Down feather identification method and system based on machine vision

Abstract

The invention discloses a down feather identification method and system based on machine vision, and the method comprises the steps: S1, obtaining an original image, and extracting H, S and V channel component values; s2, calculating a V channel distribution histogram and H, S and V channel cumulative distribution function diagrams; s3, extracting a pixel value, and accumulating the probability as a significant feature of the image; s4, screening the processed image features by adopting a random forest algorithm; s5, constructing a multi-feature weighted aggregation model, and performing weighted aggregation on the significance ranking of each feature to obtain the final feature of the image; s6, inputting the final features of the image into the decision tree model for identification, and outputting a result of white velvet or miscellaneous velvet; s7, inputting the identified image features corresponding to the white velvet or the miscellaneous velvet into a support vector machine model for identification, wherein the support vector machine model outputs whether the white velvet or the miscellaneous velvet is fresh velvet or recycled velvet; and S8, carrying out statistical superposition on the identification results to obtain a final identification result.

Description

Technical Field

[0001] The present invention relates to the technical field of machine vision, and in particular to a down identification method and system based on machine vision. Background Art

[0002] Traditional down identification and classification mainly relies on manual visual inspection, which involves packaging the down in a transparent sealed bag and having inspectors visually inspect the appearance, shape, impurity distribution and other characteristics of the down to determine whether it is fresh or recycled. However, this manual visual inspection method has many limitations. Due to the influence of the inspector's experience and subjective factors, its accuracy is difficult to be stably guaranteed during long-term inspections.

[0003] In the prior art, fully automatic down recognition methods based on machine vision have been developed, but currently these methods mainly focus on the classification of down types. For example, the patent with publication number CN111680582A discloses a down detection method and system based on a support vector machine, including: establishing a down database containing different down categories and corresponding down data; selecting a certain number of down categories and corresponding down data as samples according to the down database to form a training set; preprocessing the samples in the training set, using the preprocessed down data as input and the down category as output, training the support vector machine model to obtain a down vector machine model; obtaining the down data of the down to be detected and performing the same preprocessing, inputting the preprocessed down data into the down vector machine model, and obtaining the down category of the down to be detected. Although the above patent can detect down raw materials in the production and processing of down to achieve differentiation, such as distinguishing duck down from goose down. However, for the more critical problem of distinguishing fresh down from recycled down, the accuracy of the prior art is generally low, and there is a lack of an effective solution.

[0004] In view of the above technical problems, the present invention proposes a down recognition method and system based on machine vision. Summary of the invention

[0005] The purpose of the present invention is to provide a down identification method and system based on machine vision in view of the defects of the prior art.

[0006] In order to achieve the above purpose, the present invention adopts the following technical solutions:

[0007] A down recognition method based on machine vision, comprising: S1. Obtaining an original image corresponding to the down, and extracting the H, S, and V channel component values ​​of the original image; S2. Calculate the V channel distribution histogram and the H, S, V channel cumulative distribution function graph; S3. extracting the pixel value corresponding to the peak value in the V channel distribution histogram and the cumulative probability corresponding to the pixel value in the H, S, V channel cumulative distribution function map as the significant features of the image, and normalizing the significant features of the original image to obtain the significant features of the processed image; S4. Using the random forest algorithm to screen the processed image features, obtain the significance ranking of each feature in the image's significant features; S5. construct a multi-feature weighted aggregation model, and perform weighted aggregation on the significance ranking of each feature based on the multi-feature weighted aggregation model to obtain the final features of the image; S6. The final features of the image are input into the decision tree model for identification, and the decision tree model outputs the result of white velvet or mixed velvet; S7. Inputting the image features corresponding to the identified white down or mixed down into the support vector machine model for identification, the support vector machine model outputs whether the white down or mixed down is fresh down or recycled down; S8. Perform statistical superposition on the recognition results outputted from steps S6 and S7 to obtain a final recognition result.

[0008] Furthermore, before extracting the H, S, and V channel component values ​​of the original image in step S1, the step also includes: converting the RGB channel of the original image into an HSV channel.

