Automobile light-operated detection method and system based on image recognition

Abstract

The invention relates to the technical field of automobile detection, and discloses an automobile light-operated detection method and system based on image recognition. The method comprises the following steps: firstly, collecting real-time image data of an automobile light-operated component, and extracting a light-operated component area image after noise reduction and enhancement processing; extracting feature points of the regional image, and identifying structural features and state features to form a feature data set; classifying the feature data set through a deep learning model, and determining a current working state parameter of the light-operated component; comparing the current working state parameter with a preset standard parameter, and calculating a deviation degree to obtain a deviation evaluation value; and judging whether the light-operated part is abnormal or not by combining a preset deviation threshold range, and generating a detection result report. The method can accurately detect the working state of the light-operated part of the automobile, discover abnormity in time, improve the comprehensiveness and accuracy of detection, and is suitable for light-operated detection scenes in automobile production and maintenance.

Description

Technical Field

[0001] The present invention relates to the technical field of automobile detection, and in particular to an automobile light control detection method and system based on image recognition. Background Art

[0002] With the rapid development of the automotive industry, vehicle safety and reliability are receiving increasing attention. As a crucial component of driving safety, the operational stability of automotive light control components directly impacts visibility and road safety. Currently, automotive light control components encompass a wide range of categories, including headlights, turn signals, brake lights, and fog lights. These components perform critical lighting and signaling functions during driving, steering, and braking. Traditional methods for inspecting automotive light control components rely heavily on manual visual inspection or simple sensor data collection, which presents numerous limitations. During manual inspection, the accuracy of the results is directly determined by the inspector's experience and commitment. Long hours of work can easily lead to misjudgments or missed detections, especially subtle anomalies in light control components, which are difficult to accurately detect with the naked eye. Furthermore, manual inspection is inefficient, making it difficult to adapt to the demands of large-scale vehicle inspections. It often consumes significant manpower and time during vehicle production lines and maintenance.

[0003] Detection methods based on simple sensors can typically only obtain a single parameter of a light-control component, such as brightness or current, and cannot fully reflect the component's overall operating status. For example, some light-control components may have brightness that meets standards but the beam angle is offset. This anomaly is difficult to detect using only a single parameter, but it can cause glare to oncoming vehicles during actual driving or lead to insufficient lighting range, posing a safety hazard. Furthermore, traditional sensor detection is easily affected by environmental factors such as temperature, humidity, and electromagnetic interference, which can cause deviations in detection data and affect the judgment of the light-control component's operating status.

[0004] As vehicles become increasingly intelligent, the structure and functionality of light-control components are becoming increasingly complex. Traditional testing methods are no longer able to meet the high-precision, comprehensive testing requirements. Some high-end vehicles are equipped with advanced light-control devices such as adaptive headlights and automatic dimming systems. The operating conditions of these devices involve the coordinated coordination of multiple parameters, making comprehensive assessment difficult with traditional testing methods. If minor anomalies in light-control components are not detected and addressed promptly, they can gradually worsen, ultimately leading to component failure and potentially causing traffic accidents.

[0005] Currently, some computer vision-based inspection technologies have emerged on the market, but most of these technologies are limited to identifying surface defects in light-control components and are unable to deeply analyze their internal operating parameters. Furthermore, these technologies lack unified, standardized operating parameters as a reference, resulting in low consistency and reliability in their inspection results, hindering widespread application in the automotive manufacturing and maintenance industries. Therefore, developing a comprehensive, accurate, and efficient method for inspecting the operating status of automotive light-control components has become a pressing challenge. Summary of the Invention

[0006] The object of the present invention is to provide an automobile light control detection method and system based on image recognition to solve the problems raised in the above background technology.

[0007] To achieve the above objectives, the present invention provides an automobile light control detection method and system based on image recognition, the method comprising: Collecting real-time image data from a light control component of an automobile, performing noise reduction and enhancement processing on the real-time image data, and extracting an image of a light control component region from the real-time image data; Extracting feature points from the image of the light control component area, identifying structural features and state features of the light control component, and obtaining a feature data set; Classify the feature data set using a deep learning model to determine current working state parameters of the light control component; Comparing the current working state parameter with a preset standard working state parameter, calculating the degree of deviation between the current working state parameter and the standard working state parameter, and obtaining a deviation evaluation value; Based on the deviation evaluation value and a preset deviation threshold range, it is determined whether the automobile light control component has an abnormality, and a light control detection result report is generated.

