Method, terminal and storage device for identifying commodity categories through LOGO

Abstract

The present invention provides a method, terminal and storage device for identifying commodity categories through LOGO, the method comprising: S101: detecting commodity display areas through scene detection models, obtaining display areas of commodities to be detected according to the commodity display areas, the commodity display areas being commodity display areas of commodities of different brands; S102: training POSM detection models and LOGO recognition models respectively, and obtaining the position of POSM, the position of LOGO and the category in the display area through the POSM detection models and the LOGO recognition models; S103: removing the LOGO in the POSM according to the position of POSM and the position of LOGO, and obtaining the category corresponding to the LOGO not removed in the display area. The present invention can distinguish commodities of different categories in the same brand through LOGO recognition, avoids the problem of needing to identify commodity packaging, reduces modeling difficulty and labor cost, effectively removes background interference through scene detection, reduces misidentification, is not easily interfered by auxiliary sales materials, and improves recognition accuracy.

Description

Technical Field

[0001] The present invention relates to the field of image recognition, and in particular to a method, a terminal and a storage device for identifying commodity categories through LOGO. Background Art

[0002] In the retail field, there are often various sub-brands under a certain brand. Take Mars Wrigley as an example. It has brands that are familiar to consumers, such as Doublemint, M&M, SNICKERS, EXTRA, Dove, and Skittles. And within each sub-brand, there are products in different packages, such as bars, boxes, cans, etc. This involves the need to identify different categories of sub-brands of goods.

[0003] To address this problem, existing technical solutions are mainly based on pure image classification. First, data of commodities of various categories are collected, and then the data categories are labeled. Then, the classification model in deep learning is used for training to obtain a recognizable classification model. This technical solution is the most general and direct method, but its disadvantages are also obvious: 1) Since they belong to the same brand, the appearance of commodities of different categories is very similar and difficult to distinguish. 2) If you want to identify all the packages of all categories of commodities, you must model all newly launched products. However, in actual modeling, due to the long modeling cycle, high labor cost and high difficulty in new product mining, efficient recognition modeling cannot be achieved. 3) The shelf background is cluttered and complex, which can easily lead to misidentification of similar commodities of non-brands. 4) The scene is complex and easily interfered by POSM (Point of Sales Materials) such as top brands and shelf strips, resulting in reduced recognition accuracy. Summary of the invention

[0004] In order to overcome the shortcomings of the prior art, the present invention proposes a method, a terminal and a storage device for identifying commodity categories through LOGO, and obtains the display area of ​​commodities of the same brand through scene detection model recognition, and obtains the position of POSM and the position of LOGO in the display area through the trained POSM detection model and LOGO recognition model, and then removes the LOGO in the POSM according to the position, so as to identify the category of the commodity according to the LOGO that has not been removed. By means of LOGO recognition, different categories of commodities of the same brand can be distinguished, and the problem of needing to identify commodity packaging is avoided, and the difficulty of modeling and labor cost are reduced. In addition, background interference is effectively removed by scene detection, which reduces misidentification, and is not easily interfered by auxiliary sales materials, thereby improving recognition accuracy.

[0005] To solve the above problem, a technical solution adopted by the present invention is: a method for identifying commodity categories through LOGO, the method for identifying commodity categories through LOGO comprising: S101: detecting a commodity display area through a scene detection model, and obtaining a display area of ​​a commodity to be detected according to the commodity display area, wherein the commodity display area is a commodity display area for commodities of different brands; S102: training a POSM detection model and a LOGO recognition model respectively, and obtaining a position of the POSM, a position of the LOGO, and a category in the display area through the POSM detection model and the LOGO recognition model; S103: removing the LOGO in the POSM according to the position of the POSM and the position of the LOGO, and obtaining a category corresponding to the LOGO that has not been removed in the display area.

[0006] Furthermore, the step of detecting the commodity display area through the scene detection model specifically includes: obtaining a display image of the commodity, preprocessing the display image, and inputting the preprocessed display image into the scene detection model to obtain location information and category label information of the commodity display area.

[0007] Furthermore, the step of acquiring the display area of ​​the commodity to be detected according to the commodity display area specifically includes: acquiring the brand of the commodity to be detected, and acquiring the display area according to the brand.

