Automobile coating query method and device, electronic equipment and storage medium

Abstract

The invention provides an automobile coating query method, which comprises the following steps: acquiring a full-surface image of a to-be-queried automobile, and determining the color complexity of the to-be-queried automobile based on the full-surface image; based on the color complexity of the to-be-queried automobile, matching a target feature extraction model corresponding to the color complexity in a model library; performing feature extraction on the full-surface image through a target feature extraction model to obtain appearance color features of the to-be-queried automobile; and querying paint data of the to-be-queried automobile in a paint database according to the appearance color characteristics. By introducing a color complexity evaluation mechanism, intelligent analysis of the full-surface image of the automobile is realized, and the most suitable feature extraction model is dynamically selected according to the analysis result, so that the color recognition accuracy is improved, the waste of computing resources is effectively reduced, the paint query time is greatly shortened, and the paint query efficiency is improved. And the overall automobile paint query efficiency is improved.

Description

Technical Field

[0001] The present invention relates to the technical field of video processing, and particularly to a method, device, electronic device and storage medium for querying automotive coatings. Background Art

[0002] In the automotive industry, the selection and application of coatings are crucial for the aesthetics, durability and market attractiveness of vehicles. The traditional method of querying automotive coatings mainly relies on manual identification of vehicle colors and then comparison with product catalogs provided by coating manufacturers. This process is not only time-consuming and laborious, but also easily affected by human factors, resulting in low recognition accuracy. With the development of digital technology, some automated recognition methods have gradually been applied to this field. However, most of the existing automated recognition systems are based on a single image processing algorithm and are difficult to adapt to the actual situation of complex and variable colors on the vehicle surface, and fail to perform differential processing according to the complexity of vehicle colors. Using a unified feature extraction model may not only lead to information redundancy, but also lose key color features due to model mismatch, thus affecting the final coating matching accuracy and resulting in inaccurate coating data being queried. Summary of the Invention

[0003] An embodiment of the present invention provides a method for querying automotive coatings, aiming to improve the efficiency and accuracy of automotive coating queries. By introducing a color complexity evaluation mechanism, intelligent analysis of the full-surface image of the vehicle is realized, and the most suitable feature extraction model is dynamically selected according to the analysis result, which not only improves the accuracy of color recognition, but also effectively reduces the waste of computing resources. At the same time, by constructing a refined coating database, efficient matching between the appearance color features and coating data is realized, greatly shortening the coating query time and improving the overall efficiency of automotive coating queries.

[0004] In a first aspect, an embodiment of the present invention provides a method for querying automotive coatings, the method comprising:

[0005] Obtain a full-surface image of the vehicle to be queried, and determine the color complexity of the vehicle to be queried based on the full-surface image, where the full-surface image is obtained by stitching surface images of the vehicle to be queried;

[0006] Based on the color complexity of the vehicle to be queried, match a target feature extraction model corresponding to the color complexity in a model library, where the model library includes multiple trained feature extraction models, and different trained feature extraction models correspond to different color complexities;

[0007] Extract features from the full-surface image through the target feature extraction model to obtain the appearance color features of the vehicle to be queried;

[0008] Query the paint data of the vehicle to be queried in the paint database according to the appearance color characteristics. In the paint database, different appearance color characteristics correspond to different paint data.

[0009] Optionally, the step of determining the color complexity of the vehicle to be queried based on the full-surface image includes:

[0010] Perform image segmentation on the full-surface image to obtain image blocks of different colors and the corresponding image block information for each image block. Each image block corresponds to one piece of image block information, and the image block information includes the image block color and the image block area.

[0011] Determine the color complexity of the vehicle to be queried based on the number of image blocks and the image block information corresponding to each image block.

[0012] Optionally, the step of determining the color complexity of the vehicle to be queried based on the number of image blocks and the image block information corresponding to each image block includes:

[0013] Determine the number of image blocks;

[0014] Calculate the color similarity between the image blocks based on the image block colors;

[0015] Determine the image blocks with color similarity greater than the similarity threshold as the same-color image blocks belonging to the same color category, and add up the image areas corresponding to the same-color image blocks to obtain the color areas corresponding to each category of colors.

[0016] Determine the color complexity of the vehicle to be queried based on the number of image blocks, the distance between the same-color image blocks, and the color areas corresponding to each category of colors.

[0017] Optionally, the step of determining the color complexity of the vehicle to be queried based on the number of image blocks, the distance between the same-color image blocks, and the color areas corresponding to each category of colors includes:

[0018] Calculate the central distance between the same-color image blocks;

[0019] Calculate the average color area corresponding to all categories of colors based on the color areas corresponding to each category of colors;

[0020] Calculate the color complexity of the vehicle to be queried through a preset formula. The preset formula is:

[0021]

[0022] where f is the color complexity, N is the number of color categories, and M nis the set of homogeneous color image patches in the nth color category, and num(M n ) is the number of homogeneous color image patches in the nth color category. S n is the area of the nth color category. is the average color area corresponding to all color categories. i is the ith homogeneous color image patch in the nth color category, j is the jth homogeneous color image patch in the nth color category, o is the image center of all homogeneous color image patches in the nth color category, D(i, j) is the central distance between the ith and jth homogeneous color image patches in the nth color category, and D(i, o) is the distance between the ith homogeneous color image patch in the nth color category and the image center of all homogeneous color image patches in the nth color category.

[0023] Optionally, before the step of matching the target feature extraction model corresponding to the color complexity in the model library based on the color complexity of the query vehicle, the method further includes:

[0024] Obtain a sample image set and a feature extraction model to be trained. The sample image set includes multiple sample images corresponding to different paint data and appearance color labels. Each sample image corresponds to one appearance color label. The sample image is the full surface image of the sample vehicle, and the feature extraction model to be trained is used to extract appearance color features.

