Cigarette combustion image data acquisition and panoramic splicing method and system

Abstract

The invention provides a cigarette combustion image data acquisition and panoramic splicing method and system. Belongs to the technical field of cigarette detection. A multi-angle camera system is used for collecting images of multiple visual angles in the cigarette combustion process, and the multi-angle images cover different stages and different visual angles of cigarette combustion; the images are preprocessed, and the influence of illumination changes on the definition of the images in the cigarette combustion process is reduced; the multi-angle camera system comprises a plurality of cameras, and spatial position and angle calibration is performed on each camera through a posture estimation algorithm, so that images acquired by each camera have spatial consistency in a splicing process; performing feature point extraction and matching on each image by using a multi-level feature point screening and weighting strategy; generating a depth map of each pixel point by using a depth estimation algorithm, and correcting the relative scale and displacement of each image according to the depth map; and performing panoramic stitching on the multi-view images, and performing multi-scale smoothing processing on the stitching edge through a seamless fusion technology based on a stitching path.

Description

Technical Field

[0001] The present invention relates to the technical field of cigarette detection, and in particular to a method and system for collecting cigarette burning image data and performing panoramic stitching. Background Art

[0002] With the growing development of the cigarette market, consumers are paying more and more attention to the quality of cigarettes, and the burning process of cigarettes is directly related to consumers' smoking experience. During the burning process, the stability of the cigarette ash column, crack characteristics, and char line integrity are all important indicators that affect the quality of cigarettes. These indicators not only reflect the smoking comfort of cigarettes, but also affect the combustion stability and safety. Traditional cigarette quality control methods mostly rely on manual observation and simple physical indicator measurements. However, the visual characteristics of cigarette burning are complex and changeable, and different stages of burning will show changes in color, brightness, and texture, which makes it difficult for traditional detection methods to quantify the burning quality of cigarettes in real time and accurately.

[0003] In the detection of cigarette combustion quality, accurate detection of surface characteristics such as ash column shrinkage, crack rate, and carbon line integrity is particularly important. The generation of ash column cracks, the width and uniformity of the burning carbon line, and other characteristics are related to many factors such as cigarette paper, tobacco filling density, and combustion calorific value. Therefore, real-time monitoring and high-precision analysis of these surface characteristics are challenging. Especially during the combustion process, the color difference and complex and diverse shapes of the crack area, the different sizes of the crack and the carbon line, and the change in perspective caused by the cylindrical structure of the cigarette increase the difficulty of image processing and segmentation.

[0004] Traditional cigarette burning quality detection methods often use static images or single-angle shooting, which makes it difficult to fully record the entire process of cigarette burning, especially for the extraction of dynamic change features such as ash column cracks.

[0005] In addition, the different stages of cigarette burning require real-time adjustment of image acquisition parameters to cope with the differences in brightness, color, and background produced by burning. Therefore, existing detection methods face the following technical challenges:

[0006] 1. Multi-view image acquisition: Cigarette burning is a dynamic process, and the ash columns and cracks produced gradually change during the burning stage. Single-view acquisition cannot fully cover the cigarette surface, resulting in incomplete image data and difficulty in meeting the needs of panoramic analysis of burning characteristics.

[0007] 2. Image preprocessing and light adaptation: During the burning process of cigarettes, the light changes significantly, which directly affects the clarity and detail of the image, especially the contrast between the bright and low-light areas during burning. Traditional image acquisition systems are not adaptable enough to light changes, and often require manual adjustment of exposure parameters, resulting in a complex and inefficient detection process.

[0008] 3. Image stitching accuracy: Cigarettes are cylindrical, so it is difficult to stitch their panoramic images. Stitching at different angles often leads to ghosting, breakage and other problems due to perspective deviation, especially in detail areas such as burning charcoal lines and cracks. Existing stitching algorithms fail to effectively solve the stitching errors of multiple perspectives, resulting in errors in the stitched images, affecting the accuracy of the data.

