Defect detection method and system based on lamp strip in automobile sun shield

Abstract

The invention discloses a defect detection method and system based on an internal lamp strip of an automobile sun shield, and relates to the technical field of image processing, and the method comprises the steps: carrying out the feature extraction processing of a lamp strip image set containing different defect features, and obtaining the different defect features of the lamp strip. According to the invention, feature extraction and spectrum analysis are carried out on a lamp strip image set containing different defect features to obtain different types of defect features and spectrum data of different defect types, and then, a convolutional neural network model is trained according to the different types of defect features and the spectrum data of different defect types to obtain a convolutional neural network model; according to the method, the lamp strip defect identification model is determined, when the defect of the lamp strip needs to be identified, the type of the defect of the lamp strip can be determined only by inputting the to-be-verified lamp strip defect image into the lamp strip defect identification model, and then whether the lamp strip is put into use or not can be selected according to the type of the defect of the lamp strip. According to the mode, the working intensity of workers is reduced, and the production cost of the lamp strip is also reduced.

Description

Technical Field

[0001] This invention relates to the field of image processing technology, specifically to a method and system for detecting defects in the internal light strip of an automotive sun visor. Background Technology

[0002] The internal light strip of a car sun visor usually refers to the lighting strip around the vanity mirror of the sun visor. It is mainly used to provide local lighting at night or when there is insufficient light, making it convenient for applying makeup, tidying up one's appearance, or finding small items.

[0003] Since there are many types of defects in LED strips, not all defective LED strips are unusable. Discarding all defective LED strips would increase production costs. However, manually analyzing LED strip defects and determining their types would be time-consuming and prone to errors. Summary of the Invention

[0004] To address the aforementioned technical problems, a defect detection method and system based on the internal light strip of an automotive sun visor is provided. This technical solution solves the problems mentioned in the background section.

[0005] To achieve the above objectives, the technical solution adopted by the present invention is as follows: A defect detection method based on the internal light strip of an automotive sun visor includes: A set of LED strip images containing different defect features is obtained. Based on image processing algorithms, feature extraction processing is performed on the set of LED strip images containing different defect features to obtain different defect features of the LED strips. Based on the Fast Fourier Transform algorithm, frequency domain information analysis is performed on a set of light strip images containing different defect features to determine the spectral information of different defect features. The different defect characteristics and spectral information of the LED strip are integrated and processed to obtain a training set of LED strip defects; Based on the LED strip defect training set, a convolutional neural network model is trained to obtain an LED strip defect recognition model. The defect image of the LED strip to be verified is obtained. Based on the LED strip defect recognition model, the defect image of the LED strip to be verified is recognized and verified to determine the defect type of the LED strip.

[0006] Preferably, the step of obtaining a set of light strip images containing different defect features, based on an image processing algorithm, involves feature extraction processing on the set of light strip images containing different defect features to obtain the different defect features of the light strips, specifically including the following steps: Data retrieval and processing are performed on the database system to obtain a set of light strip images containing different defect characteristics; Data reading and processing are performed on images from a set of light strip images containing different defect characteristics to obtain image capture parameters; Data matching processing is performed on the image capture parameters to determine the type of noise in the light strip image; Select a Gaussian filter based on the type of noise in the LED strip image; Based on a Gaussian filter, images in a set of light strip images containing different defect features are denoised to obtain a denoised light strip image set. Image analysis and processing are performed on the data in the denoised LED strip image set to determine the different defect characteristics of the LED strip.

[0007] Preferably, the step of performing image analysis processing on the data in the denoised LED strip image set to determine the different defect features of the LED strip specifically includes the following steps: Based on the Laplacian operator, the images in the denoised light strip image set are processed to determine the edge positions of different defect features of the light strip; Based on the Canny algorithm, edge differentiation processing is performed on the edge positions of different defect features of the light strip to determine the true edge positions of different defect features of the light strip. Based on the true edge location of different defect features of the light strip, feature extraction is performed on the data in the set of denoised light strip images to obtain different defect features of the light strip; The specific calculation formula for determining the edge position of different defect features of the light strip is as follows: In the formula, The edge locations of different defect characteristics of the light strip; The grayscale values ​​of the images in the set of denoised light strip images; These are the pixel coordinates of the images in the set of denoised LED strip images.