[0008] Preferably, extracting feature points from the image of the light control component region, identifying the structural features and state features of the light control component, and obtaining a feature data set includes: Performing edge detection and contour extraction on the image of the light control component area to determine the geometric structure parameters of the light control component; Extracting surface state features of the light control component through color space conversion and texture analysis; Performing data fusion on the geometric structure parameters and the surface state features to generate a preliminary feature set; The preliminary feature set is subjected to dimensionality reduction processing to remove redundant features and obtain an optimized feature data set.

[0009] Preferably, the classifying and processing the feature data set by a deep learning model to determine the current working state parameters of the light control component includes: Constructing a deep learning model including a convolutional layer and a pooling layer, and inputting the feature data set into the deep learning model for feature mapping; Performing nonlinear transformation on the result of the feature mapping through a fully connected layer to output multiple working state categories; Performing probability calculation on the multiple working status categories of the energy using a normalized exponential function to determine the confidence level of each category; The working state category with the highest confidence is selected as the current working state parameter of the light control component.

[0010] Preferably, the comparing the current working state parameter with a preset standard working state parameter, calculating the degree of deviation between the current working state parameter and the standard working state parameter, and obtaining a deviation evaluation value includes: Obtain preset standard working state parameters and establish a parameter comparison matrix; Compare the current working state parameters with the corresponding standard parameters in the parameter comparison matrix one by one, and calculate the absolute deviation value of each parameter; Performing weighted summation on the absolute deviation values ​​to obtain a comprehensive deviation value; The deviation evaluation value is generated according to the comprehensive deviation value and a preset weight coefficient.

[0011] Preferably, judging whether the automobile light control component has an abnormality based on the deviation evaluation value and a preset deviation threshold range, and generating a light control detection result report, includes: Comparing the deviation evaluation value with a preset deviation threshold range to determine the threshold interval to which the deviation evaluation value belongs; Classifying the abnormality of the automobile light control component into abnormality levels according to the abnormality level standards corresponding to the threshold intervals; Collect the results of the abnormality level classification and the current working status parameters, and generate a preliminary detection report; The preliminary test report is subjected to data verification and format standardization processing to obtain a final light control test result report.

[0012] Preferably, comparing the deviation evaluation value with a preset deviation threshold range to determine the threshold interval to which the deviation evaluation value belongs includes: Obtaining a plurality of preset deviation threshold intervals, each of the deviation threshold intervals corresponding to a different abnormal risk level; Calculating the distance between the deviation evaluation value and each of the deviation threshold interval boundary values; Determine, based on the distance, whether the deviation evaluation value falls within a target threshold interval; The target threshold interval is marked as the threshold interval to which the deviation evaluation value belongs.

[0013] Preferably, the collecting of real-time image data from a light control component of an automobile, performing noise reduction and enhancement processing on the real-time image data, and extracting an image of a light control component region from the real-time image data comprises: Collecting real-time image data of the automobile light control component through a high-definition camera and performing grayscale processing on the real-time image data; The Gaussian filtering algorithm is used to perform noise reduction processing on the grayscale image data to remove image noise; The contrast of the noise-reduced image is enhanced by using the histogram equalization method to highlight the features of the light control components; The enhanced image is segmented using the target detection algorithm to extract the area image containing the light control component.

[0014] Preferably, the collecting of real-time image data from a light control component of an automobile, performing noise reduction and enhancement processing on the real-time image data, and extracting an image of a light control component region from the real-time image data further includes: Collecting ambient light intensity data outside the vehicle in real time, and adjusting exposure parameters and white balance parameters of the high-definition camera according to the ambient light intensity data; The real-time image data is re-collected based on the adjusted exposure parameters and white balance parameters, and noise reduction and enhancement processing is performed on the re-collected real-time image data to obtain an optimized light control component area image.

[0015] Preferably, the classifying and processing the feature data set by a deep learning model to determine the current working state parameters of the light control component further includes: Dividing the light control component area image into multiple areas to identify multiple sub-components in the light control component; Extracting feature data sets of each subcomponent respectively, and performing independent classification processing on the feature data sets of each subcomponent through a deep learning model to obtain working status parameters of each subcomponent; The working state parameters of the plurality of sub-components are subjected to correlation analysis to comprehensively determine the overall current working state parameters of the light control component.