[0025] Calculate the color complexity of each sample image. Each sample image corresponds to one color complexity.

[0026] Based on the color complexity, divide the multiple sample images to obtain multiple training sets. Each training set includes multiple sample images corresponding to the color complexity.

[0027] Based on the multiple training sets, train the feature extraction model to be trained respectively to obtain trained feature extraction models corresponding to different color complexities.

[0028] Associate the trained feature extraction models with the color complexity and store them in the model library.

[0029] Optionally, before the step of querying the paint data of the query vehicle in the paint database according to the appearance color features, the method further includes:

[0030] Obtain the paint data corresponding to the sample image.

[0031] After obtaining the trained feature extraction models corresponding to different color complexities, extract the appearance color features of the sample image through the trained feature extraction model to obtain the query features corresponding to the sample image.

[0032] Associate the query features with the paint data and store them in the paint database, where each query feature corresponds to one piece of paint data.

[0033] Optionally, after the step of querying the paint data of the to-be-query car in the paint database according to the appearance color features, the method further includes:

[0034] Determine the image positions corresponding to each appearance color feature in the full-surface image;

[0035] Through a guiding line, display the paint data corresponding to the appearance color feature at the corresponding image position.

[0036] In a second aspect, an embodiment of the present invention further provides an automobile paint query device, which includes:

[0037] A first acquisition module, configured to acquire a full-surface image of a to-be-query car and determine the color complexity of the to-be-query car based on the full-surface image, where the full-surface image is obtained by stitching surface images of the to-be-query car;

[0038] A matching module, configured to match a target feature extraction model corresponding to the color complexity in a model library based on the color complexity of the to-be-query car, where the model library includes multiple trained feature extraction models, and different trained feature extraction models correspond to different color complexities;

[0039] A processing module, configured to perform feature extraction on the full-surface image through the target feature extraction model to obtain the appearance color features of the to-be-query car;

[0040] A query module, configured to query the paint data of the to-be-query car in a paint database according to the appearance color features, where in the paint database, different appearance color features correspond to different paint data.

[0041] In a third aspect, an embodiment of the present invention provides an electronic device, including: a memory, a processor, and a computer program stored on the memory and executable on the processor, where when the processor executes the computer program, the steps in the automobile paint query method provided by the embodiment of the present invention are implemented.

[0042] In a fourth aspect, an embodiment of the present invention provides a computer-readable storage medium, on which a computer program is stored, and when the computer program is executed by a processor, the steps in the automobile paint query method provided by the embodiment of the present invention are implemented.

[0043] In the embodiments of the present invention, a full-surface image of a vehicle to be queried is obtained, and the color complexity of the vehicle to be queried is determined based on the full-surface image. The full-surface image is obtained by stitching surface images of the vehicle to be queried. Based on the color complexity of the vehicle to be queried, a target feature extraction model corresponding to the color complexity is matched in a model library. The model library includes multiple trained feature extraction models, and different trained feature extraction models correspond to different color complexities. The full-surface image is subjected to feature extraction through the target feature extraction model to obtain the external color feature of the vehicle to be queried. According to the external color feature, paint data of the vehicle to be queried is queried in a paint database. In the paint database, different external color features correspond to different paint data. By introducing a color complexity evaluation mechanism, the present invention realizes the intelligent analysis of the full-surface image of the vehicle, dynamically selects the most suitable feature extraction model according to the analysis result, not only improves the accuracy of color recognition, but also effectively reduces the waste of computing resources. At the same time, by constructing a refined paint database, the efficient matching between the external color feature and the paint data is realized, greatly shortening the paint query time and improving the overall vehicle paint query efficiency. BRIEF DESCRIPTION OF THE DRAWINGS

[0044] In order to more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the following will briefly introduce the drawings required for use in the description of the embodiments or the prior art. Obviously, the following drawings are only some embodiments of the present invention. For those of ordinary skill in the art, other drawings can be obtained based on these drawings without creative efforts.

[0045] Figure 1 is a flowchart of a method for querying vehicle paint provided by an embodiment of the present invention;

[0046] Figure 2 is a schematic structural diagram of a vehicle paint query device provided by an embodiment of the present invention;

[0047] Figure 3 is a schematic structural diagram of an electronic device provided by an embodiment of the present invention. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0048] The following will clearly and completely describe the technical solutions in the embodiments of the present invention with reference to the drawings in the embodiments of the present invention. Obviously, the described embodiments are only some embodiments of the present invention, rather than all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those of ordinary skill in the art without creative efforts belong to the scope of protection of the present invention.

[0049] As Figure 1 shown, Figure 1It is a flowchart of a method for querying automotive coatings provided by an embodiment of the present invention. The method for querying automotive coatings includes the steps:

[0050] 101. Obtain a full-surface image of the vehicle to be queried, and determine the color complexity of the vehicle to be queried based on the full-surface image. The full-surface image is obtained by stitching together surface images of the vehicle to be queried.

[0051] 102. Based on the color complexity of the vehicle to be queried, match a target feature extraction model corresponding to the color complexity in a model library. The model library includes multiple trained feature extraction models, and different trained feature extraction models correspond to different color complexities.

[0052] 103. Use the target feature extraction model to extract features from the full-surface image to obtain the external color features of the vehicle to be queried.

[0053] 104. According to the external color features, query the coating data of the vehicle to be queried in a coating database. In the coating database, different external color features correspond to different coating data.