[0008] Preferably, the step of performing frequency domain information analysis on a set of light strip images containing different defect features based on the Fast Fourier Transform algorithm to determine the spectral information of different defect features specifically includes the following steps: Analog-to-digital conversion is performed on a set of LED strip images containing different defect features to obtain the digital signal corresponding to each LED strip defect image; The digital signal corresponding to each LED strip defect image is preprocessed to obtain the preprocessed digital signal of each LED strip defect image. Based on the Fast Fourier Transform algorithm, the preprocessed digital signal of each LED strip defect image is subjected to frequency domain transformation to obtain the frequency domain signal of each LED strip defect image, i.e.: In the formula, The frequency domain signal for each LED strip defect image; The digital signal after preprocessing for each LED strip defect image; Based on a bandpass filter, feature extraction processing is performed on the frequency domain signal of each LED strip defect image to obtain spectral information of different defect features.

[0009] Preferably, the step of preprocessing the digital signal corresponding to each LED strip defect image to obtain the preprocessed digital signal of each LED strip defect image specifically includes the following steps: Based on a high-pass filter, noise removal processing is performed on the digital signal corresponding to each LED strip defect image to obtain a noise-free digital signal for each LED strip defect image. The noise-free digital signal of each LED strip defect image is sampled and processed to obtain the sampled digital signal of each noise-free image; The digital signal of each noise-free image is smoothed by using a moving average method to obtain the preprocessed digital signal of each LED strip defect image.

[0010] Preferably, the step of training the convolutional neural network model based on the LED strip defect training set to obtain the LED strip defect recognition model specifically includes the following steps: The set of LED strip images containing different defect features is set as the validation set for the LED strip defect recognition model; Based on the training set of LED strip defects, the convolutional neural network model is trained to obtain the model output results to be verified. Based on the validation set of the LED strip defect identification model, structural verification processing is performed on the output results of the model to obtain the LED strip defect verification results. The results of LED strip defect verification were compared and analyzed to determine the LED strip defect identification model.

[0011] Preferably, the comparative analysis of the LED strip defect verification results to determine the LED strip defect identification model specifically includes the following steps: Data matching processing is performed on the verification results of LED strip defects and the set of LED strip images containing different defect features; If the defect verification result of the LED strip is the same as the image defect features in the LED strip image set containing different defect features, the LED strip defect recognition model is determined. If the defect verification result of the LED strip is different from the image defect features in the LED strip image set containing different defect features, a new set of LED strip images containing different defect features is selected to train the convolutional neural network model and determine the LED strip defect recognition model.

[0012] Preferably, the steps of acquiring the LED strip defect image to be verified, and performing identification and verification processing on the LED strip defect image based on the LED strip defect recognition model to determine the LED strip defect type specifically include the following steps: Based on the shooting equipment, image processing is performed on the defective light strip to obtain the defective image of the light strip to be verified; Based on the LED strip defect recognition model, feature extraction and signal analysis are performed on the LED strip defect image to be verified to obtain the defect features and the spectrum of the LED strip defect image to be verified. Based on the different defect features and spectral information of the LED strip, data matching processing is performed on the defect features of the LED strip to be verified and the spectrum of the defect image of the LED strip to be verified to obtain the LED strip defect feature recognition result and the LED strip defect feature spectrum recognition result. By comparing and analyzing the results of LED strip defect feature identification and LED strip defect feature spectrum identification, the type of LED strip defect is determined.

[0013] Preferably, the step of comparing and analyzing the LED strip defect feature identification results and the LED strip defect feature spectrum identification results to determine the LED strip defect type specifically includes the following steps: If the LED strip defect feature identification result is that the defect feature to be verified is the same as one of the different defect features of the LED strip, and the LED strip defect feature spectrum identification result is that the spectrum of the LED strip defect image to be verified is the same as one of the spectrum information of the different defect features, then the LED strip defect type is determined. If the defect feature recognition result of the light strip is that the defect feature to be verified is different from the different defect features of the light strip, or if the spectrum recognition result of the defect feature is that the spectrum of the defect image of the light strip to be verified is different from the spectrum information of different defect features, the defect recognition accuracy of the light strip defect recognition model is low, and the light strip defect recognition model should be retrained.