[0016] Preferably, the present invention further includes an automobile light control detection system based on image recognition, which is used to implement the above-mentioned automobile light control detection method and system based on image recognition, and the system includes: An acquisition module is used to acquire real-time image data from a light-control component of an automobile, perform noise reduction and enhancement processing on the real-time image data, and extract an image of a light-control component region from the real-time image data; A feature extraction module is used to extract feature points from the image of the light control component area, identify the structural features and state features of the light control component, and obtain a feature data set; A state recognition module, configured to classify the feature data set using a deep learning model to determine the current working state parameters of the light control component; a deviation calculation module, configured to compare the current working state parameter with a preset standard working state parameter, calculate the degree of deviation between the current working state parameter and the standard working state parameter, and obtain a deviation evaluation value; The abnormality judgment module is used to judge whether the automobile light control component has an abnormality based on the deviation evaluation value and a preset deviation threshold range, and generate a light control detection result report.

[0017] Compared with the prior art, the present invention has the following beneficial effects: By collecting real-time image data of automotive light-control components and performing noise reduction and enhancement processing, we can effectively filter out interfering information in the image, making the extracted image of the light-control component area clearer and more complete, providing a high-quality image foundation for subsequent feature recognition. Feature point extraction and identification of structural and state characteristics in the light-control component area image comprehensively capture detailed information about the light-control component, including any physical structural deformation and uniform lighting conditions, and accurately identify and form a feature dataset.

[0018] By classifying and processing feature datasets using a deep learning model, we can fully leverage the model's powerful learning and analytical capabilities to deeply mine complex feature information, thereby accurately determining the current operating parameters of the light-control component. This approach transcends the traditional detection method's reliance on a single parameter and enables a comprehensive assessment of multiple operating parameters. By comparing the current operating parameters with the preset standard operating parameters and calculating the deviation assessment value, we can quantify the difference between the light-control component's operating state and the standard state, avoiding the errors caused by traditional detection methods that rely on subjective judgment.

[0019] Based on the deviation assessment value and the preset deviation threshold range, the optical control component is judged to determine whether there is an anomaly, and a test result report is generated, making the inspection process more standardized and regularized. This method can promptly detect subtle anomalies in the optical control component, and even those that have not yet affected the basic function of the component but have already caused deviations can be accurately identified. For automobile manufacturers, this can promptly identify potential problems with the optical control component before the vehicle leaves the factory, reducing the risk of recalls due to component failure. For maintenance companies, it provides maintenance personnel with clear test results, facilitating the rapid location of the fault point and shortening repair time.

[0020] In practical applications, this method can adapt to the testing needs of different types and models of automotive light-control components, eliminating the need for large-scale adjustments for specific models and demonstrating strong versatility. Furthermore, the entire testing process is highly automated, reducing manual intervention and the impact of human factors on test results, making them more objective and reliable. Furthermore, this method can capture and process image data in real time, rapidly generating test result reports and improving testing efficiency. This approach can meet the timeliness requirements for light-control component testing in automotive production lines and during repair and maintenance. BRIEF DESCRIPTION OF THE DRAWINGS

[0021] Figure 1 This is a working principle diagram of the automobile light control detection method based on image recognition according to the present invention; Figure 2 Flowchart for feature point extraction and feature dataset generation; Figure 3 Flowchart of classification processing for deep learning models; Figure 4 Flowchart for abnormality judgment and detection report generation; Figure 5 Flowchart for deviation threshold interval matching. DETAILED DESCRIPTION

[0022] The following will clearly and completely describe the technical solutions in the embodiments of the present invention in conjunction with the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of them. All other embodiments obtained by ordinary technicians in this field based on the embodiments of the present invention without making any creative efforts shall fall within the scope of protection of the present invention.

[0023] See also Figure 1 The present invention provides an automobile light control detection method and system based on image recognition, and the method includes:

[0024] Real-time image data from a vehicle's light-control components is collected, noise reduction and enhancement processing is performed on the real-time image data, and images of the light-control component regions within the real-time image data are extracted. During this process, an image acquisition device is used to capture real-time images of the vehicle's light-control components under different operating conditions. Noise interference is removed using a corresponding noise reduction algorithm, and enhancement techniques are used to enhance the clarity of the light-control component details in the image. The processed images are then used to precisely extract images of the region containing only the light-control components.

[0025] Feature points are extracted from the image of the light-control component area to identify its structural and state characteristics, generating a feature dataset. Using specific image analysis methods, key feature points reflecting its structure and state are captured from the extracted image of the light-control component area. The information corresponding to these feature points is then integrated to form a feature dataset for subsequent analysis.

[0026] The feature dataset is classified and processed using a deep learning model to determine the current operating state parameters of the light-control component. The resulting feature dataset is input into a pre-trained deep learning model, which analyzes and classifies the features, determines the current operating state of the light-control component based on the classification results, and outputs the corresponding parameters.