[0009] Furthermore, in step S3, the pixel value corresponding to the peak value in the V channel distribution histogram and the cumulative probability corresponding to the pixel value in the H, S, V channel cumulative distribution function graph are extracted as the significant features of the image through feature engineering.

[0010] Furthermore, in step S3, the normalization process of the significant features of the image is performed based on a normalization function, wherein the normalization function is expressed as: ; ; ; ; Among them, x1, x2, x3, and x4 represent the pixel value corresponding to the peak value in the V channel distribution histogram of the original image, and the cumulative probability corresponding to the pixel value in the H, S, and V channel cumulative distribution function diagrams, respectively; μ represents the pixel value corresponding to the peak value in the V channel distribution histogram, and the cumulative probability corresponding to the pixel value in the H, S, and V channel cumulative distribution function diagrams, and the corresponding characteristic mean; σ represents the pixel value corresponding to the peak value in the V channel distribution histogram, and the cumulative probability corresponding to the pixel value in the H, S, and V channel cumulative distribution function diagrams, and the corresponding characteristic standard deviation; v hist 、h cdf 、scdf 、v cdf They respectively represent the normalized V channel distribution histogram features, H channel cumulative probability distribution features, S channel cumulative probability distribution features, and V channel cumulative probability distribution features.

[0011] Furthermore, the final features of the image obtained in step S5 are expressed as: ; Among them, ω1, ω2, ω3, and ω4 represent the feature weight values ​​corresponding to the normalized V channel distribution histogram features, H channel cumulative probability distribution features, S channel cumulative probability distribution features, and V channel cumulative probability distribution features, respectively; f represents the final features of the image.

[0012] Furthermore, the final recognition result in step S8 includes fresh white fluff, recycled white fluff, fresh mixed fluff, and recycled mixed fluff.

[0013] Correspondingly, a down identification system based on machine vision is also provided, comprising an inspection box, an inspection table, an imaging device equipped with a down identification method based on machine vision, and a display screen; the inspection box is a square structure with a hollow interior, the inspection table is arranged on the bottom of the inspection box, and a feed port is opened on one side of the inspection box, the imaging device is arranged on the top of the inspection box, and the display screen is arranged on the side outside the inspection box, so that the down can be put into the inspection box through the feed port and placed on the inspection table, and the original image of the down is collected by the imaging device, and the identification result is displayed on the display screen.

[0014] Furthermore, the imaging device includes a bar light source, a bracket, a polarizing plate, a camera base, and an industrial camera; The bracket is arranged in the inspection box, the strip light source and the camera base are fixed on the bracket, the industrial camera is fixed on the camera base, and the polarizing plate is installed on the industrial camera.

[0015] Furthermore, the industrial camera is also connected to a Raspberry Pi via a data cable.

[0016] Furthermore, the Raspberry Pi is used to implement a down recognition method based on machine vision.

[0017] Compared with the prior art, the present invention has the following beneficial effects: 1. Put the down sample in from the feed port and place it on the inspection table so that the subsequent imaging device can collect the down image; 2. The industrial camera automatically acquires images and converts the images from RGB channels to HSV channels, while extracting the HSV channel component values. The polarizer installed on the industrial camera can solve the glare problem on the outer surface of the transparent bag of the down sample, which is conducive to the extraction of color features of the down image data; 3. Calculate the V channel distribution histogram and the H, S, and V channel cumulative distribution function diagrams to improve the algorithm's ability to analyze color information; 4. The significant features of the H, S, and V channels of the above distribution map are extracted through feature engineering, specifically the pixel value corresponding to the peak in the V channel distribution histogram and the cumulative probability corresponding to the 0-255 pixel value in the H, S, and V channel cumulative distribution function map. As the significant features of the image, the algorithm is more sensitive to lighting, color, and details, and the robustness and efficiency of image analysis are enhanced; 5. Design a down feature normalization function to normalize the significant features of the image, which can eliminate the scale differences between features, help improve the expression of features in the image, and enable subsequent machine learning models to be trained and inferred more efficiently.