[0014] Furthermore, a defect detection system based on the internal light strip of an automotive sun visor is proposed to implement the defect detection method based on the internal light strip of an automotive sun visor as described above, including: A defect detection terminal is used to control the data transmission and information interaction between various modules. The defect detection terminal is also used to control the various modules to perform model training processing on the LED strip defect identification model to determine the LED strip defect type. A database system for storing a set of light strip images containing different defect features; A noise removal module is used to perform noise removal processing on images in a set of light strip images containing different defect features to obtain a set of denoised light strip images. The feature extraction module is used to perform feature extraction processing on the set of denoised light strip images to obtain different defect features of the light strip; A defect spectrum determination module is used to perform frequency domain information analysis and processing on a set of light strip images containing different defect features to determine the spectrum information of different defect features. The model training module trains the convolutional neural network model based on the LED strip defect training set to obtain an LED strip defect recognition model. The defect type determination module performs identification and verification processing on the LED strip defect image to be verified based on the LED strip defect identification model to determine the LED strip defect type.

[0015] Compared with the prior art, the present invention provides a defect detection method and system based on the internal light strip of an automotive sun visor, which has the following beneficial effects: This invention extracts features and performs spectral analysis on a set of LED strip images containing different defect characteristics to obtain different types of defect features and spectral data of different defect types. Then, a convolutional neural network model is trained based on the different types of defect features and spectral data of different defect types to determine the LED strip defect recognition model. When it is necessary to identify defects in the LED strip, simply input the LED strip defect image to be verified into the LED strip defect recognition model to determine the type of LED strip defect. Then, based on the type of LED strip defect, it can be selected whether to put the LED strip into use. The above method not only reduces the workload of workers, but also reduces the production cost of LED strips. Attached Figure Description

[0016] Figure 1 This is a flowchart illustrating steps S100-S500 of a defect detection method based on an internal light strip in an automotive sun visor proposed in this invention. Figure 2 This is a structural block diagram of a defect detection system based on the internal light strip of an automotive sun visor, as proposed in this invention. Detailed Implementation

[0017] The following description is intended to disclose the invention and enable those skilled in the art to implement it. The preferred embodiments described below are merely examples, and other obvious variations will occur to those skilled in the art.

[0018] Reference Figure 1 As shown, a defect detection method based on the internal light strip of an automotive sun visor includes: S100. Obtain a set of light strip images containing different defect features. Based on image processing algorithms, perform feature extraction processing on the set of light strip images containing different defect features to obtain different defect features of the light strip. S200. Based on the Fast Fourier Transform algorithm, frequency domain information analysis and processing are performed on a set of light strip images containing different defect features to determine the spectral information of different defect features. S300: Integrate and process the different defect characteristics and spectral information of the light strip to obtain a training set of light strip defects; S400. Based on the LED strip defect training set, perform model training on the convolutional neural network model to obtain the LED strip defect recognition model. S500: Obtain the image of the LED strip defect to be verified; based on the LED strip defect recognition model, perform recognition and verification processing on the image of the LED strip defect to be verified to determine the type of LED strip defect. Those skilled in the art will understand that, due to the variety of defect types in LED strips, some defects do not affect their normal use. Destroying all defective LED strips would increase production costs. Therefore, it is necessary to screen out LED strips that contain defects but do not affect normal use. However, manually judging and screening LED strips not only increases the workload of workers but also increases the misjudgment rate. This is because workers' attention will decrease after a long period of defect judgment, which in turn increases the misjudgment rate. Therefore, training a convolutional neural network model with a set of LED strip images containing different defect features to determine an LED strip defect recognition model, and using this model to identify defect types in LED strips, can not only reduce the workload of workers but also reduce the misjudgment rate.

[0019] Step S100: Obtain a set of light strip images containing different defect features. Based on image processing algorithms, perform feature extraction processing on the set of light strip images containing different defect features to obtain the different defect features of the light strips. Specifically, this includes the following steps: S101. Perform data retrieval processing on the database system to obtain a set of light strip images containing different defect characteristics; S102. Perform data reading and processing on the images in the set of light strip images containing different defect characteristics to obtain image shooting parameters; S103. Perform data matching processing on the image shooting parameters to determine the type of noise in the light strip image; It is understandable that different image capturing parameters will produce different noise. Therefore, the noise type is determined by the image capturing parameters, and then a Gaussian filter is selected based on the noise type. S104. Select a Gaussian filter based on the type of noise in the LED strip image; S105. Based on a Gaussian filter, perform denoising processing on the images in the set of light strip images containing different defect features to obtain a denoised light strip image set. To make image analysis more accurate, a Gaussian filter is used to remove noise from images in a set of light strip images containing different defect features, so as to avoid interfering with the training of the subsequent light strip defect recognition model and make the recognition results of the light strip defect recognition model more accurate. S106. Perform image analysis processing on the data in the noise-reduced LED strip image set to determine the different defect characteristics of the LED strip.