[0027] The current operating state parameters are compared with preset standard operating state parameters, and the degree of deviation between the current operating state parameters and the standard operating state parameters is calculated to obtain a deviation assessment value. The standard operating state parameters of the light control components preset in the system are called, and the current operating state parameters determined in real time are compared with the standard parameters one by one. The degree of deviation between the two is calculated using a specific calculation method and used as the deviation assessment value.

[0028] Based on the deviation assessment value and the preset deviation threshold range, the system determines whether the vehicle's light-control components are abnormal, and generates a light-control test result report. The calculated deviation assessment value is matched with different preset deviation threshold ranges. Based on the threshold range to which it falls, the system determines whether the light-control components are abnormal and the severity of the abnormality. Finally, all relevant information is compiled into a light-control test result report.

[0029] Example 1: See Figure 2 , extract feature points from the image of the light-control component area, identify the structural features and state features of the light-control component, and when obtaining the feature data set, first perform edge detection and contour extraction on the image of the light-control component area. Scan the image of the light-control component area pixel by pixel, calculate the grayscale value difference between each pixel and its neighboring pixels, and when the difference value exceeds the set threshold, mark the pixel as an edge point. Continuous edge lines are formed by connecting adjacent edge points, and then the complete outline of the light-control component is outlined based on the distribution of the edge lines. The geometric structural parameters of the light-control component are extracted from the contour, including the overall contour dimensions of the component, such as linear parameters such as length, width, and height, as well as the position coordinates, shape parameters (such as the radius of a circular structure, the side length ratio of a rectangular structure), and the relative distance and angular relationship between the structures of the holes and protrusions on the component.

[0030] The surface characteristics of the light-control component are extracted through color space conversion and texture analysis. The image of the light-control component region is converted from RGB color space (red, green, and blue) to HSV color space (hue, saturation, and value), separating the three components: hue (H), saturation (S), and value (V). A statistical analysis of the hue component determines the hue range of the primary color on the light-control component's surface; a distribution calculation of the saturation component is performed to determine the vividness of the surface color; and a gradient analysis of the value component is performed to determine the surface's brightness and darkness variations. In texture analysis, a sliding window is used to traverse the image, calculating the grayscale co-occurrence matrix of each pixel within the window. Texture feature parameters such as energy, entropy, contrast, and correlation are extracted from the matrix to reflect the surface roughness, texture distribution patterns, and the presence of scratches, wear, and other state characteristics.

[0031] The geometric parameters and surface characteristics are fused to generate a preliminary feature set. A unified data format is established, arranging the linear, shape, and positional parameters of the geometric parameters with the color component and texture parameters of the surface characteristics in a pre-set order. Different parameter types are standardized to bring all values ​​within the same range to avoid the impact of differences in parameter magnitude on subsequent processing. All standardized parameters are combined to form a preliminary feature set containing comprehensive information about the light-control components.

[0032] The preliminary feature set is subjected to dimensionality reduction processing, redundant features are removed, and an optimized feature data set is obtained. The preliminary feature set is processed using the principal component analysis method, the covariance matrix of all parameters in the feature set is calculated, and the eigenvalues ​​and eigenvectors of the covariance matrix are solved. The importance of the principal component is determined according to the size of the eigenvalue, and the first several principal components with larger eigenvalues ​​are selected as new feature variables. These principal components can cover most of the information in the preliminary feature set while reducing the number of features. By calculating the load of each original feature on the selected principal component, the contribution of the original feature to the principal component is analyzed, and the original features with lower contribution are eliminated. The remaining original features are combined with the selected principal component to form an optimized feature data set, which not only retains the key feature information of the light control component but also reduces the dimension and redundancy of the data.

[0033] Example 2: See Figure 3To classify the feature dataset using a deep learning model and determine the current operating state parameters of the light-control component, a deep learning model consisting of convolutional and pooling layers is constructed. The convolutional layer uses multiple convolution kernels with sizes ranging from 1×1, 3×3, and 5×5. Each kernel performs a sliding convolution operation on the input feature dataset, extracting local features from the feature data, such as texture edges and geometric shape changes on the light-control component's surface. During the convolution operation, each kernel is element-wise multiplied by the corresponding region of the feature data and the sum is calculated to generate a corresponding feature map. The pooling layer uses max pooling or average pooling to downsample the feature map output by the convolutional layer. This involves partitioning the feature map into a preset pooling window size (e.g., 2×2) and selecting the maximum or average value within each window as the output value for that window, thereby reducing the size and data volume of the feature map. The feature dataset is then input into the deep learning model for feature mapping. After multiple layers of convolution and pooling operations, a high-level feature representation that reflects the essential characteristics of the light-control component is obtained.