[0018] 6. Use the random forest algorithm to screen and obtain the significance ranking of each feature, retain the most valuable features for prediction, and eliminate redundant features to facilitate the weighted aggregation of subsequent features.

[0019] 7. Design a multi-feature weighted aggregation model, perform weighted aggregation based on the significance of the features, and obtain the final features. This can highlight the contribution of useful features, reduce noise interference, and obtain more representative comprehensive features.

[0020] 8. The decision tree model is trained based on the above image features to preliminarily identify white and mixed wool. It can effectively use the extracted color and brightness features for classification, and has the advantages of strong interpretability, efficient training and reasoning, adaptability to nonlinear relationships, and avoidance of overfitting. 9. If it is identified as white (mixed) down, the image features of the white (mixed) down are used to train the support vector machine model to identify fresh down and recycled down in the white (mixed) down again. The support vector machine can effectively handle the subtle differences between fresh down and recycled down by maximizing the boundary and nonlinear mapping, ensuring high classification accuracy. 10. The results of the two identifications are statistically superimposed to obtain the final result (fresh white fluff, recycled white fluff, fresh mixed fluff, recycled mixed fluff). BRIEF DESCRIPTION OF THE DRAWINGS

[0021] Figure 1 is a structural diagram of a down identification system based on machine vision provided in Example 1; Figure 2 This is an interactive diagram of a down identification system based on machine vision provided in Example 1; Figure 3 This is a flow chart of a down identification method based on machine vision provided in Example 2; Figure 4 This is the HSV feature distribution diagram of the down image provided in Example 2; Figure 5This is the recognition result diagram provided in Example 2. DETAILED DESCRIPTION

[0022] The following describes the embodiments of the present invention by specific examples, and those skilled in the art can easily understand other advantages and effects of the present invention from the contents disclosed in this specification. The present invention can also be implemented or applied through other different specific embodiments, and the details in this specification can also be modified or changed in various ways based on different viewpoints and applications without departing from the spirit of the present invention. It should be noted that the following embodiments and features in the embodiments can be combined with each other without conflict.

[0023] The purpose of the present invention is to provide a down identification method and system based on machine vision in view of the defects of the prior art.

[0024] Embodiment 1

[0025] This embodiment provides a down identification system based on machine vision, such as Figure 1 , Figure 2 As shown, it includes an inspection table 1, a feed port 2, an inspection box 3, an imaging device 4, a Raspberry Pi 5, and a display screen 6.

[0026] The inspection box 3 is a closed box with a hollow interior, which is used to accommodate components such as the imaging device 4, the inspection table 1, the feed port 2, etc., and provides a relatively stable environment for imaging the down samples.

[0027] The inspection table 1 is arranged at the bottom of the inspection box 3, and the inspection table 1 is used to place the down sample; the feed port 2 is arranged on the front of the inspection box 3, and the feed port 2 is an opening, that is, the feed port 2 is opened on the front of the inspection box 3, and the down sample enters the inspection box 3 through the feed port 2, so that the down sample is placed on the inspection table 1.

[0028] The imaging device 4 is fixed inside the inspection box 3 and is used to collect image information of the down sample.

[0029] Raspberry Pi 5, as the core component of image processing and model operation, is placed in a suitable position inside or outside the inspection box 3; and Raspberry Pi 5 has multiple interfaces, such as HDMI, USB, Ethernet, etc., which are convenient for connecting with external devices such as imaging device 4 and display screen 6 to realize functions such as image data collection and result display, providing convenience for building a down recognition device.

[0030] The display screen 6 is used to display the recognition result and is placed outside the device for easy viewing by operators. It is generally located on the side or above the inspection box 3 .

[0031] In this embodiment, the imaging device 4 includes a bar light source 7 , a bracket 8 , a polarizing plate 9 , a camera base 10 , and an industrial camera 11 .