[0020] Specifically, step S106, which involves performing image analysis processing on the data in the denoised LED strip image set to determine the different defect features of the LED strip, includes the following steps: S1061. Based on the Laplacian operator, perform calculations on the images in the denoised light strip image set to determine the edge positions of different defect features of the light strip. S1062. Based on the Canny algorithm, perform edge differentiation processing on the edge positions of different defect features of the light strip to determine the true edge positions of different defect features of the light strip. S1063. Based on the real edge positions of different defect features of the light strip, feature extraction is performed on the data in the set of denoised light strip images to obtain different defect features of the light strip. The specific calculation formula for determining the edge position of different defect features of the light strip is as follows: In the formula, The edge locations of different defect characteristics of the light strip; The grayscale values ​​of the images in the set of denoised light strip images; These are the pixel coordinates of the images in the set of denoised LED strip images; It is understandable that noise can also cause changes in brightness. Therefore, there may be brightness changes caused by noise at the edge positions of different defect features of the light strip. So, the Canny algorithm is used to remove the brightness changes caused by noise, making the defect features of the subsequent light strip more complete. At the same time, it makes the training of the subsequent light strip defect recognition model more accurate. In this embodiment, when there is a defect on the surface of the light strip, there will be a difference between the defect and the surface of the light strip, that is, there will be a difference in image color. The essential difference is that there is a difference in pixel value, which can also be understood as a difference in brightness between the defect and the surface of the light strip.

[0021] Step S200: Based on the Fast Fourier Transform algorithm, frequency domain information analysis is performed on the set of light strip images containing different defect features to determine the spectral information of different defect features. This specifically includes the following steps: S201. Perform analog-to-digital conversion on the set of LED strip images containing different defect features to obtain the digital signal corresponding to each LED strip defect image; S202. Perform signal preprocessing on the digital signal corresponding to each LED strip defect image to obtain the preprocessed digital signal of each LED strip defect image. S203. Based on the Fast Fourier Transform algorithm, the preprocessed digital signal of each LED strip defect image is subjected to frequency domain transformation to obtain the frequency domain signal of each LED strip defect image, i.e.: In the formula, The frequency domain signal for each LED strip defect image; The digital signal after preprocessing for each LED strip defect image; S204. Based on the bandpass filter, feature extraction processing is performed on the frequency domain signal of each LED strip defect image to obtain the spectral information of different defect features.

[0022] Specifically, step S202, which involves preprocessing the digital signal corresponding to each LED strip defect image to obtain the preprocessed digital signal for each LED strip defect image, includes the following steps: S2021. Based on a high-pass filter, noise removal processing is performed on the digital signal corresponding to each LED strip defect image to obtain a noise-free digital signal for each LED strip defect image. S2022. Sample the noise-free digital signal of each LED strip defect image to obtain the sampled digital signal of each noise-free image. S2023. The digital signal after sampling each noise-free image is smoothed by using the moving average method to obtain the preprocessed digital signal of each LED strip defect image. In this embodiment, different types of LED strip defects manifest in different ways. For example, there may be pits on the surface of the LED strip and foreign objects on the surface of the LED strip. Therefore, the spectrum corresponding to the defect features in the LED strip defect image will also be different. So, by collecting the defect spectrum of images in the set of LED strip images containing different defect features and introducing the defect spectrum into the subsequent training process of the LED strip defect recognition model, the training process of the LED strip defect recognition model is not monotonous, thus improving the recognition accuracy of the LED strip defect recognition model.

