Cigarette authenticity identification method and system based on multi-view image acquisition and deep learning
By combining multi-view image acquisition with deep learning, along with multi-classification convolutional neural networks and object detection algorithms, the low accuracy and deployment challenges of cigarette authenticity identification in complex environments were solved, achieving efficient and accurate on-site identification.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2025-09-04
- Publication Date
- 2026-03-13
AI Technical Summary
Existing technologies have low accuracy in identifying genuine and counterfeit cigarettes with high similarity and subtle differences under complex shooting environments, poor generalization ability, and difficulty in rapid on-site deployment.
A multi-view image acquisition system combined with deep learning was adopted. A cascaded deep learning model was constructed through multi-classification convolutional neural networks and object detection algorithms. A hybrid loss function combining classification loss and contrastive loss was used to identify cigarette categories, detect key regions, and determine the authenticity of local cigarettes.
It achieves high-precision and robust authenticity identification in complex environments, can be quickly deployed on portable mobile terminals, meets the real-time identification needs of market supervision sites, and significantly improves identification efficiency and accuracy.
Smart Images
Figure CN121661475A_ABST
Abstract
Description
Technical Field
[0001] This invention relates to the field of deep learning technology, specifically to a method and system for identifying genuine and counterfeit cigarettes based on multi-view image acquisition and deep learning. Background Technology
[0002] Currently, the identification of genuine and counterfeit cigarettes mainly relies on the subjective experience of inspectors or the assistance of professional optical equipment. Manual identification is significantly affected by individual differences in experience, lacks unified and objective judgment standards, and is inefficient, making it difficult to meet the needs of large-scale market supervision. Professional equipment identification, on the other hand, requires specific instruments and operating skills, and is not readily deployable or widely applicable on-site.
[0003] With the development of artificial intelligence technology in the field of image recognition, some image processing-based authentication schemes have emerged. These methods typically involve capturing images of cigarette packaging using a mobile terminal and uploading them to a cloud server for identification. However, in practical applications, the complex and variable shooting environment presents challenges such as inconsistent lighting conditions, packaging angle shifts, film reflection interference, localized deformation, and differences in image scale, all of which severely affect the authentication results.
[0004] Furthermore, the visual differences between genuine and counterfeit cigarettes often lie in extremely subtle local features. These features occupy small areas and have complex shapes, and traditional image algorithms have significant limitations in feature extraction and discrimination capabilities. Existing deep learning-based identification methods mostly use a single model to perform end-to-end classification, which has limited ability to perceive subtle differences and does not fully consider the characteristic differences between different categories and different identification regions, resulting in the need to improve the model's generalization ability and accuracy.
[0005] Although existing technologies have attempted to combine traditional visual features (such as color, texture, and shape) with deep learning algorithms, there is still room for further optimization and innovation in terms of feature fusion methods, regional positioning accuracy, environmental interference resistance, and ease of practical deployment.
[0006] Therefore, there is an urgent need for a more accurate, robust, and applicable intelligent identification technology solution suitable for on-site law enforcement environments. Summary of the Invention
[0007] To address this, embodiments of the present invention provide a method and system for identifying genuine and counterfeit cigarettes based on multi-view image acquisition and deep learning, in order to solve the technical problems of low accuracy, poor generalization ability, and difficulty in rapid on-site deployment of existing technologies in complex shooting environments for identifying identification points with high similarity and subtle differences.
[0008] To achieve the above objectives, the embodiments of the present invention provide the following technical solutions:
[0009] According to a first aspect of the present invention, a method for identifying genuine and counterfeit cigarettes based on multi-view image acquisition and deep learning is provided, the method comprising:
[0010] S1. Acquire multiple surface images of the cigarette packaging to be identified using a multi-view image acquisition system;
[0011] S2. Input the multiple surface images into a pre-trained category recognition model to identify the specific category to which the cigarette belongs;
[0012] S3. Based on the identified category, call the corresponding region detection model to locate at least one key identification region from the surface image;
[0013] S4. Input the image of each key identification region into the corresponding region recognition model to obtain the local authenticity judgment result of each region. The region recognition model is a binary classification model trained for its category and specific region.
[0014] S5. Based on the authenticity determination results of all key identification areas, generate and output the overall authenticity identification conclusion of the cigarette.