[0009] 4. Feature point matching and image calibration: The stitching of cigarette burning images requires high-precision feature point matching and spatial calibration. The subtle features of the cigarette surface show differences at different viewing angles, making it difficult for traditional matching algorithms to accurately align. In addition, the influence of the curvature and depth of the cigarette surface increases the complexity of the perspective transformation, so more accurate image depth calibration and multi-scale feature matching strategies are needed to improve the accuracy and stability of stitching.

[0010] In view of this, the inventor, based on many years of experience in production design in this field and related fields, has designed a method for collecting cigarette burning image data and panoramic stitching after repeated experiments, in order to solve the problems existing in the prior art. Summary of the invention

[0011] The object of the present invention is to provide a method and system for collecting cigarette burning image data and panoramic stitching, which can accurately align cigarette burning images at different viewing angles.

[0012] In order to achieve the above-mentioned purpose, the present invention proposes a method and system for collecting cigarette burning image data and panoramic stitching, wherein the method for collecting cigarette burning image data and panoramic stitching includes:

[0013] Use a multi-angle camera system to collect images from multiple perspectives during the cigarette burning process. The multi-angle images cover different stages and perspectives of cigarette burning.

[0014] Preprocessing each image and reducing the influence of illumination changes during cigarette burning on the clarity of each image;

[0015] The multi-angle camera system includes multiple cameras, and the spatial position and angle of each camera are calibrated through the posture estimation algorithm, so that the images collected by each camera have spatial consistency during the stitching process;

[0016] Use multi-level feature point screening and weighting strategy to extract and match feature points of each image;

[0017] A depth estimation algorithm is used to generate a depth map for each pixel, and the relative scale and displacement of each image are corrected according to the depth map;

[0018] The multi-view images are panoramically stitched, and the stitching edges are smoothed at multiple scales using seamless fusion technology based on the stitching path.

[0019] Compared with the prior art, the present invention has the following characteristics and advantages:

[0020] The cigarette burning image data acquisition and panoramic stitching method and system proposed in the present invention constructs a high-precision cigarette burning image data set, which not only meets the panoramic analysis and dynamic monitoring needs of the apparent burning characteristics of cigarettes, but also realizes real-time query, data sharing and subsequent data analysis through the support of the big data platform, providing strong technical support for quality control, defect detection and performance optimization of the cigarette industry. In addition, the technical solution also has good adaptability and efficiency, and can provide high-precision cigarette burning image data for the big data platform, helping the quality control and performance optimization of the cigarette industry. BRIEF DESCRIPTION OF THE DRAWINGS

[0021] The accompanying drawings described herein are only for explanation purposes and are not intended to limit the scope of the present invention in any way. In addition, the shapes and proportional dimensions of the various components in the figures are only schematic, used to help understand the present invention, and are not specifically limited to the shapes and proportional dimensions of the various components of the present invention. Those skilled in the art can select various possible shapes and proportional dimensions to implement the present invention according to the teachings of the present invention.

[0022] Figure 1 It is a flow chart of the cigarette burning image data collection and panoramic stitching method of the present invention;

[0023] Figure 2a The image of the burning charcoal line of the cigarette of the linear smoking device of the present invention;

[0024] Figure 2b It is an image of the burning charcoal line of a cigarette of the rotary disk type single-channel smoking device of the present invention;

[0025] Figure 2c It is an image of burning charcoal lines of cigarettes of the turntable type multi-channel smoking device of the present invention;

[0026] Figure 3a A panoramic image of a cigarette pack ash crack of the present invention;

[0027] Figure 3b A crack template image corresponding to the cigarette ash crack of the present invention;

[0028] Figure 4 Schematic diagram of sample types of the apparent burning quality data set of the cigarette product of the present invention. DETAILED DESCRIPTION

[0029] The details of the present invention can be more clearly understood with reference to the accompanying drawings and the description of the specific embodiments of the present invention. However, the specific embodiments of the present invention described herein are only used for the purpose of explaining the present invention and cannot be understood as limiting the present invention in any way. Under the guidance of the present invention, technicians can conceive of any possible variations based on the present invention, which should all be considered to belong to the scope of the present invention.