[0054] In this embodiment, a high-definition camera or professional photography equipment can be used to photograph each surface (such as the roof, body, doors, rear of the vehicle, etc.) of the vehicle to be queried. The captured images should ensure uniform light and avoid interference factors such as shadows and reflections to ensure image quality. After obtaining the captured images, an image processing software or a professional image stitching algorithm can be used to stitch the captured surface images into a complete full-surface image.

[0055] Perform color analysis on the stitched full-surface image, calculate parameters such as the category and distribution of colors in the image, and evaluate the color complexity of the vehicle to be queried based on the category and distribution of colors in the image. The color complexity is used to represent the diversity of the automotive surface coatings. The higher the color complexity, the more diverse the automotive surface coatings. It can be divided into three levels: simple, medium, and complex, or can be divided according to the value of the color complexity.

[0056] A model library containing multiple trained feature extraction models can be established. Each feature extraction model is optimized for a specific color complexity to ensure accurate extraction of color features under different color complexities.

[0057] The training of the feature extraction model can be based on a large number of automotive images and corresponding coating data, and machine learning or deep learning algorithms can be used.

[0058] According to the color complexity of the vehicle to be queried, a corresponding target feature extraction model is matched from the model library. The target feature extraction model is used to extract features from the full-surface image of the vehicle to be queried, and key features that can represent the exterior color of the vehicle to be queried are extracted.

[0059] A database containing various coating data can be established in advance, and each coating data is associated with specific exterior color features. The extracted exterior color features are compared with the query features in the coating database, and one or more query features with successful comparison are determined as the target query features, and the coating data associated with the target query features is used as the coating data of the vehicle to be queried.

[0060] In one possible embodiment, the queried coating data can also be presented to the user in an intuitive manner, such as displaying the coating name, brand, color number, formula, etc. Detailed information about the coating, such as usage instructions and precautions, can be provided to help the user better understand the coating data of the vehicle to be queried.

[0061] In one possible embodiment, pictures or videos of the coating in actual application can also be shown to enhance the user's intuitive feeling and improve the user experience.

[0062] In the embodiment of the present invention, a full-surface image of the vehicle to be queried is obtained, and the color complexity of the vehicle to be queried is determined based on the full-surface image. The full-surface image is obtained by stitching surface images of the vehicle to be queried; based on the color complexity of the vehicle to be queried, a target feature extraction model corresponding to the color complexity is matched in the model library. The model library includes multiple trained feature extraction models, and different trained feature extraction models correspond to different color complexities; through the target feature extraction model, features are extracted from the full-surface image to obtain the exterior color features of the vehicle to be queried; according to the exterior color features, the coating data of the vehicle to be queried is queried in the coating database. In the coating database, different exterior color features correspond to different coating data. By introducing a color complexity evaluation mechanism, the present invention realizes the intelligent analysis of the full-surface image of the vehicle, dynamically selects the most suitable feature extraction model according to the analysis result, not only improves the accuracy of color recognition, but also effectively reduces the waste of computing resources. At the same time, by constructing a refined coating database, the efficient matching between the exterior color features and the coating data is realized, greatly shortening the coating query time and improving the overall vehicle coating query efficiency.

[0063] It can be understood that in the specific implementation of the present application, related data such as image data and coating data are involved. When the embodiments of the present application are applied to specific products or technologies, user permission or consent is required, and the collection, use, and processing of related data need to comply with relevant laws, regulations, and standards in relevant countries and regions.

[0064] Optionally, the step of determining the color complexity of the vehicle to be queried based on the full-surface image includes:

[0065] Performing image segmentation on the full-surface image to obtain image blocks of different colors and the corresponding image block information for each image block, where each image block corresponds to one piece of image block information, and the image block information includes the image block color and the image block area;

[0066] Based on the number of the image blocks and the image block information corresponding to each of the image blocks, determining the color complexity of the vehicle to be queried.

[0067] In this embodiment, an image segmentation algorithm (such as threshold segmentation, region growing, edge detection, etc.) can be used to segment the full-surface image into multiple image blocks with relatively consistent colors. Each image block represents a color region, thereby obtaining a set of image blocks of different colors, and each set of image blocks corresponds to multiple image blocks of one color.

[0068] For each segmented image block, the image block color can be determined by calculating the average color value of the pixels within the image block, and the RGB, HSV, or Lab color space, etc. can be used to represent the image block color.

[0069] For each segmented image block, the area information of each image block can also be extracted, that is, the total number of pixels within the image block or the proportion relative to the full-surface image.

[0070] After the image segmentation is completed, the total number of the segmented image blocks is counted. The larger the number of image blocks, the more frequent the color changes in the full-surface image, the higher the color complexity, that is, the more coatings are used on the vehicle to be queried.

[0071] The color distribution of each image block can be analyzed to determine the color complexity. If the types of colors in the image block are numerous and the differences between the colors are large, then the color complexity will also increase accordingly. On the contrary, if the types of colors in the image block are few and the differences between the colors are small, then the color complexity will also decrease accordingly.

[0072] In a possible embodiment, when evaluating the color complexity, the factor of the image block area can also be considered. If a certain color region (i.e., the image block) occupies a large area while other color regions are small, then the existence of this "dominant color" may reduce the overall color complexity. On the contrary, if the areas of multiple color regions are comparable and the color differences are significant, then the color complexity will be higher.

[0073] In a possible embodiment, the color complexity can also be comprehensively calculated by combining the color category and the area of the image block. Specifically, a comprehensive integral can be designed to calculate the color complexity. This metric can include the weighted value of the number of image blocks (considering the area factor), the measure of color difference (such as the Euclidean distance in the color space), etc. Specifically, according to the integral calculation formula, a quantified color complexity value can be obtained by combining the number, color, and area information of the image blocks.