[0032] The bracket 8 is used to support the strip light source 7 and the industrial camera 11 , and is placed inside the inspection box 3 . Its position and structure need to be designed according to the overall layout of the imaging device 4 .

[0033] The strip light source 7 is fixed on both sides of the bracket 8 and is used to illuminate the down sample to improve the imaging quality. Its position needs to be coordinated with the industrial camera 11 to ensure that the light can be evenly irradiated on the down sample.

[0034] The camera base 10 is used to fix the industrial camera 11 and is placed on the bracket 8 to ensure that the position of the industrial camera 11 is stable.

[0035] The industrial camera 11 is used to collect images of the down samples and is fixed on the camera base 10 , placed on one side of the bracket 8 , opposite to the strip light source 7 , and aimed at the down samples on the inspection table 1 .

[0036] The polarizing plate 9 is installed on the industrial camera 11 to solve the glare problem on the outer surface of the transparent bag of the down sample, and is placed in front of the lens of the industrial camera 11.

[0037] In this embodiment, the down image data collected by the industrial camera 11 in the imaging device 4 is transmitted to the Raspberry Pi 5 via a data line, and the Raspberry Pi 5 performs operations such as preprocessing, feature extraction, model training and recognition on the image.

[0038] The strip light source 7 provides lighting for the imaging device 4. Through reasonable layout and light adjustment, the imaging device 4 can clearly capture the image of the down sample, improve the image quality, and thus provide a better basis for subsequent image processing and recognition.

[0039] The polarizing plate 9 is installed on the industrial camera 11. When the industrial camera 11 shoots the down sample packaged in the transparent sealed bag, the polarizing plate 9 can effectively solve the glare problem generated by the outer surface of the transparent bag, making the captured image clearer, which is conducive to the extraction of color features of the down image data and improving the accuracy of recognition.

[0040] The Raspberry Pi 5 transmits the identified down results (such as fresh white down, recycled white down, fresh mixed down, recycled mixed down) to the display screen 6 through the data cable, and the display screen 6 displays the results, so that the operator can intuitively understand the identification status of the down sample and realize human-computer interaction.

[0041] It should be noted that the connection between each component in this embodiment can be achieved by threading, welding, gluing, etc., and the installation position of each component can be set according to actual conditions, which will not be described in detail in this embodiment.

[0042] Embodiment 2

[0043] This embodiment provides a down recognition method based on machine vision, and the recognition method is implemented based on the recognition system of the first embodiment.

[0044] like Figure 3 As shown, the identification method includes: S1. Obtaining an original image corresponding to the down, and extracting the H, S, and V channel component values ​​of the original image; S2. Calculate the V channel distribution histogram and the H, S, V channel cumulative distribution function graph; S3. extracting the pixel value corresponding to the peak value in the V channel distribution histogram and the cumulative probability corresponding to the pixel value in the H, S, V channel cumulative distribution function map as the significant features of the image, and normalizing the significant features of the original image to obtain the significant features of the processed image; S4. Using the random forest algorithm to screen the processed image features, obtain the significance ranking of each feature in the image's significant features; S5. construct a multi-feature weighted aggregation model, and perform weighted aggregation on the significance ranking of each feature based on the multi-feature weighted aggregation model to obtain the final features of the image; S6. The final features of the image are input into the decision tree model for identification, and the decision tree model outputs the result of white velvet or mixed velvet; S7. Inputting the image features corresponding to the identified white down or mixed down into the support vector machine model for identification, the support vector machine model outputs whether the white down or mixed down is fresh down or recycled down; S8. Perform statistical superposition on the recognition results outputted from steps S6 and S7 to obtain a final recognition result.

[0045] In step S1, an original image corresponding to down is acquired, and the H, S, and V channel component values ​​of the original image are extracted.