[0030] The present invention proposes a method and system for collecting cigarette burning image data and panoramic stitching, wherein the method for collecting cigarette burning image data and panoramic stitching includes:

[0031] Use a multi-angle camera system to collect images from multiple perspectives during the cigarette burning process. The multi-angle images cover different stages and perspectives of cigarette burning.

[0032] Preprocessing each image and reducing the influence of illumination changes during cigarette burning on the clarity of each image;

[0033] The multi-angle camera system includes multiple cameras, and the spatial position and angle of each camera are calibrated through the posture estimation algorithm, so that the images collected by each camera have spatial consistency during the stitching process;

[0034] Use multi-level feature point screening and weighting strategy to extract and match feature points of each image;

[0035] A depth estimation algorithm is used to generate a depth map for each pixel, and the relative scale and displacement of each image are corrected according to the depth map;

[0036] The multi-view images are panoramically stitched, and the stitching edges are smoothed at multiple scales using seamless fusion technology based on the stitching path.

[0037] The present invention proposes a method and system for collecting and panoramic stitching of cigarette burning image data, which provides a method for comprehensively and accurately capturing and analyzing the cigarette burning process, providing strong technical support for cigarette quality control, and also provides high-precision image data for the big data platform, which helps the quality control, defect detection and performance optimization of the cigarette industry.

[0038] The present invention provides a method for collecting and stitching cigarette burning images based on big data, which generates a panoramic image data set of the entire process of cigarette burning through multi-angle image collection, dynamic preprocessing, depth estimation calibration and stitching fusion algorithm. Figure 1 As shown, the specific steps include:

[0039] 1) Use a multi-angle camera system to capture multi-angle images of the cigarette burning process to ensure comprehensive coverage of different stages and angles of cigarette burning;

[0040] 2) Dynamically adjust the exposure, grayscale balance and color calibration of the collected multi-view images to adapt to the impact of illumination changes on image clarity during the combustion process;

[0041] 3) Calibrate the spatial position and angle of each camera through the attitude estimation algorithm. The attitude estimation algorithm is based on the calculation of the internal and external parameter matrices of the camera, and uses the rotation and translation matrices to correct the perspective differences of each camera in space to ensure the spatial consistency of images with different perspectives during the stitching process;

[0042] 4) Use multi-level feature point screening and weighting strategy to match image feature points. The strategy includes extracting image feature points through multi-scale feature detection algorithm, screening out low-confidence noise points and giving higher weights to key feature points to improve stitching accuracy;

[0043] 5) Using the image calibration step based on depth estimation, a depth map of each pixel is generated using a depth estimation algorithm before stitching, and the relative scale and displacement of the image are corrected according to the depth information to improve the alignment accuracy of the stitched images under different viewing angles;

[0044] 6) Perform panoramic stitching on multi-view images, and perform multi-scale smoothing on the stitching edges through seamless fusion technology based on the stitching path to reduce edge breakage or ghosting caused by angle differences and achieve high-resolution seamless stitching.

[0045] In an optional embodiment of the present invention, the multi-angle camera system includes at least three cameras, which are respectively located at the front view position, the side view position and the top view position of the cigarette burning area. By adopting the multi-angle camera system, the characteristics of cigarette burning can be captured from multiple perspectives, and the comprehensive monitoring of the cigarette burning process is realized, which not only enhances the panoramic analysis capability of cigarette burning characteristics, but also provides rich visual information for dynamic monitoring, and significantly improves the integrity and accuracy of image data acquisition.

[0046] Specifically, at least three cameras are arranged in the cigarette burning area to achieve comprehensive coverage of the entire burning process. Each camera uses a CCD sensor to ensure image clarity, and uses an adjustable focus lens to achieve clear imaging at different viewing angles. During the cigarette burning process, each camera synchronously collects image data at a set frame rate (such as 20fps) to capture the burning characteristics of different stages, such as the expansion of the crack, the width and stability of the carbon line, etc., to provide high-quality original images for subsequent splicing.