[0008] Furthermore, the steps of separately training the POSM detection model and the LOGO recognition model specifically include: separately collecting a first picture including the POSM and a second picture including the LOGO, preprocessing the first picture and the second picture, and inputting the preprocessed first picture into the POSM detection network for training, and inputting the preprocessed second picture into the LOGO detection network for training; and separately adjusting and optimizing the parameters of the POSM detection model and the LOGO recognition model formed by the training until the models converge.

[0009] Furthermore, the preprocessing includes at least one of image scaling, image expansion, and image normalization.

[0010] Furthermore, the detection networks used by the scene detection model, the POSM detection model, and the LOGO recognition model include at least one of Cascade RCNN, Faster RCNN, and YOLO.

[0011] Furthermore, the step of removing the LOGO in the POSM according to the position of the POSM and the position of the LOGO specifically includes: acquiring the LOGO overlapping with the position of the POSM according to the position of the POSM and the position of the LOGO, and removing the LOGO.

[0012] Furthermore, the step of obtaining the position and category of the LOGO in the display area through the LOGO recognition model specifically includes: preprocessing the image of the display area, performing recognition processing of the LOGO position and category on the preprocessed image through the trained LOGO recognition model, and obtaining the position and category of the LOGO according to the result of the recognition processing.

[0013] Based on the same inventive concept, the present invention also proposes an intelligent terminal, which includes a processor and a memory. The memory stores a computer program, and the computer program is used to execute the method of identifying the category of a product through a LOGO as described above.

[0014] Based on the same inventive concept, the present invention further proposes a storage device, wherein the storage device stores program data, and the program data is used to execute the method for identifying the category of a commodity through a LOGO as described above.

[0015] Compared with the prior art, the beneficial effects of the present invention are: obtaining the display area of ​​the same brand of goods through scene detection model recognition, and obtaining the position of the POSM and the position of the LOGO in the display area through the trained POSM detection model and the LOGO recognition model, and then removing the LOGO in the POSM according to the position, so as to identify the category of the goods according to the LOGO that has not been removed, and being able to distinguish different categories of goods in the same brand through LOGO recognition, avoiding the problem of needing to identify the product packaging, reducing the difficulty of modeling and labor costs, and effectively removing background interference through scene detection, reducing misidentification, and not being easily interfered by auxiliary sales materials, thereby improving recognition accuracy. BRIEF DESCRIPTION OF THE DRAWINGS

[0016] Figure 1 It is a flow chart of an embodiment of a method for identifying a commodity category through a LOGO according to the present invention;

[0017] Figure 2 A flowchart of another embodiment of the method for identifying the category of a commodity through a LOGO of the present invention;

[0018] Figure 3 It is a flow chart of an embodiment of scene detection model training and scene detection in the method for identifying commodity categories through LOGO of the present invention;

[0019] Figure 4 It is a flow chart of an embodiment of POSM detection model training and POSM detection in the method of identifying commodity categories through LOGO of the present invention;

[0020] Figure 5 It is a flow chart of an embodiment of LOGO detection model training and LOGO recognition in the method of identifying commodity categories through LOGO of the present invention;

[0021] Figure 6 This is a structural diagram of an embodiment of a smart terminal of the present invention;

[0022] Figure 7 FIG. 4 is a structural diagram of a storage device according to an embodiment of the present invention. DETAILED DESCRIPTION

[0023] The present invention is further described below in conjunction with the accompanying drawings and specific implementation methods. It should be noted that, under the premise of no conflict, the various embodiments or technical features described below can be arbitrarily combined to form a new embodiment.

[0024] See also Figure 1-5 , Figure 1 It is a flow chart of an embodiment of a method for identifying a commodity category through a LOGO according to the present invention; Figure 2 A flowchart of another embodiment of the method for identifying the category of a commodity through a LOGO of the present invention; Figure 3 It is a flow chart of an embodiment of scene detection model training and scene detection in the method for identifying commodity categories through LOGO of the present invention; Figure 4 It is a flow chart of an embodiment of POSM detection model training and POSM detection in the method of identifying commodity categories through LOGO of the present invention; Figure 5 This is a flow chart of an embodiment of LOGO detection model training and LOGO recognition in the method of identifying commodity categories through LOGO of the present invention. Figure 1-5 The method for identifying the category of a commodity through a LOGO of the present invention is described in detail.

