is used to control the vehicle windshield wipers to wipe the vehicle windshield according to the corresponding wiping parameters based on the preset clarity score range in which the second clarity score is located if the confidence level is greater than or equal to the preset confidence threshold. The third evaluation module is used to extract image spatial features from the image if the confidence level is less than a preset confidence threshold, and to determine the third sharpness score of the image and the score variance corresponding to the third sharpness score based on the image spatial features and driving environment parameters. The third control module is used to control the vehicle's windshield wipers to wipe the windshield according to the corresponding wiping parameters based on the preset clarity score range in which the third clarity score is located and the score variance.

9. An electronic device, characterized in that, include: The device includes a processor, a memory, and a bus, wherein the memory stores machine-readable instructions executable by the processor, and when the electronic device is running, the processor communicates with the memory via the bus, and the machine-readable instructions are executed by the processor to perform the steps of the vehicle windshield wiper control method as described in any one of claims 1 to 7.

10. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores a computer program that, when executed by a processor, performs the steps of the vehicle windshield wiper control method as described in any one of claims 1 to 7.

Images