[0047] In an optional embodiment of the present invention, each image is preprocessed, including dynamic exposure adjustment, grayscale balance and color calibration. Dynamic exposure adjustment ensures the clarity and contrast of the image under different lighting conditions, grayscale balance improves the brightness uniformity of the image, and color calibration ensures the color consistency between multi-view images. The combined application of these preprocessing technologies not only improves the visual quality of the image, but also enhances the availability of image data, providing a solid foundation for subsequent image stitching and feature extraction.

[0048] In an optional example of this implementation, the dynamic exposure adjustment includes, based on an automatic brightness control algorithm, adaptively adjusting exposure parameters according to brightness changes in different image regions during the combustion process. The method of adaptively adjusting exposure parameters effectively copes with the rapid changes in lighting conditions during the combustion process, and ensures the clarity and detail restoration of the image under different brightness conditions.

[0049] Specifically, dynamic exposure adjustment is performed on each image: according to the brightness difference of different image areas during cigarette burning, the automatic brightness control algorithm is used to adaptively adjust the exposure. The algorithm analyzes the average brightness value of each area in real time during image acquisition, adjusts the camera's exposure time and gain, and ensures that the image still has good contrast and detail performance under high or low brightness conditions.

[0050] In an optional example of this implementation, grayscale equalization includes adjusting the grayscale of each image through grayscale histogram equalization to make the brightness of each image evenly distributed. The application of grayscale histogram equalization technology makes the brightness distribution of the image more even, improves the contrast and clarity of the image, especially in the highlight and low light areas generated during the cigarette burning process, which not only improves the visual quality of the image, but also enhances the consistency of the image data, so that the spliced ​​panoramic image can be more visually coherent and the feature point matching is more accurate.

[0051] Specifically, grayscale equalization is performed on each image: the grayscale of the image is adjusted through histogram equalization to keep the image brightness evenly distributed in different areas. The grayscale distribution range of the image is expanded to the entire color scale to improve the problem of insufficient contrast caused by lighting changes, making subsequent stitching and feature extraction more accurate.

[0052] In an optional example of this implementation, color calibration includes adjusting the hue and saturation of each image through a color calibration algorithm to correct the color differences of each image caused by changes in viewing angle and light source. By adjusting the hue and saturation, the algorithm successfully corrects the color differences caused by changes in viewing angle and light source, so that images at different viewing angles are highly consistent in color, which not only improves the visual coherence of the image, but also enhances the usability of the image data. The stitched panoramic image is more natural and consistent in color, reducing the visual discontinuity and ghosting caused by color differences.

[0053] Specifically, color calibration is performed on each image: the hue and saturation of images from different perspectives are adjusted through a color calibration algorithm, and color differences caused by perspective changes and light source influences are corrected to ensure color consistency when multi-perspective stitching is performed.

[0054] In an optional example of this implementation, feature points of each image are extracted and matched using a multi-level feature point screening and weighting strategy, including using SIFT algorithm, SURF algorithm or ORB algorithm to extract multi-scale features of the image and obtain the key feature points of the image; multi-level screening of the key feature points according to the confidence level, eliminating low-confidence noise points, and retaining high-confidence key feature points; using the feature points after multi-level screening, matching is performed based on a weighted strategy, and by calculating the Euclidean distance or Hamming distance of the feature points, point pairs with smaller distances are identified as matching points, and marked at the corresponding positions, and different weights are assigned to the screened key feature points based on the confidence level. By implementing the feature point extraction and matching method of multi-level feature point screening and weighting strategy, the accuracy and efficiency of cigarette burning image data acquisition and panoramic stitching are significantly improved.

[0055] Specifically, multi-level feature point screening and weighting strategy are used to match image feature points to improve stitching accuracy:

[0056] 1. Feature point extraction: Extract feature points from each image using SIFT, SURF or ORB algorithms. SIFT and SURF algorithms search for key feature points in multi-scale space, while the ORB algorithm quickly extracts feature points using rotation-invariant binary descriptors.

[0057] 2. Multi-level screening and weighting: After extracting feature points, perform multi-level screening on them according to confidence, remove low-confidence noise points, and retain high-confidence feature points. At the same time, assign higher weights to significant feature points (such as carbon line edges, crack outlines, etc.) to enhance the alignment accuracy of key areas during the splicing process.

