A wheel hub model retrieval method, storage medium and device

Abstract

The present invention discloses a wheel hub model retrieval method, comprising the following steps: receiving a first wheel hub picture to be retrieved; obtaining first feature information according to the first wheel hub picture; obtaining a first view of each wheel hub model according to all wheel hub models in a model database; obtaining second feature information of each wheel hub model according to the first view; sorting all wheel hub models according to the first feature information and the second feature information; and also discloses a picture retrieval storage medium and device. According to the first feature information of the first wheel hub picture, the second feature information of the first view is compared, and pictures in the wheel hub model database are listed for retrieval. The process is simple and improves the accuracy of wheel hub graphic similarity determination, effectively reduces the calculation amount of the picture database, and achieves the purpose of saving costs and improving retrieval efficiency.

Description

Technical Field

[0001] The present invention relates to the technical field of image recognition, and in particular to a wheel hub model retrieval method, storage medium and equipment. Background Art

[0002] Image retrieval has developed rapidly in the past decade. It uses computers to analyze images, establish image feature descriptions and store them in image feature libraries. When a user inputs a query image, the same feature extraction method is used to extract the features of the query image to obtain query information. Then, the similarity of each feature in the query feature library is calculated under a certain similarity measurement criterion. Finally, the images are sorted by similarity and the corresponding images are output in sequence. However, there is currently no retrieval specifically for wheel images, and other image retrieval methods have some shortcomings, mainly manifested in the complex calculation process and the inability to accurately and efficiently implement image retrieval. Summary of the invention

[0003] The present invention aims to solve at least one of the technical problems existing in the prior art. To this end, the present invention provides a wheel hub model retrieval method, storage medium and device.

[0004] To achieve the above-mentioned objectives, the present invention provides the following technical solutions: a wheel hub model retrieval method comprises the following steps: receiving a first wheel hub image to be retrieved; obtaining first feature information based on the first wheel hub image; obtaining a first view of each wheel hub model based on all wheel hub models in a model database; obtaining second feature information of each wheel hub model based on the first view; and sorting all wheel hub models according to the first feature information and the second feature information.

[0005] To be more specific, in the above technical solution, the first view of each wheel hub model is obtained based on all the wheel hub models in the model database, which specifically includes: the wheel hub model includes multiple triangular vertices, and a square box is calculated based on the coordinates of each vertex of all the triangles, and the square box includes all vertices. The square box is orthogonally projected to obtain the first view of each of the wheel hub models.

[0006] More specifically, in the above technical solution, the first view is a front view or a rear view.

[0007] To be more specific, in the above technical solution, the first characteristic information includes the number of hollowed-out areas, the number of types of hollowed-out areas, and the area size of each hollowed-out area in the first wheel hub image; the second characteristic information includes the number of hollowed-out areas, the number of types of hollowed-out areas, and the area size of each hollowed-out area in the first view.

[0008] More specifically, in the above technical solution, color processing is performed on the first wheel hub image before acquiring the first characteristic information, and the color of the hollowed-out portion of the first wheel hub image is set to a color consistent with the background color.

[0009] To be more specific, in the above technical solution, all wheel hub models are sorted according to the first feature information and the second feature information, specifically including: judging whether the number of hollowed-out parts and the number of types of hollowed-out area types in the first hub picture and each of the first views are consistent; if so, comparing the first hub picture with the sum of the areas of the hollowed-out parts in each of the first views, and arranging the ones with the smaller difference between the sum of the areas of the hollowed-out parts in the first view and the sum of the areas of the hollowed-out parts in the first hub picture in the front; if not, performing a fuzzy search.

[0010] More specifically, in the above technical solution, the fuzzy search includes binarizing the first wheel hub image and each of the first views, respectively obtaining multiple contours of the first wheel hub image and each of the first views, and counting the number of contours and the ratio of the total area of ​​the contours to the area of ​​the graphic, and displaying the first views with smaller difference values ​​in front;

[0011] Difference value = k1×|number of contours in the first hub image - number of contours in the first view|+k2×|ratio of total contour area to graphic area in the first hub image - ratio of total contour area to graphic area in the first view|, k1 and k2 are weight values, k1+k2=100%.