[0015] Furthermore, multiple surface images of the cigarette packaging to be identified are acquired using a multi-view image acquisition system, including:
[0016] The multi-view image acquisition system includes multiple industrial cameras and a matching lighting unit. The multiple industrial cameras are arranged to simultaneously acquire images of the front, back, top, bottom, and two sides of a cigarette pack.
[0017] Furthermore, the category recognition model is a multi-class convolutional neural network model used to identify the brand, specifications, and style of cigarettes.
[0018] Furthermore, the category identification model employs a deep learning network based on the MobileNet architecture.
[0019] Furthermore, the region detection model is a model trained based on the YOLO series of object detection algorithms, used to accurately select the regions on the packaging that are printed with anti-counterfeiting features.
[0020] Furthermore, the region identification model is a deep learning network based on the ResNet architecture, and it is optimized for the specific region it is responsible for.
[0021] Furthermore, the region identification model is a binary classification model trained for its category and a specific region, including:
[0022] Data augmentation techniques were used to expand the training sample set when training the region recognition model. These data augmentation techniques included random flipping, rotation, and color adjustment.
[0023] Furthermore, the region identification model is a binary classification model trained for its category and a specific region, and also includes:
[0024] The loss function used to train the region recognition model consists of a combination of classification loss and contrastive loss. The contrastive loss is used to increase the distance between the features of genuine and counterfeit images, while reducing the distance between the features of genuine images. Specifically, it includes:
[0025] The loss function expression of the region identification model is:
[0026]
[0027] The expression for the classification loss function is:
[0028]
[0029] Where n is the number of image samples, Let i be the true label of the i-th image. Let be the predicted probability of the i-th image;
[0030] The expression for the contrastive loss function is:
[0031]
[0032]
[0033] in, Let represent the feature vector of the i-th sample.
[0034] Furthermore, the region identification model is a binary classification model trained for its category and a specific region, and also includes:
[0035] The genuine and fake samples are defined as positive and negative samples respectively. The contents of the loss function are constructed into a matrix multiplication form to reduce the feature differences between genuine products of the same type.
[0036] According to a second aspect of the present invention, a cigarette authenticity identification system based on multi-view image acquisition and deep learning is provided, the system comprising:
[0037] The surface image acquisition module is used to acquire multiple surface images of the cigarette packaging to be identified through a multi-view image acquisition system;
[0038] The category recognition module is used to input the multiple surface images into a pre-trained category recognition model to identify the specific category to which the cigarette belongs;
[0039] The key identification region localization module is used to call the corresponding region detection model based on the identified category to locate at least one key identification region from the surface image;
[0040] The local authenticity determination module is used to input the image of each key identification region into the corresponding region recognition model to obtain the local authenticity determination result of each region. The region recognition model is a binary classification model trained for its category and specific region.
[0041] The authenticity determination result output module is used to integrate the authenticity determination results of all key identification areas, generate and output the overall authenticity determination conclusion of the cigarette.
[0042] The embodiments of the present invention have the following advantages:
[0043] This invention employs a multi-camera synchronous acquisition system to obtain multi-angle images of cigarette packaging. A cascaded deep learning model architecture is used to sequentially perform category identification, key region detection, and region authenticity determination. A hybrid loss function combining classification and contrastive losses is introduced to enhance the model's ability to distinguish subtle feature differences. This invention effectively overcomes interference from ambient light, angle changes, and packaging deformation, achieving high-precision and robust authenticity verification. It can be deployed on portable mobile terminals, meeting the needs of rapid, real-time identification in market supervision, and significantly improving identification efficiency and accuracy. Attached Figure Description
[0044] To more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are merely exemplary, and those skilled in the art can derive other embodiments based on the provided drawings without creative effort.
[0045] The structures, proportions, sizes, etc. illustrated in this specification are only for the purpose of assisting those skilled in the art in understanding and reading the content disclosed herein, and are not intended to limit the conditions under which the present invention can be implemented. Therefore, they have no substantial technical significance. Any modifications to the structure, changes in the proportions, or adjustments to the size, without affecting the effects and objectives that the present invention can produce, should still fall within the scope of the technical content disclosed in the present invention.