[0074] By segmenting the image, the complex full-surface image is simplified into multiple image blocks with relatively consistent colors, which can reduce the computational amount and improve the overall processing efficiency. Through the above color complexity evaluation method, it can adapt to the car surfaces with different color distributions and complexities, and can process various types of cars more flexibly, that is, whether it is a single-color body or a complex design with multi-color splicing, it can be adapted by adjusting the calculation of the color complexity.

[0075] Optionally, the step of determining the color complexity of the car to be queried based on the number of the image blocks and the image block information corresponding to each of the image blocks includes:

[0076] Determine the number of the image blocks;

[0077] Based on the colors of the image blocks, calculate the color similarity between the image blocks;

[0078] Determine the image blocks with the color similarity greater than the similarity threshold as the same-color image blocks belonging to the same color category, and add up the image areas corresponding to the same-color image blocks to obtain the color areas corresponding to each category of colors;

[0079] Based on the number of the image blocks, the distance between the same-color image blocks, and the color areas corresponding to each category of colors, determine the color complexity of the car to be queried.

[0080] In this embodiment, an image segmentation algorithm (such as K-means clustering, SLIC superpixel segmentation, etc.) can be used to segment the full-surface image of the car into multiple image blocks with relatively consistent colors. The image blocks of different colors represent different color regions in the image. After the image segmentation is completed, K image blocks can be obtained, and the number of color categories of the image blocks can be statistically obtained, denoted as N, that is, there are N image blocks of color categories, and the number K of the image blocks is greater than or equal to N.

[0081] The color features of each image block can be calculated, and the color similarity between the image blocks can be calculated based on these features. The above color features can be the color feature values in modes such as RGB value, HSV value, Lab value, etc.

[0082] For each image patch, its color feature vector can be extracted and compared with the color feature vectors of other image patches to calculate the similarity value between them. A K×K similarity matrix is obtained, where the matrix elements represent the similarity between the corresponding image patches.

[0083] Set a similarity threshold T, and consider the image patches with similarity greater than T as the same-color image patches belonging to the same color category. By traversing the similarity matrix, all the same-color image patches can be found and divided into different color categories.

[0084] For each color category, calculate its corresponding color area. Specifically, it can be obtained by adding up the areas of the same-color image patches corresponding to this color category. The area of an image patch can be obtained by calculating the sum of the number of pixels within the image patch.

[0085] In a possible embodiment, after obtaining the color area of each color category, the area proportion can be further calculated, that is, the proportion of the area of each color category in the total area.

[0086] Determine the color complexity based on the number of image patches, the distance between the same-color image patches, and the color areas corresponding to each category of colors. Specifically, an integral function can be designed to quantify the color complexity.

[0087] A possible integral function is: Color complexity = f(K, D, A)

[0088] Where N is the number of image patches, D is a certain measure of the distance between the same-color image patches (such as the average distance, the maximum distance, etc.), and A is the combination of the color areas corresponding to each category of colors (such as the variance of the area proportion, entropy, etc.).

[0089] Specifically, the color complexity can be calculated as:

[0090] Color complexity = w1*K + w2*D_avg + w3*Var(A)

[0091] Where w1, w2, and w3 are weight coefficients that need to be adjusted according to the actual situation. D_avg represents the average distance between the same-color image patches, and Var(A) represents the variance of the area proportion of the colors. It can be seen that the more the number of image patches K, the more frequent the color changes, and the higher the color complexity. The greater the distance D between the same-color image patches, the more dispersed the color distribution, and the higher the color complexity. The greater the variance or entropy of the area proportion A of the colors, the more uneven the color distribution, and the higher the color complexity.

[0092] Through the color complexity, an objective quantitative index can be provided to evaluate the color distribution and changes on the car surface. The above color complexity calculation method is more accurate and repeatable than subjective evaluation, avoiding personal bias and subjective judgment.

[0093] Optionally, the step of determining the color complexity of the vehicle to be queried based on the number of the image blocks, the distance between the same-color image blocks, and the color area corresponding to each category color includes:

[0094] Calculating the central distance between the same-color image blocks;

[0095] Based on the color area corresponding to each category color, calculating the average color area corresponding to all category colors;

[0096] Calculating the color complexity of the vehicle to be queried through a preset formula, where the preset formula is:

[0097]

[0098] where f is the color complexity, N is the number of color categories, M n is the set of same-color image blocks in the nth color category, num(M n ) is the number of same-color image blocks in the nth color category, S n is the area of the nth color category, is the average color area corresponding to all category colors, i is the ith same-color image block in the nth color category, j is the jth same-color image block in the nth color category, o is the image center of all same-color image blocks in the nth color category, D(i, j) is the central distance between the ith same-color image block and the jth same-color image block in the nth color category, and D(i, o) is the distance between the ith same-color image block and the image center of all same-color image blocks in the nth color category in the nth color category.

[0099] It can be seen from the above formula that the more the number of color categories N is, the more frequent the color change is, and the higher the color complexity is. In a color category, the more the number of same-color image blocks is, the more dispersed the image blocks are, and the higher the color complexity is. It means that the greater the deviation between the color area of a color category and the average color area is, the higher the color complexity is. |D(i, j)-D(i, o)| means that in a color category, the greater the distance between one image block and another image block and the image center of all same-color image blocks of this image block is, the farther the interval between the same-color image blocks is, and the higher the color complexity is.