[0034] The feature mapping results are nonlinearly transformed through a fully connected layer, which consists of multiple neurons. Each neuron is connected to all neurons in the previous layer, and each connection corresponds to a weight parameter. After the feature mapping results are input into the fully connected layer, they are weighted and summed with the weight parameters. They are then processed through an activation function (such as a linear rectification function or a sigmoid function) to achieve nonlinear transformation. The feature mapping results are converted into an output vector related to the operating state categories of the light-control component. The dimensions of this output vector match the number of possible operating state categories, and the value of each dimension corresponds to the raw score of a working state category, thereby outputting multiple possible working state categories.

[0035] A normalized exponential function is used to calculate the probability of multiple possible working state categories, converting the raw scores in the output vector into probability values. Specifically, an exponential operation is performed on each value in the output vector, and each exponential operation result is divided by the sum of all exponential operation results to obtain a probability value for each working state category. This probability value represents the likelihood that the light control component is in the corresponding working state category, which is used to determine the confidence level of each category.

[0036] The working state category with the highest confidence is selected as the current working state parameter of the light control component.

[0037] In addition, the image of the light-control component region is segmented into multiple subregions based on the component's structural composition and according to pre-set segmentation rules. Each subregion corresponds to a subcomponent within the light-control component, such as the light source module, adjustment knob, or circuit interface. A feature dataset is extracted for each subcomponent. For each subregion, the same methods used for extracting overall light-control component features, including edge detection, contour extraction, color analysis, and texture analysis, are used to obtain the geometric structural parameters and surface characteristics of each subcomponent. After data fusion and dimensionality reduction, a feature dataset is generated for each subcomponent. Each subcomponent feature dataset is independently classified using a deep learning model. Each subcomponent feature dataset is then input into the constructed deep learning model. After feature mapping, nonlinear transformation, and probability calculation, the operating state parameters of each subcomponent are obtained. Correlation analysis is performed on the operating state parameters of multiple subcomponents to analyze the inherent connections and mutual influences between their operating states. For example, the matching relationship between the brightness of the light source module and the position of the adjustment knob, and the influence of the connection status of the circuit interface on the operating state of the light source module, are analyzed. The overall current operating state parameters of the light-control component are determined by combining the operating state parameters of all subcomponents.

[0038] Example 3: See Figure 4 Based on the deviation assessment value and the preset deviation threshold range, it is determined whether there is any abnormality in the vehicle's light control components. When generating the light control test result report, the deviation assessment value is compared with the preset deviation threshold range. The preset deviation threshold range consists of multiple continuous and non-overlapping intervals. The division of these intervals is based on the design standards, operating specifications and historical test data of the light control components. Each interval corresponds to a different abnormality judgment standard. For example, a certain interval may correspond to the light control component being in normal working condition, another interval may correspond to the existence of a minor abnormality, and another interval may correspond to the existence of a serious abnormality, etc. By comparing the deviation assessment value with the upper and lower boundaries of these intervals one by one, it is determined in which threshold interval the deviation assessment value specifically falls.

[0039] Abnormalities in the vehicle's light-control components are classified according to the abnormality level standards corresponding to the threshold intervals. Different threshold intervals have pre-set abnormality levels, including no abnormality, level 1 abnormality, level 2 abnormality, and level 3 abnormality, each representing the severity of the abnormality in the light-control component. Once the threshold interval to which the deviation assessment value belongs is determined, the abnormality in the light-control component is classified into the corresponding level according to the abnormality level standards corresponding to that interval. For example, if the deviation assessment value falls within the threshold interval corresponding to no abnormality, the light-control component is determined to have no abnormality; if it falls within the threshold interval corresponding to level 2 abnormality, the light-control component is determined to have a level 2 abnormality.

[0040] Collect the abnormality classification results and current operating status parameters to generate a preliminary test report. The abnormality classification results include the name or code of the abnormality level and the corresponding threshold range information. The current operating status parameters include the various operating parameters and characteristic data of the light-control component. This information is organized in a logical order, such as first listing the basic information of the light-control component, then presenting the current operating status parameters, deviation assessment values, corresponding threshold ranges, and abnormality classification results in sequence to form a preliminary test report. During this organization process, ensure the integrity of the information and do not omit any critical data.

[0041] Perform data verification and format standardization on the preliminary test report. During data verification, check whether the data in the report is accurate, for example, whether the value of the current working status parameter is consistent with the actual test result, whether the calculation of the deviation assessment value is correct, whether the abnormality level division corresponds to the threshold range, etc. At the same time, check whether the logical relationship between the data is reasonable, such as whether there is a contradiction between the values ​​of different parameters, whether the correspondence between the abnormality level and the deviation assessment value complies with the preset rules, etc. Format standardization is to adjust the preliminary test report according to unified format requirements, including font type and size, line spacing, paragraph format, data alignment, etc., so that the report layout is neat and standardized, easy to read and understand. After data verification and format standardization, the final light control test result report is obtained.