[0025] The present invention solves the problems that different subcategories of the same brand of goods are difficult to distinguish, the modeling cycle is long, the modeling range is difficult to fully cover, the background is messy and complex and easy to be misidentified, and the scene is complex and easy to be interfered by POSM. The method based on scene detection and LOGO (logotype, trademark) recognition is proposed to solve the above problems. The scheme is mainly divided into the following stages: the first stage is scene detection; the second stage is background removal; the third stage is POSM detection; the fourth stage is LOGO detection and recognition; the fifth stage is to remove the LOGO in POSM; the sixth stage is to output the category information of the goods to be identified.

[0026] Specifically, the methods for identifying product categories through LOGO include:

[0027] S101: Detecting a product display area through a scene detection model, and obtaining a display area of ​​a product to be detected according to the product display area, where the product display area is a product display area of ​​products of different brands.

[0028] In this embodiment, the commodities for identification are placed in a secondary display manner on shelves, counters and other areas, wherein commodities of the same brand are placed together.

[0029] In this embodiment, the step of detecting the commodity display area through the scene detection model specifically includes: obtaining a display image of the commodity, preprocessing the display image, and inputting the preprocessed display image into the scene detection model to obtain the location information and category label information of the commodity display area.

[0030] In this embodiment, the category label information includes display area categories such as pile head, end rack, Australian rack, etc.

[0031] In this embodiment, the preprocessing step includes image scaling, image expansion, image normalization and other preprocessing methods. The preprocessing method obtains an image with a consistent size that can be processed by the scene detection mode.

[0032] In a specific embodiment, the display image is a photographed shelf image of a commodity, and the photographed shelf image is preprocessed, mainly including image scaling, image expansion, image normalization, etc. The preprocessed shelf image is sent to a trained scene detection model to detect the display areas and category labels of commodities of different brands.

[0033] In this embodiment, the steps of training the scene detection model include: image data collection (collecting images including display images of different brands of goods, such as large shelf images), preprocessing the image data (normalization, data enhancement, etc.), inputting the preprocessed images into the pre-trained detection network, and training the model. During model training, the model is tuned until the model converges, and the optimal model after convergence is used as the model detection model and exported, and the display images of user product recognition are input into the scene detection model to obtain the product display areas (scene frame position information) of different brands of goods, that is, category information.

[0034] In this embodiment, the step of obtaining the display area of ​​the commodity to be detected according to the commodity display area specifically includes: obtaining the brand of the commodity to be detected, and obtaining the display area of ​​the commodity to be detected according to the brand. Among them, the scene detection model marks the commodity display areas of different brands of commodities in the form of scene detection frames, obtains the commodity display area of ​​the brand of commodities through the scene detection frames, and obtains the position information of the display area of ​​the commodity to be detected according to the coordinate information of the four corners of the scene detection frame.

[0035] S102: Train the POSM detection model and the LOGO recognition model respectively, and obtain the position of the POSM, the position and the category of the LOGO in the display area through the POSM detection model and the LOGO recognition model.

[0036] In this embodiment, the steps of respectively training the POSM detection model and the LOGO recognition model specifically include: respectively collecting a first picture including the POSM and a second picture including the LOGO, preprocessing the first picture and the second picture, and inputting the preprocessed first picture into the POSM detection network for training, and inputting the preprocessed second picture into the LOGO detection network for training; respectively adjusting and optimizing the parameters of the trained POSM detection model and the LOGO recognition model until the models converge. The LOGO of the second picture includes the LOGOs of different sub-brands in the same brand of goods.

[0037] In this embodiment, the preprocessing method is the same as the preprocessing method in the above scene detection model detection of the commodity display area, and both include at least one of image scaling, image expansion, and image normalization.

[0038] In this embodiment, the detection networks used by the scene detection model, the POSM detection model, and the LOGO recognition model include at least one of Cascade RCNN, Faster RCNN, and YOLO algorithms.

[0039] In this embodiment, the step of obtaining the location information of POSM through the POSM detection model includes: obtaining the display area of ​​the commodity to be detected, and performing image preprocessing on the area, mainly including image scaling, image expansion, image normalization, etc.; sending the preprocessed image to the trained POSM detection model to detect the POSM location information; and outputting the location information of the POSM in the display area.

[0040] In this embodiment, the step of obtaining the position and category of the LOGO in the display area through the LOGO recognition model specifically includes: preprocessing the image of the display area, performing recognition processing of the LOGO position and category on the preprocessed image through the trained LOGO recognition model, and obtaining the position and category of the LOGO based on the result of the recognition processing.