[0023] Step S400: Based on the LED strip defect training set, train the convolutional neural network model to obtain the LED strip defect recognition model. This specifically includes the following steps: S401. Set the set of LED strip images containing different defect features as the validation set of the LED strip defect recognition model; S402. Based on the LED strip defect training set, perform model training on the convolutional neural network model and obtain the model output results to be verified. S403. Based on the validation set of the LED strip defect identification model, perform structural verification processing on the verification results output by the model to obtain the LED strip defect verification results. S404. Compare and analyze the verification results of LED strip defects to determine the LED strip defect identification model. Step S404, which involves comparing and analyzing the results of the LED strip defect verification to determine the LED strip defect identification model, specifically includes the following steps: S4041. Perform data matching processing on the LED strip defect verification results and the set of LED strip images containing different defect features; S4042. If the verification result of the LED strip defect is the same as the image defect features in the LED strip image set containing different defect features, determine the LED strip defect recognition model. S4043. If the defect verification result of the light strip is different from the image defect features in the set of light strip images containing different defect features, reselect the set of light strip images containing different defect features to train the convolutional neural network model and determine the light strip defect recognition model. In this embodiment, if the training of the LED strip defect recognition model is error-free and the training range is appropriate, the accuracy of the LED strip defect recognition model will be high, and the defect type identified by the LED strip defect recognition model will be correct. Therefore, by verifying the LED strip defect verification results, the accuracy of the LED strip defect recognition model is determined, and it is determined whether the LED strip defect recognition model needs to be retrained.

[0024] Step S500: Obtain the image of the LED strip defect to be verified. Based on the LED strip defect recognition model, perform recognition and verification processing on the image of the LED strip defect to be verified to determine the type of LED strip defect. This specifically includes the following steps: S501. Based on the shooting equipment, perform image processing on the defective light strip to obtain the defective image of the light strip to be verified. S502. Based on the LED strip defect recognition model, feature extraction and signal analysis are performed on the LED strip defect image to be verified to obtain the LED strip defect features and the spectrum of the LED strip defect image to be verified. S503. Based on the different defect features of the light strip and the spectral information of the different defect features, perform data matching processing on the defect features of the light strip to be verified and the spectrum of the defect image of the light strip to be verified to obtain the defect feature recognition result and the spectrum recognition result of the light strip defect feature. S504. Compare and analyze the results of the LED strip defect feature identification and the LED strip defect feature spectrum identification to determine the LED strip defect type.

[0025] Specifically, step S504, comparing and analyzing the results of the LED strip defect feature identification and the results of the LED strip defect feature spectrum identification to determine the LED strip defect type, includes the following steps: S5041. If the strip defect feature identification result is that the strip defect feature to be verified is the same as one of the different defect features of the strip, and the strip defect feature spectrum identification result is that the strip defect image spectrum to be verified is the same as one of the spectrum information of different defect features, then determine the strip defect type. S5042. If the defect feature identification result of the light strip is that the defect feature of the light strip to be verified is different from the different defect features of the light strip, or the spectrum identification result of the light strip defect feature is that the spectrum of the defect image of the light strip to be verified is different from the spectrum information of different defect features, the defect identification accuracy of the light strip defect identification model is low, and the light strip defect identification model is retrained. In this embodiment, after the LED strip defect recognition model is trained, the defect type of the LED strip defect image to be verified can be determined by the LED strip defect recognition model, and it can be determined whether the defective LED strip corresponding to the LED strip defect image to be verified can be put into normal use. This can not only reduce the workload of the staff, but also reduce the production cost of the LED strip.

[0026] Reference Figure 2 As shown, a defect detection system based on an internal light strip in a car sun visor is used to implement the defect detection method for an internal light strip in a car sun visor as described above, including: A defect detection terminal is used to control the data transmission and information interaction between various modules. The defect detection terminal is also used to control the various modules to perform model training processing on the LED strip defect identification model to determine the LED strip defect type. A database system for storing a set of light strip images containing different defect features; A noise removal module is used to perform noise removal processing on images in a set of light strip images containing different defect features to obtain a set of denoised light strip images. The feature extraction module is used to perform feature extraction processing on the set of denoised light strip images to obtain different defect features of the light strip; A defect spectrum determination module is used to perform frequency domain information analysis and processing on a set of light strip images containing different defect features to determine the spectrum information of different defect features. The model training module trains the convolutional neural network model based on the LED strip defect training set to obtain an LED strip defect recognition model. The defect type determination module performs identification and verification processing on the LED strip defect image to be verified based on the LED strip defect identification model to determine the LED strip defect type.