[0012] More specifically, in the above technical solution, the wheel hub model retrieval method further includes further accurately arranging the first ten views with the smallest difference values, specifically including:

[0013] Using a median filter to perform low-pass filtering on the first view and the first wheel image to remove salt and pepper noise;

[0014] Detect the obtained profile and delete the profile of the wheel hub;

[0015] The remaining contours are clustered and merged to form feature vectors for each of the first views and the first wheel hub image, and the feature vectors are composed as follows:

[0016] {K, [(A1, N1, C1), (A2, N2, C2), ..., (Ai, Ni, Ci)]}, where 1 ≤ i ≤ K, K represents the number of contour types, A i represents the average area of ​​the i-th type of contour, N i represents the number of contours of the i-th type, C iIt represents the coordinate set of the i-th type of contour on the coordinate axis with the hub center as the origin;

[0017] comparing each of the first views with a feature vector of the first wheel image;

[0018] Filter out eigenvectors with equal K values ​​and eliminate N i Eigenvectors with different values;

[0019] The contour similarity is calculated by using the linear combination of normalized moments;

[0020] k3 and k4 are weight values, k3+k4=100%.

[0021] More specifically, in the above technical solution, after the first hub image and each of the first views are binarized, an expansion and erosion algorithm is used to cluster the hollowed-out areas to obtain a clearer outline.

[0022] A computer storage medium stores a computer program, which implements the wheel hub model retrieval method as described above when executed by a processor.

[0023] A picture retrieval device, comprising:

[0024] Memory for storing computer programs;

[0025] The processor is used to implement the wheel hub model retrieval method as described above when executing the computer program.

[0026] Compared with the prior art, the beneficial effects of the present invention are: by comparing and analyzing the first feature information of the first wheel hub image to be retrieved and the second feature information of the first view, similar wheel hub images in the wheel hub model database are listed, and a wheel hub image retrieval process with a simple process and high accuracy in determining graphic similarity is achieved. The method has a simple retrieval process, improves the accuracy of determining wheel hub graphic similarity, reduces the amount of retrieval calculations in the wheel hub model database, and achieves the purpose of saving costs and improving retrieval efficiency. BRIEF DESCRIPTION OF THE DRAWINGS

[0027] Figure 1 It is a schematic diagram of the wheel hub model retrieval method of the present invention;

[0028] Figure 2 A first wheel hub picture or first view of the present invention;

[0029] Figure 3 A schematic diagram of a hollowed-out portion for obtaining a picture according to the present invention;

[0030] Figure 4 This is a schematic diagram of a picture after binarization processing according to the present invention;

[0031] Figure 5 This is a schematic diagram of a picture after the expansion and corrosion algorithm of the present invention;

[0032] Figure 6 This is a schematic diagram of a picture after low-pass filtering according to the present invention;

[0033] Figure 7 This is a schematic diagram of a picture after removing salt and pepper noise according to the present invention;

[0034] Figure 8 A schematic diagram of a picture of a detected contour of the present invention;

[0035] Fig. 9 This is a schematic diagram of a picture after deleting the outer contour of the wheel hub according to the present invention. DETAILED DESCRIPTION

[0036] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0037] The terms "first", "second", etc. (if any) in the specification and claims and the above drawings are used to distinguish similar objects and are not necessarily used to describe a specific order or sequence. It should be understood that the terms used in this way are interchangeable under appropriate circumstances so that the embodiments described herein can be implemented in an order other than that shown or described herein. In addition, the terms "including" and "having" and any variations thereof are intended to cover non-exclusive inclusions, for example, a process, method, system, product or apparatus that includes a series of steps or units is not necessarily limited to those steps or units explicitly listed, but may include other steps or units that are not explicitly listed or inherent to these processes, methods, products or apparatus.

[0038] In some scenarios, the staff needs to retrieve similar and completed wheel hub models from the model database based on the first wheel hub picture provided by the customer, so as to directly provide the relatively similar wheel hub model to the customer or provide the relatively similar wheel hub model to the customer after simple modification. Therefore, a wheel hub model retrieval method is proposed, such as Figure 1As shown, a wheel hub model retrieval method includes the following steps: receiving a first wheel hub image to be retrieved; obtaining first feature information based on the first wheel hub image; obtaining a first view of each wheel hub model based on all wheel hub models in a model database; obtaining second feature information of each wheel hub model based on the first view; and sorting all wheel hub models according to the first feature information and the second feature information.

[0039] By comparing and analyzing the first feature information of the first wheel hub image to be retrieved and the second feature information of the first view, similar wheel hub images in the wheel hub model database are listed, and a wheel hub image retrieval process with a simple process and high accuracy in graphic similarity judgment is implemented. The method has a simple retrieval process, improves the accuracy of wheel hub graphic similarity judgment, reduces the retrieval calculation amount of the wheel hub model database, and achieves the purpose of saving costs and improving retrieval efficiency.