[0058] 3. Feature point matching: Utilize the feature points after multi-level screening and match them based on a weighted strategy. By calculating the Euclidean distance or Hamming distance of the feature points, point pairs with smaller distances are identified as matching points and marked at the corresponding positions to improve the accuracy and consistency of the stitching effect.

[0059] In an optional embodiment of the present invention, a multi-scale smoothing process is performed on the stitching edge, including eliminating edge breakage or ghosting caused by angle differences by adjusting the smoothness of each level of features on the stitching path. The multi-scale smoothing process based on the stitching path effectively eliminates the edge breakage or ghosting caused by angle differences when stitching multi-view images, and the smoothing process is performed on each level of features on the stitching path, so that the edge transition of the stitched image is natural and the visual effect is coherent, which greatly improves the visual quality of the panoramic image.

[0060] Specifically, in order to avoid edge breakage or ghosting problems caused by viewing angle differences, a seamless fusion technology based on the splicing path is used to optimize the visual effect of the splicing edge:

[0061] 1. Stitching path optimization: During the stitching process, the best stitching path is determined based on the edge features of the image, and the visual coherence of the stitching is ensured by adjusting the smoothness of each layer of the image in the stitching area.

[0062] 2. Multi-scale smooth fusion: On the stitching path, a multi-scale smoothing method is used to smoothly fuse images of different scales in the stitching area layer by layer, making the transition of the stitching edge natural, thereby reducing visual breaks and ghosting caused by differences in viewing angles or angles.

[0063] In an optional embodiment of the present invention, the cigarette burning image data collection and panoramic stitching method also includes transmitting the stitched image data generated after panoramic stitching to a data platform and forming a cigarette burning image database, so that the database supports real-time query, data sharing and data analysis.

[0064] Specifically, the generated high-resolution burning panoramic image data is transmitted to the big data platform to form a complete cigarette burning image database. During the data transmission process, the spliced ​​panoramic image data is uploaded to the big data platform via a high-speed network and encrypted for transmission to ensure data security.

[0065] In an optional example of this implementation, multi-dimensional annotation is performed on key features in each spliced ​​image data in the cigarette burning image database. By implementing multi-dimensional annotation technology, not only the data in the database is made richer and more structured, but also the data is greatly enhanced in terms of searchability and analyzability.

[0066] Specifically, data storage and structured annotation are carried out by classifying and storing data in the database according to time, combustion stage and image features, and annotating key features in the image (such as crack expansion, ash column integrity, etc.) to form structured data to support subsequent data analysis and model training.

[0067] The present invention also proposes a cigarette burning image data acquisition and panoramic stitching system for implementing the above method, wherein the cigarette burning image data acquisition and panoramic stitching system comprises:

[0068] Multi-angle image acquisition module, used to acquire multi-angle images of the cigarette burning process in real time;

[0069] Data pre-processing module, including dynamic exposure adjustment, grayscale equalization, color calibration and noise removal to optimize image quality;

[0070] Image stitching module, which achieves high-precision stitching based on posture estimation algorithm, multi-level feature point screening and weighting strategy, depth estimation and seamless fusion technology;

[0071] The data storage and analysis module is used to store the generated panoramic image data set and support real-time query, feature annotation and analysis of dynamic combustion changes.

[0072] Specifically, the modular design can realize the full process automation from image acquisition to data analysis. The multi-angle image acquisition module ensures the comprehensiveness of the data, while the data preprocessing module improves the quality of the image data and lays a solid foundation for subsequent analysis. The image stitching module generates high-precision panoramic images through advanced algorithms and technologies, greatly improving the accuracy and reliability of image stitching. The data storage and analysis module not only saves valuable panoramic image data, but also makes the monitoring and evaluation of cigarette burning characteristics more efficient and accurate through real-time query and dynamic analysis functions.

[0073] The present invention also provides a computer-readable storage medium on which a computer program is stored, wherein the computer program implements the steps of the above method when executed by a processor.