[0027] The foregoing has shown and described the basic principles, main features, and advantages of the present invention. Those skilled in the art should understand that the present invention is not limited to the above embodiments. The embodiments and descriptions in the specification are merely principles of the invention. Various changes and modifications can be made to the invention without departing from its spirit and scope, and all such changes and modifications fall within the scope of the claimed invention. The scope of protection claimed by the appended claims and their equivalents is defined.

Claims

1. A defect detection method based on the internal light strip of an automotive sun visor, characterized in that, include: A set of LED strip images containing different defect features is obtained. Based on image processing algorithms, feature extraction processing is performed on the set of LED strip images containing different defect features to obtain different defect features of the LED strips. Based on the Fast Fourier Transform algorithm, frequency domain information analysis is performed on a set of light strip images containing different defect features to determine the spectral information of different defect features. The different defect characteristics and spectral information of the LED strip are integrated and processed to obtain a training set of LED strip defects; Based on the LED strip defect training set, a convolutional neural network model is trained to obtain an LED strip defect recognition model. The defect image of the LED strip to be verified is obtained. Based on the LED strip defect recognition model, the defect image of the LED strip to be verified is recognized and verified to determine the defect type of the LED strip.

2. The defect detection method based on the internal light strip of an automotive sun visor according to claim 1, characterized in that, The process of obtaining a set of LED strip images containing different defect features involves, based on image processing algorithms, performing feature extraction on the set of LED strip images containing different defect features to obtain the different defect features of the LED strips. Specifically, this includes the following steps: Data retrieval and processing are performed on the database system to obtain a set of light strip images containing different defect characteristics; Data reading and processing are performed on images from a set of light strip images containing different defect characteristics to obtain image capture parameters; Data matching processing is performed on the image capture parameters to determine the type of noise in the light strip image; Select a Gaussian filter based on the type of noise in the LED strip image; Based on a Gaussian filter, images in a set of light strip images containing different defect features are denoised to obtain a denoised light strip image set. Image analysis and processing are performed on the data in the denoised LED strip image set to determine the different defect characteristics of the LED strip.

3. The defect detection method based on the internal light strip of an automotive sun visor according to claim 2, characterized in that, The step of performing image analysis processing on the data in the denoised LED strip image set to determine the different defect features of the LED strip specifically includes the following steps: Based on the Laplacian operator, the images in the denoised light strip image set are processed to determine the edge positions of different defect features of the light strip; Based on the Canny algorithm, edge differentiation processing is performed on the edge positions of different defect features of the light strip to determine the true edge positions of different defect features of the light strip. Based on the true edge location of different defect features of the light strip, feature extraction is performed on the data in the set of denoised light strip images to obtain different defect features of the light strip; The specific calculation formula for determining the edge position of different defect features of the light strip is as follows: In the formula, The edge locations of different defect characteristics of the light strip; The grayscale values ​​of the images in the set of denoised light strip images; These are the pixel coordinates of the images in the set of denoised LED strip images.

4. The defect detection method based on the internal light strip of an automotive sun visor according to claim 3, characterized in that, The method of analyzing and processing the frequency domain information of a set of light strip images containing different defect features based on the Fast Fourier Transform algorithm to determine the spectral information of different defect features specifically includes the following steps: Analog-to-digital conversion is performed on a set of LED strip images containing different defect features to obtain the digital signal corresponding to each LED strip defect image; The digital signal corresponding to each LED strip defect image is preprocessed to obtain the preprocessed digital signal of each LED strip defect image. Based on the Fast Fourier Transform algorithm, the preprocessed digital signal of each LED strip defect image is subjected to frequency domain transformation to obtain the frequency domain signal of each LED strip defect image, i.e.: In the formula, The frequency domain signal for each LED strip defect image; The digital signal after preprocessing for each LED strip defect image; Based on a bandpass filter, feature extraction processing is performed on the frequency domain signal of each LED strip defect image to obtain spectral information of different defect features.

5. The defect detection method based on the internal light strip of an automotive sun visor according to claim 4, characterized in that, The step of preprocessing the digital signal corresponding to each LED strip defect image to obtain the preprocessed digital signal of each LED strip defect image specifically includes the following steps: Based on a high-pass filter, noise removal processing is performed on the digital signal corresponding to each LED strip defect image to obtain a noise-free digital signal for each LED strip defect image. The noise-free digital signal of each LED strip defect image is sampled and processed to obtain the sampled digital signal of each noise-free image; The digital signal of each noise-free image is smoothed by using a moving average method to obtain the preprocessed digital signal of each LED strip defect image.