[0046] First, a white down or mixed down sample is put into the inspection box 3 from the feed port 2 and placed on the inspection table 1. Then, the imaging device 4 collects images. Specifically, the sample is illuminated by the strip light source 7 fixed on both sides of the bracket 8. An industrial camera 11 equipped with a polarizing plate 9 is used to collect the original image of the white down or mixed down sample, and the original image is transmitted to the image preprocessing algorithm in the Raspberry Pi 5 for processing, specifically: first, the RGB image is read, and the RGB image is converted into an HSV image using the existing method, and the component values ​​of the H, S, and V channels of the converted HSV image are extracted.

[0047] In this embodiment, adding a polarizing plate to the industrial camera can solve the glare problem on the outer surface of the transparent bag of the down sample, which is beneficial to the extraction of color features of the down image data.

[0048] In step S2, the V channel distribution histogram and the H, S, and V channel cumulative distribution function graphs are calculated.

[0049] The V channel distribution histogram describes the distribution of the V channel (brightness) values ​​in the image. The specific steps are as follows: extract the component values ​​of the V channel from the HSV image, count the frequency of each possible value (0 to 255) in the V channel, and obtain a histogram, where the parameters of the histogram represent the image, channel index, mask, histogram size and range respectively; the histogram can intuitively display the distribution.

[0050] The cumulative distribution function (CDF) graph describes the cumulative probability that the H, S, and V channel values ​​in the image are less than or equal to a certain value. The specific steps are as follows: First, calculate the distribution histogram of the H, S, and V channels, cumulatively sum the histograms of each channel to obtain the CDF, and normalize the CDF to the range of [0, 1] to represent the cumulative probability; The CDF graph can intuitively display the distribution.

[0051] In step S3, the pixel value corresponding to the peak value in the V channel distribution histogram and the cumulative probability corresponding to the pixel value in the H, S, and V channel cumulative distribution function diagrams are extracted as the salient features of the image, and the salient features of the original image are normalized to obtain the processed image salient features.

[0052] like Figure 4 The figure shows the HSV feature distribution of white down (WHITE) or other down (OTHER) down images. The significant features of the H, S, and V channels of the above distribution map are extracted through feature engineering, specifically the pixel value corresponding to the peak value in the V channel distribution histogram and the cumulative probability corresponding to the 0-255 pixel value in the H, S, and V channel cumulative distribution function diagram as the significant features of the image.

[0053] This embodiment directly obtains the value of each pixel from the H, S, and V channels, and these values ​​can be used as features for model training. The cumulative probability of each pixel value is also obtained by calculating the cumulative sum of the histogram and normalizing it, and these cumulative probabilities can be used as features for model training. This embodiment combines the peak value of the V channel distribution histogram and the cumulative probability of the H, S, and V channels to extract significant features, which can better capture the key information in the image data and improve the performance of the model.

[0054] This embodiment also designs a feather feature normalization function to normalize the significant features of the image, which is expressed as: ; ; ; ; Among them, x1, x2, x3, and x4 represent the pixel value corresponding to the peak value in the V channel distribution histogram of the original image, and the cumulative probability corresponding to the pixel value in the H, S, and V channel cumulative distribution function diagrams, respectively; μ represents the pixel value corresponding to the peak value in the V channel distribution histogram, and the cumulative probability corresponding to the pixel value in the H, S, and V channel cumulative distribution function diagrams, and the corresponding characteristic mean; σ represents the pixel value corresponding to the peak value in the V channel distribution histogram, and the cumulative probability corresponding to the pixel value in the H, S, and V channel cumulative distribution function diagrams, and the corresponding characteristic standard deviation; v hist 、h cdf 、s cdf 、v cdf They respectively represent the normalized V channel distribution histogram features, H channel cumulative probability distribution features, S channel cumulative probability distribution features, and V channel cumulative probability distribution features. The above formula can be used to normalize the features and eliminate the scale differences between the features.

[0055] In step S4, the processed image features are screened using a random forest algorithm to obtain the significance ranking of each feature in the image's significant features. Specifically: S41. Prepare data; First, it is necessary to prepare the feature data of the down sample images, including the pixel value, histogram, cumulative probability, etc. of the image, and divide the feature data into a training set and a test set.