[0042] During the anomaly classification process, the following method is used to quantify the relationship between anomaly level and deviation assessment value:

[0043] in, Represents the abnormal level value, represents the deviation evaluation value, represents the interval coefficient, Represents the interval reference value. Interval coefficient and interval benchmark values According to the range setting of different threshold intervals, each threshold interval corresponds to a specific set of and The deviation evaluation value is converted into the corresponding abnormality level value through this formula, and then the specific abnormality level is determined.

[0044] Example 4: See Figure 5, compare the deviation assessment value with the preset deviation threshold range, and when determining the threshold interval to which the deviation assessment value belongs, obtain multiple preset deviation threshold intervals. These intervals are defined based on the performance parameters, operating requirements, and historical detection records of the light control components. Each interval has clear upper and lower boundary values, and there is no overlap between intervals. For example, for the light control components of automobile headlights, the preset deviation threshold intervals include [0,5), [5,10), [10,15), [15,20), etc. Each interval corresponds to a different abnormal risk level, such as no risk, low risk, medium risk, and high risk.

[0045] Calculate the distance between the deviation assessment value and each deviation threshold interval boundary. For each interval's upper and lower boundaries, calculate the absolute value of the difference between the deviation assessment value and the boundary value. For example, if the deviation assessment value is 8, for the interval [0,5), its lower boundary is 0 and its upper boundary is 5. The calculated distance between 8 and 0 is 8, and the distance between 8 and 5 is 3. For the interval [5,10), its lower boundary is 5 and its upper boundary is 10. The calculated distance between 8 and 5 is 3, and the distance between 8 and 10 is 2. For the interval [10,15), its lower boundary is 10 and its upper boundary is 15. The calculated distance between 8 and 10 is 2, and the distance between 8 and 15 is 7. For the interval [15,20), its lower boundary is 15 and its upper boundary is 20. The calculated distance between 8 and 15 is 7, and the distance between 8 and 20 is 12.

[0046] The target threshold interval into which the deviation assessment value falls is determined based on the distance. By comparing the distance between the deviation assessment value and the boundaries of each interval, the position of the deviation assessment value is analyzed. When the deviation assessment value is greater than or equal to the lower boundary of a certain interval and less than the upper boundary of the interval, it falls into that interval. Taking the deviation assessment value of 8 as an example, it is greater than the lower boundary of the interval [5,10) by 5 and less than the upper boundary of the interval by 10, so it is determined to fall into the interval [5,10). If the deviation assessment value is equal to the boundary value of a certain interval, such as the deviation assessment value of 5, then the interval to which it belongs is determined according to the preset rules, for example, it is classified into the interval [5,10).

[0047] Mark the target threshold interval as the threshold interval to which the deviation assessment value belongs. After determining the interval to which the deviation assessment value belongs, mark the interval, for example, by identifying it with its upper and lower boundary values ​​or the corresponding abnormal risk level. For example, mark the interval [5, 10) to which the deviation assessment value belongs as "low-risk interval" and store this marking information in the detection system database as the basis for subsequent abnormality level determination and detection report generation. During the marking process, ensure that the marking information is accurately associated with the deviation assessment value, interval boundary values, and other data to avoid confusion or errors.

[0048] Example 5: Real-time image data from a car's light-control components is collected, noise reduction and enhancement processing is performed on the real-time image data, and when extracting images of the light-control component area from the real-time image data, the real-time image data of the car's light-control components is collected using a high-definition camera. The high-definition camera is installed at a preset position inside the car, with the lens facing the light-control component to ensure that the shooting range can fully cover all parts of the light-control component. During the acquisition process, the camera continuously shoots at a set frame rate to obtain image data of the light-control component at different working stages. The real-time image data is grayscaled, and the RGB three-color value of each pixel in the color image is converted into a single grayscale value. The conversion method is to fuse the values ​​of the three RGB channels at a specific ratio to form a grayscale image, thereby reducing the amount of data for subsequent processing.

[0049] A Gaussian filter algorithm is used to reduce noise in grayscale image data. Based on the noise distribution characteristics of the image, a Gaussian kernel of appropriate size is selected, determined by the size of the noise particles. The Gaussian kernel is then slid across the grayscale image, replacing the grayscale value of each pixel with the sum of the weight of that pixel multiplied by the Gaussian kernel's corresponding position. This smoothing process weakens the noise signal in the image, resulting in a softer overall image.