[0100] Optionally, before the step of matching the target feature extraction model corresponding to the color complexity in the model library based on the color complexity of the vehicle to be queried, the method further includes:

[0101] Obtain a sample image set and a feature extraction model to be trained. The sample image set includes multiple sample images corresponding to different paint data and appearance color labels. Each sample image corresponds to one appearance color label. The sample images are full-surface images of sample cars. The feature extraction model to be trained is used to extract appearance color features;

[0102] Calculate the color complexity of each of the sample images. Each sample image corresponds to one color complexity;

[0103] Based on the color complexity, divide the multiple sample images to obtain multiple training sets. Each training set includes multiple sample images corresponding to the color complexity;

[0104] Based on the multiple training sets, train the feature extraction model to be trained respectively to obtain trained feature extraction models corresponding to different color complexities;

[0105] Associate the trained feature extraction model with the color complexity and store it in the model library.

[0106] In this embodiment, a sample image set containing multiple sample images can be collected. Specifically, images of cars with known paint data can be collected and stitched according to the specifications of full-surface images to obtain sample images of cars with known paint data. The above sample images represent different paint data and appearance colors. Each sample image corresponds to a clear appearance color label. Here, the sample images are full-surface images of sample cars.

[0107] At the same time, obtain a feature extraction model to be trained. The feature extraction model to be trained can extract the appearance color features in the image and prepare for subsequent color complexity calculation and training. It can be a deep convolutional neural network model.

[0108] For each sample image, the color complexity of the sample image can be calculated by the above color complexity calculation method. Each sample image will obtain a color complexity value corresponding to it.

[0109] According to the calculated color complexity, divide the sample image set into multiple training sets. Each training set contains multiple sample images with similar color complexities. By dividing the sample image set according to the color complexity, in subsequent model training, the model can perform more effective feature learning and extraction for images with different color complexities. For example, the color complexity can be divided into multiple intervals, and all sample images whose color complexity falls within an interval are used as a training set. Then, multiple intervals correspond to multiple training sets, and each interval corresponds to a training set.

[0110] Using multiple divided training sets, train the feature extraction model to be trained respectively. Each training set corresponds to a training process, aiming to enable the model to learn the feature representation of images under this color complexity.

[0111] The training process can be supervised training. The feature extraction model to be trained extracts the appearance color features of the sample images, obtains the output appearance color features corresponding to the sample images, calculates the loss between the output appearance color features corresponding to the sample images and the appearance color labels corresponding to the sample images, obtains the error loss between the output appearance color features corresponding to the sample images and the appearance color labels corresponding to the sample images. Taking minimizing the error loss as the optimization goal, adjust the model parameters of the feature extraction model to be trained through the backpropagation algorithm, iterate the adjustment process of the model parameters until the error loss reaches the target value or the number of iterations reaches the preset number, then stop the training process to obtain the trained feature extraction model.

[0112] After all the training sets are completed, multiple feature extraction models trained for different color complexities can be obtained. Associate the trained feature extraction models with their corresponding color complexities and store these associated models in the model library. In this way, in subsequent applications, the most suitable feature extraction model can be quickly matched in the model library according to the color complexity of the car to be queried. At the same time, storing these associated models in the model library is also for subsequent use and maintenance.

[0113] Optionally, before the step of querying the paint data of the car to be queried in the paint database according to the appearance color features, the method further includes:

[0114] Obtain the paint data corresponding to the sample image;

[0115] After obtaining the trained feature extraction models corresponding to different color complexities, extract the appearance color features of the sample image through the trained feature extraction model to obtain the query features corresponding to the sample image;

[0116] Associate the query features with the paint data and store them in the paint database, and each query feature corresponds to one paint data.

[0117] In this embodiment, for one paint data, multiple sample images of sample cars using this paint data can be collected, so that a series of sample images and the paint data corresponding to these images can be collected or obtained. The paint data may include information such as the type, color code, brand, and formula of the paint.

[0118] After the feature extraction model is trained, these trained feature extraction models can be used to extract features from the sample images of the corresponding color complexity collected previously. For the same set of paint data, that is, cars of the same color and paint, there can be multiple sample images. The trained corresponding feature extraction models are used to extract features from these sample images, obtaining multiple appearance color features, and calculating the average feature of the multiple appearance color features as the query feature corresponding to this set of paint data.

[0119] Query features are an abstract representation of the color information in the sample images, and they can be used to find matching paint data in the database.

[0120] The extracted query features need to be associated with the corresponding paint data, and each query feature will be marked or linked to the paint data of its corresponding group. These associated data (including query features and paint data) will be stored in a dedicated paint database. This paint database can be regarded as a knowledge base, which stores information about different paint data and their visual features.

[0121] In the database, each query feature uniquely corresponds to a set of paint data, which ensures the accuracy and efficiency of subsequent queries.

[0122] Optionally, after the step of querying the paint data of the to-be-query car in the paint database according to the appearance color feature, the method further includes:

[0123] In the full-surface image, determine the image positions corresponding to each appearance color feature;

[0124] Through guiding lines, display the paint data corresponding to the appearance color feature at the corresponding image positions.

[0125] The above full-surface image is the full-surface image for the above query (i.e., the complete image of the to-be-query car), and accurately locates the specific positions of each appearance color feature on this image.

[0126] The appearance color features can be regressed to the full-surface image through a regression algorithm, so as to obtain the image positions corresponding to each appearance color feature.

[0127] After determining the image positions corresponding to each appearance color feature, during the subsequent display process, the queried paint data can be accurately associated with the corresponding regions on the image.

[0128] After determining the positions of each appearance color feature in the image, the paint data corresponding to them can be displayed at these positions through guiding lines (such as arrows, lines, or annotation boxes, etc.).

[0129] This display method can intuitively tell the user which specific area in the image corresponds to a certain paint data (such as color code, paint type, etc.).