[0041] S103: Remove the LOGO in the POSM according to the position of the POSM and the position of the LOGO, and obtain the category corresponding to the LOGO that has not been removed in the display area.

[0042] In this embodiment, the step of removing the LOGO in the POSM according to the position of the POSM and the position of the LOGO specifically includes: acquiring the LOGO overlapping with the position of the POSM according to the position of the POSM and the position of the LOGO, and removing the LOGO.

[0043] After taking out the LOGO in the POSM, the commodity category corresponding to the remaining LOGO in the display area is used as the category of the commodity in the display area (the brand and sub-brand information to which the commodity belongs).

[0044] Compared with the existing technical solutions, the technical solution of the present invention has quite obvious advantages, which are mainly reflected in:

[0045] 1) It is not necessary to model all new products to achieve full coverage of product recognition, saving time, effort and cost. In the recognition process, this solution does not need to classify and identify all product packaging. After all, mining requires time and manpower costs. It only needs to identify the LOGO in the product to achieve the recognition of the product and its different series and different packaging. In addition, all models in this solution can be reused, such as scene detection model, POSM detection model, and LOGO detection model. There is no need to iterate the labeled data multiple times, saving manpower costs.

[0046] 2) High recognition accuracy. This solution effectively filters out irrelevant background and interference background through scene detection model and POSM detection model, and effectively solves the misidentification of highly similar non-original products in the background in the existing solution (for the target class to be identified, non-original products, especially non-original SKUs with extremely similar appearance, belong to interference background. Scene detection can effectively remove this background interference) and the interference of LOGO in POSM;

[0047] 3) Multiple-scenario reuse and strong versatility. The implementation scheme of the present invention is not limited or affected by the complexity of the scene, can be adapted to multiple different scenes, and can achieve a high recognition accuracy.

[0048] Based on the same inventive concept, the present invention also proposes an intelligent terminal, see Figure 6 , Figure 6 This is a structural diagram of an embodiment of the intelligent terminal of the present invention, combined with Figure 6 The intelligent terminal of the present invention is described.

[0049] In this embodiment, the intelligent terminal includes: a processor and a memory. The processor is in communication with the memory. The memory stores a computer program. The processor executes the method of identifying the category of a product through a LOGO as described in the above embodiment through the computer program.

[0050] Beneficial effect: The intelligent terminal of the present invention obtains the display area of ​​the same brand of goods through scene detection model recognition, and obtains the position of the POSM and the position of the LOGO in the display area through the trained POSM detection model and the LOGO recognition model, and then removes the LOGO in the POSM according to the position, so as to identify the category of the goods according to the LOGO that has not been removed. It can distinguish different categories of goods in the same brand through LOGO recognition, avoids the problem of needing to identify the product packaging, reduces the difficulty of modeling and labor costs, and effectively removes background interference through scene detection, reduces misidentification, and is not easily interfered by auxiliary sales materials, thereby improving recognition accuracy.

[0051] Based on the same inventive concept, the present invention also proposes a storage device, please refer to Figure 7 , Figure 7 FIG. 1 is a structural diagram of an embodiment of a storage device of the present invention. Figure 7 The storage device of the present invention will be described.

[0052] In this embodiment, the storage device stores program data, and the program data is used to execute the method for identifying the category of a commodity through a LOGO as described in the above embodiment.

[0053] In the embodiments provided by the present invention, it should be understood that the disclosed devices / terminals and methods can be implemented in other ways. For example, the device / terminal embodiments described above are only schematic. For example, the division of the modules or units is only a logical function division. There may be other division methods in actual implementation, such as multiple units or components can be combined or integrated into another system, or some features can be ignored or not executed. Another point is that the mutual coupling or direct coupling or communication connection shown or discussed can be through some interfaces, indirect coupling or communication connection of devices or units, which can be electrical, mechanical or other forms.

[0054] The units described as separate components may or may not be physically separated, and the components shown as units may or may not be physical units, that is, they may be located in one place or distributed on multiple network units. Some or all of the units may be selected according to actual needs to achieve the purpose of the solution of this embodiment.