[0050] The contrast of the denoised image is enhanced through histogram equalization. The grayscale histogram of the denoised image is calculated to determine the frequency of occurrence of different grayscale values. The cumulative distribution function of the grayscale values ​​is then calculated and mapped to a new grayscale range using the cumulative distribution function. This results in a more uniform distribution of grayscale values ​​within the image, revealing details in previously low-contrast areas and sharper edges and textures of light-control components.

[0051] The enhanced image is segmented using an object detection algorithm. Using a pre-trained model, the algorithm identifies the outline of the light-control component in the image and determines its position and size. Based on the recognition results, the enhanced image is cropped to capture only the light-control component, creating an image of the light-control component region and eliminating any interference from irrelevant background.

[0052] Simultaneously, real-time data on the vehicle's external ambient light intensity is collected via a light sensor mounted on the vehicle's exterior. This sensor can detect changes in ambient light intensity at different times and under different weather conditions, and transmits this data in real time to the image processing system. The high-definition camera's exposure and white balance parameters are adjusted based on this ambient light intensity data. When the ambient light intensity is high, the camera's exposure time and sensitivity are reduced to prevent the image from being too bright, which could lead to loss of detail in the light-control components. When the ambient light intensity is low, the exposure time and sensitivity are increased to prevent the image from being too dark, which could affect feature extraction. The white balance parameters are adjusted based on the ambient light's color temperature. When the ambient light is cool, the white balance parameters are adjusted to warm the image; when the ambient light is warm, the parameters are adjusted to cool the image, ensuring that the captured color of the light-control components matches the actual color.

[0053] Based on the adjusted exposure and white balance parameters, real-time image data is recaptured and the aforementioned noise reduction and enhancement processes are performed on the recaptured real-time image data, namely, Gaussian filtering for noise reduction and histogram equalization for enhancement. This ultimately yields an optimized image of the light-control component area. This image can adapt to varying external lighting environments, accurately depicting the structure and state characteristics of the light-control component, providing reliable image data for subsequent feature extraction and state recognition.

[0054] It should be noted that, in this document, relational terms such as first and second, etc., are used to distinguish one entity or operation from another entity or operation, and do not necessarily require or imply any actual relationship or order between these entities or operations. Moreover, the terms "comprises," "comprising," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that includes a list of elements includes not only those elements but also other elements not explicitly listed, or elements inherent to such process, method, article, or apparatus.

[0055] Although the embodiments of the present invention have been shown and described, it will be apparent to those skilled in the art that various changes, modifications, substitutions and alterations may be made to these embodiments without departing from the principles and spirit of the invention, and the scope of the invention is defined by the appended claims and their equivalents.

Claims

1. A method and system for detecting automobile light control based on image recognition, characterized in that: include: Collecting real-time image data from a light control component of an automobile, performing noise reduction and enhancement processing on the real-time image data, and extracting an image of a light control component region from the real-time image data; Extracting feature points from the image of the light control component area, identifying structural features and state features of the light control component, and obtaining a feature data set; Classify the feature data set using a deep learning model to determine current working state parameters of the light control component; Comparing the current working state parameter with a preset standard working state parameter, calculating the degree of deviation between the current working state parameter and the standard working state parameter, and obtaining a deviation evaluation value; Based on the deviation evaluation value and a preset deviation threshold range, it is determined whether the automobile light control component has an abnormality, and a light control detection result report is generated.

2. The method and system for automobile light control detection based on image recognition according to claim 1, characterized in that: The extracting feature points from the image of the light control component region, identifying the structural features and state features of the light control component, and obtaining a feature data set includes: Performing edge detection and contour extraction on the image of the light control component area to determine the geometric structure parameters of the light control component; Extracting surface state features of the light control component through color space conversion and texture analysis; Performing data fusion on the geometric structure parameters and the surface state features to generate a preliminary feature set; The preliminary feature set is subjected to dimensionality reduction processing to remove redundant features and obtain an optimized feature data set.

3. The method and system for automobile light control detection based on image recognition according to claim 2, characterized in that: The classifying and processing the feature data set by using a deep learning model to determine the current working state parameters of the light control component includes: Constructing a deep learning model including a convolutional layer and a pooling layer, and inputting the feature data set into the deep learning model for feature mapping; Performing nonlinear transformation on the result of the feature mapping through a fully connected layer to output multiple working state categories; Performing probability calculation on the multiple working status categories using a normalized exponential function to determine the confidence level of each category; The working state category with the highest confidence is selected as the current working state parameter of the light control component.