[0046] Figure 1A schematic diagram of the logical structure of a cigarette authenticity identification system based on multi-view image acquisition and deep learning provided in an embodiment of the present invention;
[0047] Figure 2 This is a flowchart illustrating a method for identifying genuine and counterfeit cigarettes based on multi-view image acquisition and deep learning, provided in an embodiment of the present invention.
[0048] Figure 3 A schematic diagram of the contrast loss function in a cigarette authenticity identification method based on multi-view image acquisition and deep learning provided in an embodiment of the present invention;
[0049] Figure 4 This is a schematic diagram illustrating the discrimination of a specific category in a cigarette authenticity identification method based on multi-view image acquisition and deep learning provided in an embodiment of the present invention.
[0050] Figure 5 This is a schematic diagram illustrating the category identification in a cigarette authenticity identification method based on multi-view image acquisition and deep learning, provided in an embodiment of the present invention.
[0051] Figure 6 A schematic diagram of positive and negative sample pairs in a cigarette authenticity identification method based on multi-view image acquisition and deep learning provided in an embodiment of the present invention;
[0052] Figure 7 This is a schematic diagram illustrating the usage process of a cigarette authenticity identification method based on multi-view image acquisition and deep learning, provided in an embodiment of the present invention. Detailed Implementation
[0053] The following specific embodiments illustrate the implementation of the present invention. Those skilled in the art can easily understand other advantages and effects of the present invention from the content disclosed in this specification. Obviously, the described embodiments are only some, not all, of the embodiments of the present invention. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.
[0054] Currently, the main method for identifying genuine and counterfeit cigarettes relies on manual inspection. Professionals need extensive experience to judge authenticity based on factors such as the cigarette's appearance and packaging printing quality. However, this method has several limitations. First, manual inspection is inefficient, as a single inspection may require repeated observations and is time-consuming, making it unsuitable for large-scale market inspections. Second, manual inspection is highly subjective; differences in experience and judgment standards among different inspectors can easily lead to misjudgments. Furthermore, with the continuous upgrading of counterfeiting technology, counterfeit cigarettes are becoming increasingly realistic in appearance, making manual identification increasingly difficult and hindering the effective curbing of counterfeit cigarette circulation.
[0055] Currently, the identification of genuine and counterfeit cigarettes mainly relies on manual methods, which are inefficient and difficult. There are no readily available AI algorithm models or technologies for reference in the industry. To implement this technology, a highly efficient and accurate intelligent identification system needs to be built from scratch, encompassing data collection, data processing, algorithm development, model training, model validation, and the commercialization of scientific and technological achievements. Traditional image processing methods (such as contour extraction, edge detection, and corner detection) are effective for simple tasks but struggle to effectively identify genuine and counterfeit cigarettes because their appearances are extremely similar and their patterns are complex, making it difficult for traditional image processing methods to handle.
[0056] To address the aforementioned technical challenges of low accuracy, poor generalization ability, and difficulty in rapid on-site deployment when identifying highly similar images with subtle differences in complex shooting environments, this invention directly uses deep learning algorithms to extract image features and constructs a loss function specifically designed to handle images with high similarity but only subtle differences. This significantly improves the model's ability to distinguish between complex and similar images of genuine and counterfeit cigarettes, with recognition performance and accuracy far exceeding traditional image processing and traditional neural network algorithm solutions. Furthermore, this model system can be deployed on portable mobile terminals, enabling a closed loop for the commercialization of scientific research results.
[0057] refer to Figure 1 This invention discloses a cigarette authenticity identification system based on multi-view image acquisition and deep learning. The system includes: a surface image acquisition module 1; a product category identification module 2; a key identification area positioning module 3; a local authenticity determination module 4; and an authenticity determination result output module 5.
[0058] This invention constructs a high-definition multi-view camera system and a matching lighting system to acquire images of single packs and cartons of cigarettes from multiple perspectives. The acquired image data is labeled and preprocessed, and then augmented. Image segmentation techniques are used to process key areas of some product images, such as hidden and subtle features. A rich sample dataset corresponding to currently circulating and counterfeit cigarettes is built from scratch. A deep learning algorithm system is then constructed and trained. The trained models are cascaded to form a comprehensive expert model system, completing a fully automated intelligent identification system from input to output. Embodiments of this invention can be directly deployed and integrated into portable mobile terminal devices for market enforcement personnel, allowing them to easily carry the device for enforcement inspections and provide identification results in extremely short times, such as seconds.