[0130] This not only improves the readability and understandability of the query results, but also helps the user understand and make decisions more quickly. For example, when choosing car repair or modification, which paint can be used for a certain area.

[0131] In the embodiments of the present invention, through image processing and visualization techniques, complex paint data is combined with real car images, thereby providing the user with an accurate and convenient query and decision support tool. This has significant practical value and convenience in the automotive-related industries, especially in scenarios involving paint selection and application.

[0132] Such as Figure 2 As shown, the embodiments of the present invention provide a car paint query device, which includes:

[0133] A first acquisition module 201, configured to acquire a full-surface image of the car to be queried, and determine the color complexity of the car to be queried based on the full-surface image, where the full-surface image is obtained by stitching surface images of the car to be queried;

[0134] A matching module 202, configured to match a target feature extraction model corresponding to the color complexity in a model library based on the color complexity of the car to be queried, where the model library includes multiple trained feature extraction models, and different trained feature extraction models correspond to different color complexities;

[0135] A processing module 203, configured to perform feature extraction on the full-surface image through the target feature extraction model to obtain the external color feature of the car to be queried;

[0136] A query module 204, configured to query the paint data of the car to be queried in a paint database according to the external color feature, where in the paint database, different external color features correspond to different paint data.

[0137] Optionally, the first acquisition module 201 is further configured to:

[0138] Perform image segmentation on the full-surface image to obtain image blocks of different colors and the image block information corresponding to the image blocks, each image block corresponding to an image block information, and the image block information including the image block color and the image block area;

[0139] Determine the color complexity of the car to be queried based on the number of the image blocks and the image block information corresponding to each of the image blocks.

[0140] Optionally, the first acquisition module 201 is further configured to:

[0141] Determine the number of the image blocks;

[0142] Calculate the color similarity between the image blocks based on the colors of the image blocks;

[0143] Determine the image blocks with color similarity greater than the similarity threshold as the same-color image blocks belonging to the same color category, add up the image areas corresponding to the same-color image blocks, and obtain the color areas corresponding to each category of colors;

[0144] Determine the color complexity of the vehicle to be queried based on the number of the image blocks, the distances between the same-color image blocks, and the color areas corresponding to each category of colors.

[0145] Optionally, the first acquisition module 201 is further configured to:

[0146] Calculate the central distance between the same-color image blocks;

[0147] Calculate the average color area corresponding to all categories of colors based on the color areas corresponding to each category of colors;

[0148] Calculate the color complexity of the vehicle to be queried through a preset formula, where the preset formula is:

[0149]

[0150] where f is the color complexity, N is the number of color categories, M n is the set of same-color image blocks in the nth color category, num(M n ) is the number of same-color image blocks in the nth color category, S n is the area of the nth color category, is the average color area corresponding to all categories of colors, i is the ith same-color image block in the nth color category, j is the jth same-color image block in the nth color category, o is the image center of all same-color image blocks in the nth color category, D(i, j) is the central distance between the ith same-color image block and the jth same-color image block in the nth color category, and D(i, o) is the distance between the ith same-color image block and the image center of all same-color image blocks in the nth color category in the nth color category.

[0151] Optionally, the device further includes:

[0152] A second acquisition module, configured to acquire a sample image set and an untrained feature extraction model to be trained. The sample image set includes a plurality of sample images corresponding to different coating data and appearance color labels, and each sample image corresponds to one appearance color label. The sample images are full-surface images of sample cars, and the untrained feature extraction model is used to extract appearance color features;

[0153] A calculation module, configured to calculate the color complexity of each of the sample images, and each sample image corresponds to one color complexity;

[0154] A partitioning module, configured to partition the plurality of sample images based on the color complexity to obtain a plurality of training sets, and each training set includes a plurality of the sample images corresponding to the color complexity;

[0155] A training module, configured to train the untrained feature extraction model respectively based on the plurality of training sets to obtain trained feature extraction models corresponding to different color complexities;

[0156] An association module, configured to associate the trained feature extraction models with the color complexities and store them in a model library.

[0157] Optionally, the apparatus further includes:

[0158] A third acquisition module, configured to acquire coating data corresponding to the sample images;

[0159] A feature extraction module, configured to, after obtaining the trained feature extraction models corresponding to different color complexities, extract appearance color features of the sample images through the trained feature extraction models to obtain query features corresponding to the sample images;

[0160] A storage module, configured to associate the query features with the coating data and store them in the coating database, and each query feature corresponds to one coating data.

[0161] Optionally, the apparatus further includes:

[0162] A position determination module, configured to determine image positions corresponding to respective appearance color features in the full-surface images;

[0163] A display module, configured to display the coating data corresponding to the appearance color features at the corresponding image positions through guiding lines.

[0164] It should be noted that the car coating query apparatus provided in the embodiments of the present invention can be applied to devices such as computers and servers capable of performing car coating query methods.

[0165] The automotive paint query device provided by the embodiments of the present invention can implement each process implemented by the automotive paint query method in the above method embodiments, and can achieve the same beneficial effects. To avoid repetition, it will not be elaborated here.

[0166] See Figure 3 , Figure 3 which is a schematic structural diagram of an electronic device provided by an embodiment of the present invention. As Figure 3 shown, it includes: a memory 302, a processor 301, and a computer program of the automotive paint query method stored on the memory 302 and executable on the processor 301, where:

[0167] The processor 301 is configured to call the computer program stored in the memory 302 and execute the following steps:

[0168] Obtain a full-surface image of the vehicle to be queried, and determine the color complexity of the vehicle to be queried based on the full-surface image, where the full-surface image is obtained by stitching surface images of the vehicle to be queried;

[0169] Based on the color complexity of the vehicle to be queried, match a target feature extraction model corresponding to the color complexity in the model library, where the model library includes multiple trained feature extraction models, and different trained feature extraction models correspond to different color complexities;

[0170] Extract features from the full-surface image through the target feature extraction model to obtain the external color features of the vehicle to be queried;

[0171] Query the paint data of the vehicle to be queried in the paint database according to the external color features, where in the paint database, different external color features correspond to different paint data.