[0040] In some embodiments, the first view of each hub model is obtained according to all hub models in the model database, specifically including: the hub model includes multiple triangular vertices, a square box is calculated according to the coordinates of each vertex of all triangles, the square box includes all vertices, and the square box is orthogonally projected to obtain the first view of each hub model. According to the triangular vertices of the front, rear, left, right, top or bottom view of the hub model, the square box of each view can be accurately obtained, the square box covers the vertices of all triangles, and the obtained first view can accurately express the shape characteristics of the first view of the hub model.

[0041] In some embodiments, the first view is a front view or a rear view. In some cases, the entire search process can be completed by obtaining the front view or the rear view of the wheel hub model. Only obtaining the front view or the rear view of the wheel hub model can greatly reduce the search calculation amount of the wheel hub model database and improve the search efficiency.

[0042] In some embodiments, the first characteristic information includes the number of hollow outs, the number of types of hollow out areas, and the area size of each hollow out in the first wheel hub image; the second characteristic information includes the number of hollow outs, the number of types of hollow out areas, and the area size of each hollow out in the first view.

[0043] like Figure 2 and Figure 3 As shown, in some embodiments, the first hub image is color processed before obtaining the first characteristic information, and the color of the hollowed-out portion of the first hub image is set to a color consistent with the background color, so that the characteristics of the hub shape in the first hub image and the shape characteristics of the hollowed-out portion can be obtained more accurately.

[0044] In some embodiments, all wheel hub models are sorted according to the first feature information and the second feature information, specifically including: determining whether the number of hollowed-out parts and the number of types of hollowed-out area types in the first hub image and each first view are consistent; if so, comparing the first hub image with the sum of the areas of the hollowed-out parts in each first view, and arranging the models with the smaller difference between the sum of the areas of the hollowed-out parts in the first view and the sum of the areas of the hollowed-out parts in the first hub image in the front; if not, performing a fuzzy search.

[0045] If the number of hollowed-out parts and the number of types of hollowed-out area in the first hub image are consistent with those in each first view retrieved, it is a precise search and a first view that is extremely similar to the first hub image can be obtained; if the number of hollowed-out parts and the number of types of hollowed-out area in the first hub image are inconsistent with those in each first view retrieved, the search scope is expanded and a fuzzy search is performed on the first hub image.

[0046] In some embodiments, the fuzzy search includes binarizing the first hub image and each first view, and the obtained image is as follows: Figure 4 As shown, the first hub image and multiple contours of each first view are obtained respectively, and the number of contours and the ratio of the total contour area to the graphic area are counted, and the first view with the smallest difference value is listed in the front for display;

[0047] Difference value = k1×|number of contours in the first hub image - number of contours in the first view|+k2×|ratio of total contour area to graphic area in the first hub image - ratio of total contour area to graphic area in the first view|, k1 and k2 are weight values, k1+k2=100%.

[0048] By calculating the difference value, the first view with a small difference value can be arranged in the front, making it easier for the searcher to obtain the most similar first view.

[0049] In some embodiments, the wheel hub model retrieval method further includes further accurately arranging the first ten first views with small difference values, specifically including:

[0050] The median filter is used to perform low-pass filtering on the first view and the first wheel image to remove the salt and pepper noise. The image is processed by low-pass filtering. Figure 6 As shown in Figure 2, after removing the salt and pepper noise, Figure 7 As shown;

[0051] The obtained contour is detected and the obtained contour is as follows Figure 8 As shown in the figure, delete the outline of the wheel hub shape. After deleting the outline of the wheel hub shape, the remaining outline is as follows Fig. 9 As shown;

[0052] The remaining contours are clustered and merged to form feature vectors for each first view and the first hub image. The feature vectors are composed as follows:

[0053] {K, [(A1, N1, C1), (A2, N2, C2), ..., (Ai, Ni, Ci)]}, where 1 ≤ i ≤ K, K represents the number of contour types, A i represents the average area of ​​the i-th type of contour, N i represents the number of contours of the i-th type, C i Represents the coordinate set of the i-th type of contour on the coordinate axis with the hub center as the origin, such as Fig. 9 As shown, i=2;

[0054] Compare each first view with the feature vector of the first wheel image;

[0055] Filter out eigenvectors with equal K values ​​and eliminate N i Eigenvectors with different values;

[0056] The contour similarity is calculated by using the linear combination of normalized moments;

[0057] k3 and k4 are weight values, k3+k4=100%.

[0058] The method of further accurately arranging the first view is relatively complicated and labor-intensive. Therefore, only performing further accurate arrangement on the first ten first views with small difference values ​​can greatly reduce the workload, improve work efficiency, and obtain a more accurate final similarity.

[0059] like Figure 5 As shown, in some embodiments, after the first hub image and each first view are binarized, an expansion and erosion algorithm is used to cluster the hollowed-out areas to obtain a clearer contour and obtain multiple contour data, and the number of contours obtained is shown in the figure.