[0059] The present invention's embodiments for identifying genuine and counterfeit cigarettes overcome the limitations of traditional methods, improving identification efficiency and accuracy while reducing labor intensity. It provides intelligent support for cigarette market authentication and offers strong technical support to tobacco market regulators in combating the illegal production and sale of counterfeit cigarettes and protecting consumer rights. It also serves as a deterrent to counterfeiters and sellers, effectively preventing the loss of national tax revenue and economic losses.
[0060] Corresponding to the aforementioned cigarette authenticity identification system based on multi-view image acquisition and deep learning, this invention also discloses a cigarette authenticity identification method based on multi-view image acquisition and deep learning. The following details a cigarette authenticity identification method based on multi-view image acquisition and deep learning disclosed in this invention, in conjunction with the aforementioned cigarette authenticity identification system based on multi-view image acquisition and deep learning.
[0061] refer to Figure 2 and Figure 3 This invention discloses a method for identifying genuine and counterfeit cigarettes based on multi-view image acquisition and deep learning. The method includes: S1, acquiring multiple surface images of the cigarette packaging to be identified through a multi-view image acquisition system; S2, inputting the multiple surface images into a pre-trained category recognition model to identify the specific category to which the cigarette belongs; S3, based on the identified category, calling the corresponding region detection model to locate at least one key identification region from the surface images; S4, inputting the image of each key identification region into the corresponding region recognition model to obtain the local authenticity determination result of each region, wherein the region recognition model is a binary classification model trained for its category and specific region; S5, combining the authenticity determination results of all key identification regions to generate and output the overall authenticity identification conclusion of the cigarette.
[0062] Step 1: Build a multi-view camera system and lighting system. Since each pack or carton of cigarettes has six sides, if a single camera detects one side at a time, it would require a maximum of six detections. Therefore, a multi-camera system is built to detect multiple sides simultaneously, reducing the number of operations. Simultaneously, to better capture the patterns, textures, and other appearance information on the cigarette packaging, the camera is equipped with a lighting system. This system makes the captured images of the cigarette packaging clearer and of higher quality, and high-quality image data is the foundation of the algorithm.
[0063] Step 2: Collect image data of genuine and counterfeit cigarettes for various categories. Input genuine and counterfeit cigarettes (including single packs and cartons) from different categories into the system built in the previous step, and collect a large amount of image data of the outer packaging of genuine and counterfeit cigarettes through operation. During the collection process, it is necessary to collect as much sample data as possible to provide a solid foundation for subsequent model training.
[0064] Step 3: Process and label the acquired images: Remove blurry images and label the data using labeling tools.
[0065] Step 31: Label the data used for the category identification model: Divide the image data of genuine and counterfeit cigarettes of all categories into categories, where categories refer to subcategories. For example, a certain brand of cigarettes may have multiple different specifications (such as hard pack, soft pack, thick pack, thin pack, etc.). Cigarettes of the same series and with the same packaging appearance belong to one category.
[0066] Step 32: Label the data used for the region detection model: Label the region detection data for each of the multiple categories built in the previous step. For example, label the regions in the soft packaging of a certain brand of cigarettes that can be used to determine authenticity. Each region belongs to a category.
[0067] Figure 4 The image shows an expert system for a specific product category. The key area localization shown in the image is achieved through the category's area detection model, while the feature recognition and judgment of that area are achieved through the final area recognition model.
[0068] Step 33: Label the data for the region recognition model: Label the individual regions labeled in the previous step with true / false labels. Each region's data is individually labeled with a binary classification label, i.e., a true / false label. Convert the labeled data into the format required by each level of the model, and divide the data into training set, validation set, and test set.
[0069] Step 4: Building and Training the Category Recognition Model: For multi-class classification tasks, the MobileNet network model was chosen to determine the categories. The model was trained using a complete dataset, hyperparameters were adjusted, and cross-validation was performed to optimize performance. Performance metrics such as precision and recall were evaluated.