[0056] S42. Train the random forest model; The training set is input into the training random forest algorithm, the model is trained using the random forest algorithm, and the importance score of each feature is obtained.

[0057] S43. Obtain feature importance; After training, the random forest model will provide a feature importance score, indicating the contribution of each feature to the model prediction. These scores can be used for feature selection, retaining important features, removing unimportant features, and facilitating sorting and visualization.

[0058] S44. Screening of image features: Sort in descending order of importance scores to get the significance ranking of features, select the top N important features as needed, and reconstruct the feature matrix.

[0059] Through the above steps, the random forest algorithm can be used to filter the significance ranking of features and retain the most valuable features for prediction, thereby improving the performance and interpretability of the model.

[0060] In step S5, a multi-feature weighted aggregation model is constructed, and the significance ranking of each feature is weighted and aggregated based on the multi-feature weighted aggregation model to obtain the final features of the image.

[0061] According to the importance score of the feature, a weight ω is assigned to each feature. The weight ω can be the direct value of the feature importance score or the value after normalization. Therefore, according to the significance ranking of each feature obtained in step S4, the weight values ​​ω are assigned from high to low, such as 1, 0.75, 0.5, and 0.25. After the features are aggregated, a more representative aggregate feature can be obtained. Multiply each feature by its corresponding weight, and then add all the weighted feature values ​​to obtain the final aggregate feature, which is expressed as: ; Among them, ω1, ω2, ω3, and ω4 represent the feature weight values ​​corresponding to the normalized V channel distribution histogram features, H channel cumulative probability distribution features, S channel cumulative probability distribution features, and V channel cumulative probability distribution features, respectively; f represents the final features of the image.

[0062] In step S6, the final features of the image are input into the decision tree model for identification, and the decision tree model outputs the result of white velvet or mixed velvet. Specifically: S61.Data preparation: First, you need to prepare training data, including the aggregated feature matrix and the corresponding label data. The feature matrix is ​​obtained through weighted aggregation, and the label data indicates whether each sample is white or mixed.

[0063] S62. Divide the training set and test set: The data is divided into a training set and a test set so that the model can be trained on the training set and the performance of the model can be evaluated on the test set.

[0064] S63. Training decision tree model: Use the DecisionTreeClassifier class to train a decision tree model. The decision tree builds a tree-like model for classification tasks by learning the relationship between features and labels.

[0065] S64. Model evaluation: Use the accuracy_score and classification_report functions to evaluate the performance of the model. Evaluate the performance of the model on the test set. Common evaluation indicators include accuracy, precision, recall, and F1 score. Output accuracy and classification reports.

[0066] Through the above steps, the weighted aggregated features can be used to train a decision tree model to preliminarily identify whether the down sample is white down or mixed down, thereby improving the performance and interpretability of the model.

[0067] In step S7, the image features corresponding to the identified white down or mixed down are input into the support vector machine model for identification, and the support vector machine model outputs whether the white down or mixed down is fresh down or recycled down. S71.Data preparation: First, we need to prepare training data, including the image feature matrix of white or mixed wool and the corresponding label data. The feature matrix is ​​obtained through weighted aggregation, and the label data indicates whether each sample is fresh or recycled wool.

[0068] S72. Divide the training set and test set: The data is divided into a training set and a test set so that the model can be trained on the training set and the performance of the model can be evaluated on the test set.

[0069] S73. Training support vector machine model: The support vector machine model is trained using the SVC class, where the kernel parameter selects the linear kernel and the C parameter controls the regularization strength. SVM achieves classification tasks by finding the optimal hyperplane to maximize the interval between different classes.

[0070] S74. Model evaluation: Use the accuracy_score and classification_report functions to evaluate the performance of the model, output the accuracy and classification reports, and evaluate the performance of the model on the test set. Common evaluation indicators include accuracy, precision, recall, and F1 score.