6. The defect detection method based on the internal light strip of an automotive sun visor according to claim 5, characterized in that, The process of training a convolutional neural network model based on a training set of LED strip defects to obtain an LED strip defect recognition model specifically includes the following steps: The set of LED strip images containing different defect features is set as the validation set for the LED strip defect recognition model; Based on the training set of LED strip defects, the convolutional neural network model is trained to obtain the model output results to be verified. Based on the validation set of the LED strip defect identification model, structural verification processing is performed on the output results of the model to obtain the LED strip defect verification results. The results of LED strip defect verification were compared and analyzed to determine the LED strip defect identification model.

7. The defect detection method based on the internal light strip of an automotive sun visor according to claim 6, characterized in that, The process of comparing and analyzing the verification results of LED strip defects to determine the LED strip defect identification model includes the following steps: Data matching processing is performed on the verification results of LED strip defects and the set of LED strip images containing different defect features; If the defect verification result of the LED strip is the same as the image defect features in the LED strip image set containing different defect features, the LED strip defect recognition model is determined. If the defect verification result of the LED strip is different from the image defect features in the LED strip image set containing different defect features, a new set of LED strip images containing different defect features is selected to train the convolutional neural network model and determine the LED strip defect recognition model.

8. The defect detection method based on the internal light strip of an automotive sun visor according to claim 1, characterized in that, The process of acquiring the LED strip defect image to be verified, and performing identification and verification processing on the LED strip defect image based on the LED strip defect recognition model to determine the LED strip defect type specifically includes the following steps: Based on the shooting equipment, image processing is performed on the defective light strip to obtain the defective image of the light strip to be verified; Based on the LED strip defect recognition model, feature extraction and signal analysis are performed on the LED strip defect image to be verified to obtain the defect features and the spectrum of the LED strip defect image to be verified. Based on the different defect features and spectral information of the LED strip, data matching processing is performed on the defect features of the LED strip to be verified and the spectrum of the defect image of the LED strip to be verified to obtain the LED strip defect feature recognition result and the LED strip defect feature spectrum recognition result. By comparing and analyzing the results of LED strip defect feature identification and LED strip defect feature spectrum identification, the type of LED strip defect is determined.

9. A defect detection method based on an internal light strip in an automotive sun visor according to claim 8, characterized in that, The process of comparing and analyzing the results of LED strip defect feature identification and LED strip defect feature spectrum identification to determine the type of LED strip defect includes the following steps: If the LED strip defect feature identification result is that the defect feature to be verified is the same as one of the different defect features of the LED strip, and the LED strip defect feature spectrum identification result is that the spectrum of the LED strip defect image to be verified is the same as one of the spectrum information of the different defect features, then the LED strip defect type is determined. If the defect feature recognition result of the light strip is that the defect feature to be verified is different from the different defect features of the light strip, or if the spectrum recognition result of the defect feature is that the spectrum of the defect image of the light strip to be verified is different from the spectrum information of different defect features, the defect recognition accuracy of the light strip defect recognition model is low, and the light strip defect recognition model should be retrained.

10. A defect detection system based on an internal LED strip in an automotive sun visor, used to implement the defect detection method based on an internal LED strip in an automotive sun visor as described in any one of claims 1-9, characterized in that, include: A defect detection terminal is used to control the data transmission and information interaction between various modules. The defect detection terminal is also used to control the various modules to perform model training processing on the LED strip defect identification model to determine the LED strip defect type. A database system for storing a set of light strip images containing different defect features; A noise removal module is used to perform noise removal processing on images in a set of light strip images containing different defect features to obtain a set of denoised light strip images. The feature extraction module is used to perform feature extraction processing on the set of denoised light strip images to obtain different defect features of the light strip; A defect spectrum determination module is used to perform frequency domain information analysis and processing on a set of light strip images containing different defect features to determine the spectrum information of different defect features. The model training module trains the convolutional neural network model based on the LED strip defect training set to obtain an LED strip defect recognition model. The defect type determination module performs identification and verification processing on the LED strip defect image to be verified based on the LED strip defect identification model to determine the LED strip defect type.

Images