[0074] The computer-readable storage medium proposed by the present invention stores executable instructions. When the executable instructions are executed by a processor, the processor will be caused to execute any implementation of a system for implementing a cigarette burning charcoal line detection method based on a hybrid attention mechanism in the above-mentioned embodiments.

[0075] In another optional embodiment of the present invention, the computer-readable storage medium may be a memory such as FRAM, ROM, PROM, EPROM, EE PROM, flash memory, magnetic surface memory, optical disk, or CD-ROM; or various devices including one or any combination of the above memories. In some embodiments, the executable instructions may be in the form of a program, software, software module, script, or code, written in any form of programming language (including compiled or interpreted language, or declarative or procedural language), and may be deployed in any form, including being deployed as an independent program or being deployed as a module, component, subroutine, or other unit suitable for use in a computing environment. As an example, the executable instructions may be deployed to be executed on one computing device, or on multiple computing devices located at one location, or on multiple computing devices distributed at multiple locations and interconnected by a communication network.

[0076] The present invention proposes a computer device, including a memory and a processor, wherein a computer program that can be run on the processor is stored in the memory, wherein when the processor executes the computer program, steps for implementing a cigarette burning charcoal line detection method based on a hybrid attention mechanism are implemented.

[0077] The present invention proposes a computer device, a memory for storing executable instructions; a processor for executing the executable instructions stored in the memory to implement a cigarette burning charcoal line detection method based on a hybrid attention mechanism.

[0078] The present invention proposes a method and system for collecting cigarette burning image data and stitching panoramic images. To ensure spatial consistency during multi-angle image stitching, a posture estimation algorithm is used to calibrate the spatial position and angle of each camera:

[0079] 1. Camera parameter calibration: Based on the camera's intrinsic parameters (focal length, optical center position, radial distortion coefficient) and extrinsic parameters (rotation and translation parameters relative to the reference coordinate system), the camera's shooting posture is accurately calculated.

[0080] 2. Rotation and displacement correction: Based on the rotation matrix and translation matrix, the viewing position of each camera in the multi-view image is uniformly corrected to ensure that images at different angles can be effectively aligned during the stitching process and reduce the spatial deviation caused by perspective differences.

[0081] The present invention proposes a method and system for collecting cigarette burning image data and panoramic stitching. Before image stitching, the present invention further calibrates the image based on depth estimation to ensure the spatial consistency of images from different perspectives:

[0082] 1. Depth map generation: Multi-view stereo matching technology is used to generate a depth map for each pixel based on images from different camera perspectives. The depth map can reflect the distance from each area on the cigarette surface to the camera.

[0083] 2. Image correction: Use the generated depth map to correct the relative scale and displacement of multi-view images to ensure that images from different perspectives can be accurately aligned when stitched together, reducing stitching errors caused by differences in the curvature or depth of the cigarette surface.

[0084] The specific implementation process of the present invention is now described in detail with reference to the embodiments:

[0085] The present invention collects samples with representative size specifications in different scenarios, collects images of the combustion process to establish a cigarette burning char line image dataset, and provides basic experimental image data for the subsequent development of a multi-channel ash burning char line simultaneous, rapid, and accurate detection algorithm. The dataset contains a total of 1,200 images, such as Figure 2a As shown, it includes linear suction equipment detection scenes (500 images); Figure 2b As shown in the figure, the detection scene of the turntable type single-channel suction equipment (200 pictures); Figure 2c As shown in the figure, there are three categories of turntable multi-channel smoking equipment detection scenes (300 images), each of which includes 30 regular, medium and thin cigarette burning images. Considering that there are multiple targets in a single image of a linear smoking device and a turntable multi-channel smoking device, the data set is randomly divided into a training set (600 images), a validation set (360 images), and a test set (240 images) in a ratio of 5:3:2. During the test, the test set was divided into 30 sub-test sets according to the cigarette brand, with 8 images in each test set. The training set data was manually annotated using LabelImg software. The annotated data was strictly produced in the PascalVOC format of the public data set. The annotation file is in xml format, and the data set has been made public through common channels.