[0172] Optionally, the step of the processor 301 determining the color complexity of the vehicle to be queried based on the full-surface image includes:

[0173] Perform image segmentation on the full-surface image to obtain image blocks of different colors and the corresponding image block information for the image blocks, each image block corresponding to one piece of image block information, and the image block information including the image block color and the image block area;

[0174] Determine the color complexity of the vehicle to be queried based on the number of the image blocks and the image block information corresponding to each of the image blocks.

[0175] Optionally, the step of the processor 301 determining the color complexity of the vehicle to be queried based on the number of the image blocks and the image block information corresponding to each of the image blocks includes:

[0176] Determine the number of the image blocks;

[0177] Calculate the color similarity between the image blocks based on the colors of the image blocks;

[0178] Determine the image blocks with the color similarity greater than the similarity threshold as the same-color image blocks belonging to the same color category, add up the image areas corresponding to the same-color image blocks, and obtain the color areas corresponding to each category of colors;

[0179] Determine the color complexity of the vehicle to be queried based on the number of the image blocks, the distance between the same-color image blocks, and the color areas corresponding to each category of colors.

[0180] Optionally, the step of determining the color complexity of the vehicle to be queried by the processor 301 based on the number of the image blocks, the distance between the same-color image blocks, and the color areas corresponding to each category of colors includes:

[0181] Calculate the central distance between the same-color image blocks;

[0182] Calculate the average color area corresponding to all categories of colors based on the color areas corresponding to each category of colors;

[0183] Calculate the color complexity of the vehicle to be queried through a preset formula, and the preset formula is:

[0184]

[0185] where f is the color complexity, N is the number of color categories, M n is the set of same-color image blocks in the nth color category, num(M n ) is the number of same-color image blocks in the nth color category, S n is the area of the nth color category, is the average color area corresponding to all categories of colors, i is the ith same-color image block in the nth color category, j is the jth same-color image block in the nth color category, o is the image center of all same-color image blocks in the nth color category, D(i, j) is the central distance between the ith same-color image block and the jth same-color image block in the nth color category, and D(i, o) is the distance between the ith same-color image block and the image center of all same-color image blocks in the nth color category in the nth color category.

[0186] Optionally, before the step of matching the target feature extraction model corresponding to the color complexity in the model library based on the color complexity of the vehicle to be queried, the method further executed by the processor 301 includes:

[0187] Obtain a sample image set and a feature extraction model to be trained. The sample image set includes multiple sample images corresponding to different coating data and appearance color labels. Each sample image corresponds to one appearance color label. The sample image is a full-surface image of a sample car. The feature extraction model to be trained is used to extract appearance color features;

[0188] Calculate the color complexity of each sample image. Each sample image corresponds to one color complexity;

[0189] Based on the color complexity, divide the multiple sample images to obtain multiple training sets. Each training set includes multiple sample images corresponding to the color complexity;

[0190] Based on the multiple training sets, train the feature extraction model to be trained respectively to obtain trained feature extraction models corresponding to different color complexities;

[0191] Associate the trained feature extraction model with the color complexity and store it in the model library.

[0192] Optionally, before the step of querying the coating data of the car to be queried in the coating database according to the appearance color feature, the method executed by the processor 301 further includes:

[0193] Obtain the coating data corresponding to the sample image;

[0194] After obtaining the trained feature extraction models corresponding to different color complexities, use the trained feature extraction models to extract appearance color features from the sample images to obtain query features corresponding to the sample images;

[0195] Associate the query features with the coating data and store them in the coating database. Each query feature corresponds to one coating data.

[0196] Optionally, after the step of querying the coating data of the car to be queried in the coating database according to the appearance color feature, the method executed by the processor 301 further includes:

[0197] In the full-surface image, determine the image positions corresponding to each appearance color feature;

[0198] Through guide lines, display the coating data corresponding to the appearance color features at the corresponding image positions.

[0199] It should be noted that the electronic device provided by the embodiments of the present invention can be applied to devices such as computers and servers that can perform the automotive paint query method.

[0200] The electronic device provided by the embodiments of the present invention can implement each process realized by the automotive paint query method in the above method embodiments, and can achieve the same beneficial effects. To avoid repetition, it will not be elaborated here.

[0201] The embodiments of the present invention also provide a computer-readable storage medium. A computer program is stored on the computer-readable storage medium. When the computer program is executed by a processor, it implements each process of the automotive paint query method provided by the embodiments of the present invention, and can achieve the same technical effects. To avoid repetition, it will not be elaborated here.

[0202] Those of ordinary skill in the art can understand that all or part of the processes in implementing the methods of the above embodiments can be completed by instructing relevant hardware through a computer program. The computer program can be stored in a computer-readable storage medium. When the program is executed, it can include the processes of the embodiments of the above methods. Among them, the computer-readable storage medium can be a magnetic disk, an optical disc, a read-only memory (ROM), or a random access memory (RAM), etc.

[0203] The above-disclosed are only the preferred embodiments of the present invention. Of course, the scope of the rights of the present invention cannot be limited thereby. Therefore, equivalent changes made according to the claims of the present invention still fall within the scope covered by the present invention.