[0071] Through the above steps, the weighted aggregated features can be used to train the support vector machine model to further identify fresh and recycled wool in white wool or mixed wool, thereby improving the performance and interpretability of the model.

[0072] In step S8, the recognition results outputted from steps S6 and S7 are statistically superimposed to obtain a final recognition result.

[0073] The prediction results of the decision tree model and the SVM model are statistically superimposed to determine the final classification result. The final classification result is displayed on the display screen 6. The final recognition results include fresh white wool, recycled white wool, fresh mixed wool, and recycled mixed wool. Figure 5 Shown is a schematic diagram of the recognition results.

[0074] Compared with the prior art, the present invention has the following beneficial effects: 1. Put the down sample in from the feed port and place it on the inspection table so that the subsequent imaging device can collect the down image; 2. The industrial camera automatically acquires images and converts the images from RGB channels to HSV channels, while extracting the HSV channel component values. The polarizer installed on the industrial camera can solve the glare problem on the outer surface of the transparent bag of the down sample, which is conducive to the extraction of color features of the down image data; 3. Calculate the V channel distribution histogram and the H, S, and V channel cumulative distribution function diagrams to improve the algorithm's ability to analyze color information; 4. The significant features of the H, S, and V channels of the above distribution map are extracted through feature engineering, specifically the pixel value corresponding to the peak in the V channel distribution histogram and the cumulative probability corresponding to the 0-255 pixel value in the H, S, and V channel cumulative distribution function map. As the significant features of the image, the algorithm is more sensitive to lighting, color, and details, and the robustness and efficiency of image analysis are enhanced; 5. Design a down feature normalization function to normalize the significant features of the image, which can eliminate the scale differences between features, help improve the expression of features in the image, and enable subsequent machine learning models to be trained and inferred more efficiently.

[0075] 6. Use the random forest algorithm to screen and obtain the significance ranking of each feature, retain the most valuable features for prediction, and eliminate redundant features to facilitate the weighted aggregation of subsequent features.

[0076] 7. Design a multi-feature weighted aggregation model, perform weighted aggregation based on the significance of the features, and obtain the final features. This can highlight the contribution of useful features, reduce noise interference, and obtain more representative comprehensive features.

[0077] 8. The decision tree model is trained based on the above image features to preliminarily identify white and mixed wool. It can effectively use the extracted color and brightness features for classification, and has the advantages of strong interpretability, efficient training and reasoning, adaptability to nonlinear relationships, and avoidance of overfitting. 9. If it is identified as white (mixed) down, the image features of the white (mixed) down are used to train the support vector machine model to identify fresh down and recycled down in the white (mixed) down again. The support vector machine can effectively handle the subtle differences between fresh down and recycled down by maximizing the boundary and nonlinear mapping, ensuring high classification accuracy. 10. The results of the two identifications are statistically superimposed to obtain the final result (fresh white fluff, recycled white fluff, fresh mixed fluff, recycled mixed fluff).

[0078] Note that the above are only preferred embodiments of the present invention and the technical principles used. Those skilled in the art will understand that the present invention is not limited to the specific embodiments described herein, and that various obvious changes, readjustments and substitutions can be made by those skilled in the art without departing from the scope of protection of the present invention. Therefore, although the present invention has been described in more detail through the above embodiments, the present invention is not limited to the above embodiments, and may include more other equivalent embodiments without departing from the concept of the present invention, and the scope of the present invention is determined by the scope of the appended claims.