[0060] A computer storage medium stores a computer program, which implements the above wheel hub model retrieval method when executed by a processor.

[0061] A picture retrieval device, comprising:

[0062] Memory for storing computer programs;

[0063] The processor is used to implement the above wheel hub model retrieval method when executing a computer program.

[0064] Although embodiments of the present invention have been shown and described, it will be appreciated by those skilled in the art that various changes, modifications, substitutions and variations may be made to the embodiments without departing from the principles and spirit of the present invention, and that the scope of the present invention is defined by the appended claims and their equivalents.

Claims

1. A wheel hub model retrieval method, characterized in that: The wheel hub model retrieval method comprises the following steps: receiving a first hub image to be retrieved; Acquire first feature information according to the first wheel hub image; Acquire a first view of each wheel hub model according to all wheel hub models in a model database; Acquire second feature information of each wheel hub model according to the first view; sorting all wheel hub models according to the first feature information and the second feature information; The first characteristic information includes the number of hollowing-outs, the number of types of hollowing-out areas, and the area size of each hollowing-out in the first wheel hub image; the second characteristic information includes the number of hollowing-outs, the number of types of hollowing-out areas, and the area size of each hollowing-out in the first view; Sorting all wheel hub models according to the first feature information and the second feature information specifically includes: determining whether the number of hollowed-out parts and the number of types of hollowed-out areas in the first wheel hub picture and each of the first views are consistent, and if so, comparing the first wheel hub picture with the sum of the areas of the hollowed-out parts in each of the first views, and displaying the models with the smallest difference between the sum of the areas of the hollowed-out parts in the first view and the sum of the areas of the hollowed-out parts in the first wheel hub picture in the front; if not, performing fuzzy search; The fuzzy search includes binarizing the first wheel hub image and each of the first views, respectively obtaining multiple contours of the first wheel hub image and each of the first views, and counting the number of contours and the ratio of the total area of ​​the contours to the area of ​​the graphic, and displaying the first views with smaller difference values ​​in front; Difference value = k1×|number of contours in the first hub image - number of contours in the first view|+k2×|ratio of the total contour area to the graphic area in the first hub image - ratio of the total contour area to the graphic area in the first view|, k1 and k2 are weight values, k1+k2=100%; The wheel hub model retrieval method further includes further accurately arranging the first ten views with the smallest difference values, specifically including: Using a median filter to perform low-pass filtering on the first view and the first wheel image to remove salt and pepper noise; Detect the obtained profile and delete the profile of the wheel hub; The remaining contours are clustered and merged to form feature vectors for each of the first views and the first wheel hub image, and the feature vectors are composed as follows: {K,[(A1,N1,C1),(A2,N2,C2),……,(Ai,Ni,Ci)]}, where 1≤i≤K, K represents the number of contour types, A i represents the average area of ​​the i-th type of contour, N i represents the number of contours of the i-th type, C i It represents the coordinate set of the i-th type of contour on the coordinate axis with the hub center as the origin; comparing each of the first views with a feature vector of the first wheel image; Filter out eigenvectors with equal K values ​​and eliminate N i Eigenvectors with different values; The contour similarity is calculated by using the linear combination of normalized moments; k3 and k4 are weight values, k3+k4=100%.

2. The wheel hub model retrieval method according to claim 1, characterized in that: The first view of each wheel hub model is obtained according to all the wheel hub models in the model database, specifically including: the wheel hub model includes multiple triangular vertices, a square box is calculated according to the coordinates of each vertex of all the triangles, the square box includes all the vertices, and the square box is orthogonally projected to obtain the first view of each wheel hub model.

3. The wheel hub model retrieval method according to claim 1 or 2, characterized in that: The first view is a front view or a rear view.

4. The wheel hub model retrieval method according to claim 1, characterized in that: Before acquiring the first characteristic information, color processing is performed on the first wheel hub image, and the color of the hollowed-out portion of the first wheel hub image is set to a color consistent with the background color.

5. The wheel hub model retrieval method according to claim 1, characterized in that: After binarizing the first wheel hub image and each of the first views, a dilation and erosion algorithm is used to cluster the hollowed-out areas to obtain a clearer outline.

6. A computer storage medium having a computer program stored thereon, wherein when the program is executed by a processor, the wheel hub model retrieval method as described in any one of claims 1-2, 4-5 is implemented.

7. A picture retrieval device, characterized in that: include: Memory for storing computer programs; A processor, used to implement the wheel hub model retrieval method as described in any one of claims 1-2, 4-5 when executing the computer program.

Images