4. The method and system for detecting automobile light control based on image recognition according to claim 3, characterized in that: The comparing the current working state parameter with the preset standard working state parameter, calculating the deviation between the current working state parameter and the standard working state parameter, and obtaining the deviation evaluation value includes: Obtain preset standard working state parameters and establish a parameter comparison matrix; Compare the current working state parameters with the corresponding standard parameters in the parameter comparison matrix one by one, and calculate the absolute deviation value of each parameter; Performing weighted summation on the absolute deviation values ​​to obtain a comprehensive deviation value; The deviation evaluation value is generated according to the comprehensive deviation value and a preset weight coefficient.

5. The method and system for automobile light control detection based on image recognition according to claim 1, characterized in that: The step of determining whether the automobile light control component has an abnormality based on the deviation evaluation value and a preset deviation threshold range, and generating a light control detection result report, includes: Comparing the deviation evaluation value with a preset deviation threshold range to determine the threshold interval to which the deviation evaluation value belongs; Classifying the abnormality of the automobile light control component into abnormality levels according to the abnormality level standards corresponding to the threshold intervals; Collect the results of the abnormality level classification and the current working status parameters, and generate a preliminary detection report; The preliminary test report is subjected to data verification and format standardization processing to obtain a final light control test result report.

6. The method and system for automobile light control detection based on image recognition according to claim 5, characterized in that: The comparing the deviation evaluation value with a preset deviation threshold range to determine the threshold interval to which the deviation evaluation value belongs includes: Obtaining a plurality of preset deviation threshold intervals, each of the deviation threshold intervals corresponding to a different abnormal risk level; Calculating the distance between the deviation evaluation value and each of the deviation threshold interval boundary values; Determine, based on the distance, whether the deviation evaluation value falls within a target threshold interval; The target threshold interval is marked as the threshold interval to which the deviation evaluation value belongs.

7. The method and system for automobile light control detection based on image recognition according to claim 1, characterized in that: The collecting of real-time image data from the automobile light control component, performing noise reduction and enhancement processing on the real-time image data, and extracting the light control component area image from the real-time image data include: Collecting real-time image data of the automobile light control component through a high-definition camera and performing grayscale processing on the real-time image data; The Gaussian filtering algorithm is used to perform noise reduction processing on the grayscale image data to remove image noise; The contrast of the noise-reduced image is enhanced by using the histogram equalization method to highlight the features of the light control components; The enhanced image is segmented using the target detection algorithm to extract the area image containing the light control component.

8. The method and system for automobile light control detection based on image recognition according to claim 7, characterized in that: The collecting of real-time image data from the automobile light control component, performing noise reduction and enhancement processing on the real-time image data, and extracting the light control component area image from the real-time image data further includes: Collecting ambient light intensity data outside the vehicle in real time, and adjusting exposure parameters and white balance parameters of the high-definition camera according to the ambient light intensity data; The real-time image data is re-collected based on the adjusted exposure parameters and white balance parameters, and noise reduction and enhancement processing is performed on the re-collected real-time image data to obtain an optimized light control component area image.

9. The method and system for detecting automobile light control based on image recognition according to claim 3, characterized in that: The classifying and processing the feature data set by the deep learning model to determine the current working state parameters of the light control component further includes: Dividing the light control component area image into multiple areas to identify multiple sub-components in the light control component; Extracting feature data sets of each subcomponent respectively, and performing independent classification processing on the feature data sets of each subcomponent through a deep learning model to obtain working status parameters of each subcomponent; The working state parameters of the plurality of sub-components are subjected to correlation analysis to comprehensively determine the overall current working state parameters of the light control component.

10. An automobile light control detection system based on image recognition, used to implement the automobile light control detection method and system based on image recognition according to any one of claims 1 to 9, characterized in that: include: An acquisition module is used to acquire real-time image data from a light-control component of an automobile, perform noise reduction and enhancement processing on the real-time image data, and extract an image of a light-control component region from the real-time image data; A feature extraction module is used to extract feature points from the image of the light control component area, identify the structural features and state features of the light control component, and obtain a feature data set; A state recognition module, configured to classify the feature data set using a deep learning model to determine the current working state parameters of the light control component; a deviation calculation module, configured to compare the current working state parameter with a preset standard working state parameter, calculate the degree of deviation between the current working state parameter and the standard working state parameter, and obtain a deviation evaluation value; The abnormality judgment module is used to judge whether the automobile light control component has an abnormality based on the deviation evaluation value and a preset deviation threshold range, and generate a light control detection result report.

Images