Claims

1. A down identification method based on machine vision, characterized in that: include: S1. Obtaining an original image corresponding to the down, and extracting the H, S, and V channel component values ​​of the original image; S2. Calculate the V channel distribution histogram and the H, S, V channel cumulative distribution function graph; S3. extracting the pixel value corresponding to the peak value in the V channel distribution histogram and the cumulative probability corresponding to the pixel value in the H, S, V channel cumulative distribution function map as the significant features of the image, and normalizing the significant features of the original image to obtain the significant features of the processed image; S4. Using the random forest algorithm to screen the processed image features, obtain the significance ranking of each feature in the image's significant features; S5. construct a multi-feature weighted aggregation model, and perform weighted aggregation on the significance ranking of each feature based on the multi-feature weighted aggregation model to obtain the final features of the image; S6. The final features of the image are input into the decision tree model for identification, and the decision tree model outputs the result of white velvet or mixed velvet; S7. Inputting the image features corresponding to the identified white down or mixed down into the support vector machine model for identification, the support vector machine model outputs whether the white down or mixed down is fresh down or recycled down; S8. Perform statistical superposition on the recognition results outputted from steps S6 and S7 to obtain a final recognition result.

2. A down identification method based on machine vision according to claim 1, characterized in that: Before extracting the H, S, V channel component values ​​of the original image in step S1, the step also includes: converting the RGB channel of the original image into an HSV channel.

3. The down identification method based on machine vision according to claim 1, characterized in that: In step S3, the pixel value corresponding to the peak value in the V channel distribution histogram and the cumulative probability corresponding to the pixel value in the H, S, V channel cumulative distribution function diagram are extracted as the significant features of the image through feature engineering.

4. The down identification method based on machine vision according to claim 1, characterized in that: In step S3, the normalization process of the significant features of the image is performed based on a normalization function, wherein the normalization function is expressed as: ； ； ； ； Among them, x1, x2, x3, and x4 represent the pixel value corresponding to the peak value in the V channel distribution histogram of the original image, and the cumulative probability corresponding to the pixel value in the H, S, and V channel cumulative distribution function diagrams, respectively; μ represents the pixel value corresponding to the peak value in the V channel distribution histogram, and the cumulative probability corresponding to the pixel value in the H, S, and V channel cumulative distribution function diagrams, and the corresponding characteristic mean; σ represents the pixel value corresponding to the peak value in the V channel distribution histogram, and the cumulative probability corresponding to the pixel value in the H, S, and V channel cumulative distribution function diagrams, and the corresponding characteristic standard deviation; v hist 、h cdf 、s cdf 、v cdf They respectively represent the normalized V channel distribution histogram features, H channel cumulative probability distribution features, S channel cumulative probability distribution features, and V channel cumulative probability distribution features.

5. The down identification method based on machine vision according to claim 4, characterized in that: The final features of the image obtained in step S5 are expressed as: ； Among them, ω1, ω2, ω3, and ω4 represent the feature weight values ​​corresponding to the normalized V channel distribution histogram features, H channel cumulative probability distribution features, S channel cumulative probability distribution features, and V channel cumulative probability distribution features, respectively; f represents the final features of the image.

6. The down identification method based on machine vision according to claim 1, characterized in that: The final recognition result in step S8 includes fresh white down, recycled white down, fresh mixed down, and recycled mixed down.

7. A down identification system based on machine vision, characterized in that: It comprises an inspection box, an inspection table, an imaging device equipped with a down identification method based on machine vision as described in any one of claims 1 to 6, and a display screen; the inspection box is a square structure with a hollow interior, the inspection table is arranged on the bottom of the inspection box, and a feed port is opened on one side of the inspection box, the imaging device is arranged on the top of the inspection box, and the display screen is arranged on the side outside the inspection box, so that down can be put into the inspection box through the feed port and placed on the inspection table, and the original image of the down is collected by the imaging device, and the identification result is displayed on the display screen.

8. The down identification system based on machine vision according to claim 7, characterized in that: The imaging device includes a strip light source, a bracket, a polarizing plate, a camera base, and an industrial camera; The bracket is arranged in the inspection box, the strip light source and the camera base are fixed on the bracket, the industrial camera is fixed on the camera base, and the polarizing plate is installed on the industrial camera.

9. The down identification system based on machine vision according to claim 8, characterized in that: The industrial camera is also connected to a Raspberry Pi via a data cable.

10. The down identification system based on machine vision according to claim 9, characterized in that: The Raspberry Pi is used to implement a down identification method based on machine vision as described in any one of claims 1 to 6.

Images