[0070] Step 5: Construction and training of region detection models for each product category: A separate region detection model is constructed for each product category. Here, the YOLOv11 model is selected as the most advanced and stable model with the best overall performance. Due to its high efficiency and accuracy, YOLOv11 can effectively extract the target regions for each product category.
[0071] Step 6: Construction and Training of Region Recognition Models: A region recognition model is constructed for each region of each product category. The ResNet18 model is selected here as it offers the best overall performance in terms of both advancement and stability. The ResNet18 model provides the optimal solution for edge computing and mobile model deployment. The model structure is modified as follows: Figure 3As shown. To avoid insufficient image data samples (especially due to a lack of counterfeit cigarette samples leading to inadequate image data), data augmentation (such as image flipping and cropping) is used to expand the data samples, increase the diversity of image data, enable the model to learn more robust feature representations, and improve the model's generalization ability. Since the outer packaging of genuine and counterfeit cigarettes is extremely similar, making it difficult to distinguish them intuitively, an additional contrastive loss function is introduced to widen the feature differences between genuine and counterfeit image data and narrow the feature differences between genuine cigarette packaging image data.
[0072] The loss function expression for the region recognition model is:
[0073]
[0074] The expression for the classification loss function is:
[0075]
[0076] Where n is the number of image samples, Let i be the true label of the i-th image. Let be the predicted probability of the i-th image;
[0077] The expression for the contrastive loss function is:
[0078]
[0079]
[0080] in, Let represent the feature vector of the i-th sample.
[0081] True and false samples are defined as positive and negative samples, respectively. In the formula, P is the set of positive samples, and i,j correspond to positive sample pairs formed between two different positive samples in the set. M is the number of positive sample pairs, and Q is the set of negative samples. Let represent the feature vector of the i-th sample. To improve efficiency during training, the loss function is constructed as a matrix multiplication. Assuming there are 3 positive samples and 3 negative samples, their relationship is shown in the graph below. Figure 6 As shown, a positive sample and all other positive samples constitute a positive sample pair, and each negative sample in the negative sample set constitutes a negative sample pair.
[0082]
[0083] Among them, matrix It is a matrix composed of feature vectors obtained from positive samples through feature extraction branches.
[0084]
[0085] Here, matrix B is a matrix composed of feature vectors obtained from the feature extraction branch of the negative samples.
[0086]
[0087] Here, matrix C is composed of positive sample pairs formed between positive samples. Corresponding to the previous formula .
[0088]
[0089] matrix It consists of negative sample pairs formed between positive and negative samples. Corresponding to the previous formula .
[0090] Step 7: Export models for each level and region: After completing the above series of operations, many models for specific tasks at different levels are generated. Export these models to prepare for building an expert system.
[0091] Step 8: Construct a multi-model expert system: (Refer to...) Figure 5 The multiple models obtained above are cascaded to form a complete expert system from input to output. First, the cigarette is placed in a multi-camera system, which acquires the input image. After the input image is processed by the category recognition model, the system outputs the category to which the cigarette belongs. Then, based on the output, the system enters the corresponding category's expert subsystem. The subsystem first extracts the target region to be identified using a region extraction model, and then sends the target region to the corresponding region recognition model to obtain the true or false identification result for each region, thereby comprehensively judging the authenticity of the cigarette.
[0092] This invention constructs a camera and lighting system to prevent interference from ambient light on cigarette packaging images, providing stable input for subsequent models. By designing a dedicated region recognition algorithm, the accuracy of identifying genuine and counterfeit cigarettes is significantly improved. Furthermore, embodiments of this invention can be integrated and deployed in portable mobile terminal devices. (See reference...) Figure 7 This enables the commercialization of the entire equipment set. In the scenario of identifying genuine and counterfeit cigarettes, relying on the intelligent identification of genuine and counterfeit cigarettes based on the embodiments of this invention can greatly improve efficiency, greatly improve accuracy, reduce labor intensity, and provide real-time intelligent support for cigarette market law enforcement personnel to identify genuine and counterfeit cigarettes.
[0093] Although the present invention has been described in detail above with general descriptions and specific embodiments, modifications or improvements can be made to it, which will be obvious to those skilled in the art. Therefore, all such modifications or improvements made without departing from the spirit of the present invention fall within the scope of protection claimed by the present invention.