Claims

1. A method for querying automobile coatings, characterized in that: The method comprises the following steps: Acquire a full-surface image of the car to be queried, and determine the color complexity of the car to be queried based on the full-surface image, wherein the full-surface image is obtained by splicing various surface images of the car to be queried; Based on the color complexity of the automobile to be queried, a target feature extraction model corresponding to the color complexity is matched in a model library, wherein the model library includes a plurality of trained feature extraction models, and different trained feature extraction models correspond to different color complexities; By using the target feature extraction model, feature extraction is performed on the full surface image to obtain the exterior color features of the vehicle to be queried; According to the exterior color feature, the paint data of the to-be-queried vehicle is searched in a paint database, wherein different exterior color features correspond to different paint data in the paint database.

2. The automobile coating query method according to claim 1, characterized in that: The step of determining the color complexity of the automobile to be queried based on the full surface image comprises: Performing image segmentation on the full surface image to obtain image blocks of different colors and image block information corresponding to the image blocks, wherein each image block corresponds to an image block information, and the image block information includes image block color and image block area; Based on the number of the image blocks and the image block information corresponding to each of the image blocks, the color complexity of the automobile to be queried is determined.

3. The automobile coating query method according to claim 2, characterized in that: The step of determining the color complexity of the automobile to be queried based on the number of the image blocks and the image block information corresponding to each of the image blocks includes: Determining the number of the image blocks; Based on the colors of the image blocks, calculating the color similarity between the image blocks; Determine the image blocks whose color similarity is greater than a similarity threshold as same-color image blocks belonging to the same color, add and sum the image areas corresponding to the same-color image blocks to obtain the color areas corresponding to the colors of each category; The color complexity of the car to be queried is determined based on the number of the image blocks, the distance between the image blocks of the same color, and the color areas corresponding to the colors of each category.

4. The automobile coating query method according to claim 3, characterized in that: The step of determining the color complexity of the to-be-queried automobile based on the number of the image blocks, the distance between the same-color image blocks, and the color areas corresponding to the colors of each category includes: Calculate the center distance between image blocks of the same color; Based on the color area corresponding to each category color, calculate the average color area corresponding to all category colors; The color complexity of the car to be queried is calculated by a preset formula, and the preset formula is: Among them, f is the color complexity, N is the number of color categories, M n is the set of image blocks of the same color in the nth color category, num(M n ) is the number of image blocks of the same color in the nth color category, S n is the area of ​​the nth color category, is the average color area corresponding to all categories of colors, i is the ith image block of the same color in the nth color category, j is the jth image block of the same color in the nth color category, o is the image center of all image blocks of the same color in the nth color category, D(i,j) is the center distance between the ith image block of the same color in the nth color category and the jth image block of the same color, and D(i,o) is the distance between the ith image block of the same color in the nth color category and the image center of all image blocks of the same color in the nth color category.

5. The automobile coating query method according to any one of claims 1 to 4, characterized in that: Before the step of matching a target feature extraction model corresponding to the color complexity in a model library based on the color complexity of the automobile to be queried, the method further includes: Obtaining a sample image set and a feature extraction model to be trained, wherein the sample image set includes a plurality of sample images corresponding to different paint data and an appearance color label, each of the sample images corresponds to one appearance color label, the sample image is a full-surface image of a sample car, and the feature extraction model to be trained is used to extract appearance color features; Calculating the color complexity of each of the sample images, each of the sample images corresponds to one color complexity; Based on the color complexity, the plurality of sample images are divided to obtain a plurality of training sets, each of the training sets including a plurality of sample images corresponding to the color complexity; Based on the plurality of training sets, the feature extraction models to be trained are trained respectively to obtain trained feature extraction models corresponding to different color complexities; The trained feature extraction model is associated with the color complexity and stored in a model library.

6. The automobile coating query method according to claim 5, characterized in that: Before the step of searching for the paint data of the to-be-queried vehicle in a paint database according to the exterior color feature, the method further comprises: Acquiring coating data corresponding to the sample image; After obtaining the trained feature extraction models corresponding to the different color complexities, extracting the surface color features of the sample image through the trained feature extraction model to obtain the query features corresponding to the sample image; The query feature is associated with the coating data and stored in the coating database, each query feature corresponds to one coating data.

7. The automobile coating query method according to any one of claims 1 to 4, characterized in that: After the step of searching for the paint data of the vehicle to be queried in the paint database according to the exterior color feature, the method further comprises: In the full surface image, determining the image position corresponding to each surface color feature; The paint data corresponding to the surface color feature is displayed at the corresponding image position through the guide line.

8. An automobile paint query device, characterized in that: The automobile coating query device comprises: A first acquisition module is used to acquire a full-surface image of the car to be queried, and determine the color complexity of the car to be queried based on the full-surface image, wherein the full-surface image is obtained by splicing various surface images of the car to be queried; A matching module, used for matching a target feature extraction model corresponding to the color complexity in a model library based on the color complexity of the automobile to be queried, wherein the model library includes a plurality of trained feature extraction models, and different trained feature extraction models correspond to different color complexities; A processing module, used to extract features from the full surface image through the target feature extraction model to obtain the exterior color features of the vehicle to be queried; The query module is used to query the paint data of the car to be queried in the paint database according to the appearance color characteristics, and in the paint database, different appearance color characteristics correspond to different paint data.

9. An electronic device, characterized in that: include: A memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor implements the steps of the automobile paint query method as claimed in any one of claims 1 to 7 when executing the computer program.

10. A computer-readable storage medium, characterized in that: The computer-readable storage medium stores a computer program, and when the computer program is executed by a processor, the steps in the automobile paint query method according to any one of claims 1 to 7 are implemented.

Images