[0055] If the integrated module is implemented in the form of a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the present invention implements all or part of the processes in the above-mentioned embodiment method, and can also be completed by instructing the relevant hardware through a computer program. The program data can be stored in a computer-readable storage medium, and the computer program can implement the steps of the above-mentioned various method embodiments when it is executed. Among them, the program data includes computer program code, and the computer program code can be in source code form, object code form, executable file or some intermediate form. The computer-readable medium may include: any entity or device capable of carrying the computer program code, recording medium, U disk, mobile hard disk, disk, optical disk, computer memory, read-only memory (ROM, Read-Only Memory), random access memory (RAM, Random Access Memory), electric carrier signal, telecommunication signal and software distribution medium. It should be noted that the content contained in the computer-readable medium can be appropriately increased or decreased according to the requirements of legislation and patent practice in the jurisdiction. For example, in some jurisdictions, according to legislation and patent practice, computer-readable media do not include electric carrier signals and telecommunication signals.

[0056] The various embodiments in this specification are described in a progressive manner, and each embodiment focuses on the differences from other embodiments. The same or similar parts between the various embodiments can be referenced to each other.

[0057] The above description of the disclosed embodiments enables those skilled in the art to implement or use the present invention. Various modifications to these embodiments will be apparent to those skilled in the art, and the general principles defined herein may be implemented in other embodiments without departing from the spirit or scope of the present invention. Therefore, the present invention will not be limited to the embodiments shown herein, but rather to the widest scope consistent with the principles and novel features disclosed herein.

Claims

1. A method for identifying commodity categories through LOGO, characterized in that: The method for identifying the category of a product through a LOGO includes: S101: Detecting a commodity display area through a scene detection model, and obtaining a display area of ​​a commodity to be detected according to the commodity display area, wherein the commodity display area is a commodity display area of ​​commodities of different brands; S102: training a POSM detection model and a LOGO recognition model respectively, and obtaining the position of the POSM, the position and the category of the LOGO in the display area through the POSM detection model and the LOGO recognition model; S103: removing the LOGO in the POSM according to the position of the POSM and the position of the LOGO, and obtaining the category corresponding to the LOGO that has not been removed in the display area; The step of removing the LOGO in the POSM according to the position of the POSM and the position of the LOGO specifically includes: According to the position of the POSM and the position of the LOGO, the LOGO overlapping with the position of the POSM is obtained, and the LOGO is removed.

2. The method for identifying the category of a commodity by using a LOGO as claimed in claim 1, characterized in that: The step of detecting the commodity display area by using the scene detection model specifically includes: Acquire a display image of a commodity, preprocess the display image, and input the preprocessed display image into the scene detection model to acquire location information and category label information of the commodity display area.

3. The method for identifying the category of a commodity by using a LOGO as claimed in claim 1, characterized in that: The step of obtaining the display area of ​​the commodity to be detected according to the commodity display area specifically includes: The brand of the commodity to be detected is obtained, and the display area is obtained according to the brand.

4. The method for identifying the category of a commodity by using a LOGO as claimed in claim 1, characterized in that: The steps of respectively training the POSM detection model and the LOGO recognition model specifically include: Collect a first picture including a POSM and a second picture including a LOGO respectively, pre-process the first picture and the second picture, input the pre-processed first picture into a POSM detection network for training, and input the pre-processed second picture into a LOGO detection network for training; The parameters of the trained POSM detection model and LOGO recognition model are adjusted and optimized respectively until the models converge.

5. The method for identifying the category of a commodity through a LOGO according to claim 4, characterized in that: The preprocessing includes at least one of image scaling, image expansion, and image normalization.

6. The method for identifying the category of a commodity by using a LOGO as claimed in claim 4, characterized in that: The detection networks used by the scene detection model, the POSM detection model, and the LOGO recognition model include at least one of Cascade RCNN, Faster RCNN, and YOLO.

7. The method for identifying the category of a commodity by using a LOGO as claimed in claim 1, characterized in that: The step of obtaining the position and category of the LOGO in the display area through the LOGO recognition model specifically includes: The image of the display area is preprocessed, and the LOGO position and category are recognized on the preprocessed image by a trained LOGO recognition model, and the position and category of the LOGO are obtained according to the result of the recognition process.

8. An intelligent terminal, characterized in that: The intelligent terminal includes a processor and a memory, wherein the memory stores a computer program, and the computer program is used to execute the method for identifying a commodity category by using a LOGO as described in any one of claims 1 to 7.

9. A storage device, characterized in that: The storage device stores program data, and the program data is used to execute the method for identifying the category of a commodity through a LOGO as described in any one of claims 1 to 7.

Images