Claims
1. A method for identifying genuine and counterfeit cigarettes based on multi-view image acquisition and deep learning, characterized in that, The method includes: S1. Acquire multiple surface images of the cigarette packaging to be identified using a multi-view image acquisition system; S2. Input the multiple surface images into a pre-trained category recognition model to identify the specific category to which the cigarette belongs; S3. Based on the identified category, call the corresponding region detection model to locate at least one key identification region from the surface image; S4. Input the image of each key identification region into the corresponding region recognition model to obtain the local authenticity judgment result of each region. The region recognition model is a binary classification model trained for its category and specific region. S5. Based on the authenticity determination results of all key identification areas, generate and output the overall authenticity identification conclusion of the cigarette.
2. The method for identifying genuine and counterfeit cigarettes based on multi-view image acquisition and deep learning as described in claim 1, characterized in that, Multiple surface images of the cigarette packaging to be identified are acquired using a multi-view image acquisition system, including: The multi-view image acquisition system includes multiple industrial cameras and a matching lighting unit. The multiple industrial cameras are arranged to simultaneously acquire images of the front, back, top, bottom, and two sides of a cigarette pack.
3. The method for identifying genuine and counterfeit cigarettes based on multi-view image acquisition and deep learning as described in claim 1, characterized in that, The category recognition model is a multi-class convolutional neural network model used to identify the brand, specifications, and style of cigarettes.
4. The method for identifying genuine and counterfeit cigarettes based on multi-view image acquisition and deep learning as described in claim 3, characterized in that, The category identification model uses a deep learning network based on the MobileNet architecture.
5. The method for identifying genuine and counterfeit cigarettes based on multi-view image acquisition and deep learning as described in claim 1, characterized in that, The region detection model is a model trained based on the YOLO series of object detection algorithms, used to accurately select the regions on the packaging that are printed with anti-counterfeiting features.
6. The method for identifying genuine and counterfeit cigarettes based on multi-view image acquisition and deep learning as described in claim 1, characterized in that, The region identification model is a deep learning network based on the ResNet architecture, and it is optimized for the specific region it is responsible for.
7. The method for identifying genuine and counterfeit cigarettes based on multi-view image acquisition and deep learning as described in claim 6, characterized in that, The region identification model is a binary classification model trained on its category and a specific region, including: Data augmentation techniques were used to expand the training sample set when training the region recognition model. These data augmentation techniques included random flipping, rotation, and color adjustment.
8. The method for identifying genuine and counterfeit cigarettes based on multi-view image acquisition and deep learning as described in claim 7, characterized in that, The region identification model is a binary classification model trained on its category and a specific region, and also includes: The loss function used to train the region recognition model consists of a combination of classification loss and contrastive loss. The contrastive loss is used to increase the distance between the features of genuine and counterfeit images, while reducing the distance between the features of genuine images. Specifically, it includes: The loss function expression of the region identification model is: L loss =L cls_loss +αL con_loss The expression for the classification loss function is: Where n is the number of image samples, For the true label of the i-th image, Let be the predicted probability of the i-th image; The expression for the contrastive loss function is: Where, f(x) i ) represents the feature vector of the i-th sample.
9. The method for identifying genuine and counterfeit cigarettes based on multi-view image acquisition and deep learning as described in claim 8, characterized in that, The region identification model is a binary classification model trained on its category and a specific region, and also includes: The genuine and fake samples are defined as positive and negative samples respectively. The contents of the loss function are constructed into a matrix multiplication form to reduce the feature differences between genuine products of the same type.
10. A cigarette authenticity identification system based on multi-view image acquisition and deep learning, characterized in that, The system includes: The surface image acquisition module is used to acquire multiple surface images of the cigarette packaging to be identified through a multi-view image acquisition system; The category recognition module is used to input the multiple surface images into a pre-trained category recognition model to identify the specific category to which the cigarette belongs; The key identification region localization module is used to call the corresponding region detection model based on the identified category to locate at least one key identification region from the surface image; The local authenticity determination module is used to input the image of each key identification region into the corresponding region recognition model to obtain the local authenticity determination result of each region. The region recognition model is a binary classification model trained for its category and specific region. The authenticity determination result output module is used to integrate the authenticity determination results of all key identification areas, generate and output the overall authenticity determination conclusion of the cigarette.