[0086] The cigarette burning, smoking and imaging module constructed by the present invention is used to perform cigarette burning and smoking in the ISO standard smoking mode with reference to the linear smoking device. Based on the above designed cigarette panoramic imaging method, a DFK 23GP274 camera (The Imaging Source, Germany) is used in combination with a lens with a focal length of 12.5 mm (Fujifilm, Japan) to collect a 360° panoramic image of the ash area when the cigarette is burning.

[0087] We selected 20 cigarette samples with representative features in terms of size, appearance, specification, type, etc., collected panoramic images of the burning ash area of ​​50 cigarettes of each specification, and established a panoramic image dataset of cigarette burning ash cracks, which provides basic experimental image data for the subsequent development of crack precision segmentation and recognition algorithms. The dataset contains a total of 1,000 panoramic images of cigarette burning ash cracks ( Figure 3a ), it can be seen that the contrast between the target crack area and the background is small, and it is difficult to accurately segment it. By manually using ImageJ software to create the real label (ground-truth) crack template of 1000 images ( Figure 3b ).

[0088] In order to realize the correlation analysis and application of cigarette apparent burning quality data and cigarette product data, based on the data specifications and collection methods constructed by the "Construction and Application Research of Big Data for Quality Monitoring in the Tobacco Industry", a major scientific and technological project of China National Tobacco Corporation, the apparent burning quality index detection and analysis of 1,458 cigarette products in the tobacco industry from 2019 to 2023 were carried out, and the test results were associated with the cigarette product quality information subject database to construct a cigarette product apparent burning quality data set. The information types of the data set include product source information, specification information, physical indicators, smoke indicators, sensory indicators, test results of apparent burning quality indicators, and cigarette burning cone head quality test results. The cigarettes tested in the data set cover more than 90% of the industry's product specifications, with a production coverage rate of more than 95%. The sample types include the current mainstream medium, short, thin, regular and other cigarette specifications, such as Figure 4 shown.

[0089] The data information of the constructed dataset is shown in Table 1, which contains about 700,000 structured data, 120,000 images, and a total data volume of 189G. The constructed dataset provides basic product information for the research paper on the one hand, and provides basic research data for the analysis of the correlation between the apparent combustion quality of cigarettes and physical, smoke, sensory and other indicators on the other hand, for the subsequent improvement of product quality.

[0090] Table 1:

[0091]

[0092] The detailed explanation of the above-mentioned embodiments is only intended to explain the present invention so as to facilitate a better understanding of the present invention. However, these descriptions cannot be interpreted as limitations on the present invention for any reason. In particular, the various features described in different embodiments may also be arbitrarily combined with each other to form other embodiments. Unless there is a clear description to the contrary, these features should be understood to be applicable to any embodiment and are not limited to the described embodiments.

Claims

1. A method for collecting cigarette burning image data and panoramic stitching, characterized in that: The cigarette burning image data collection and panoramic stitching method comprises: Using a multi-angle camera system to collect images from multiple perspectives during the cigarette burning process, the multi-angle images covering different stages and perspectives of cigarette burning; Preprocessing each of the images and reducing the impact of illumination changes during cigarette burning on the clarity of each of the images; The multi-angle camera system includes multiple cameras, and the spatial position and angle of each camera are calibrated by a posture estimation algorithm so that the images collected by each camera have spatial consistency during the stitching process; Extracting and matching feature points of each of the images using a multi-level feature point screening and weighting strategy; Generating a depth map of each pixel using a depth estimation algorithm, and correcting the relative scale and displacement of each of the images according to the depth map; The multi-view images are panoramically stitched, and the stitching edges are smoothed at multiple scales using seamless fusion technology based on the stitching path.

2. The cigarette burning image data acquisition and panoramic stitching method according to claim 1, characterized in that: The multi-angle camera system includes at least three cameras, and the three cameras are respectively located at a front viewing position, a side viewing position and a top viewing position of the cigarette burning area.

3. The cigarette burning image data acquisition and panoramic stitching method according to claim 1, characterized in that: Preprocessing is performed on each of the images, including dynamic exposure adjustment, grayscale balancing and color calibration.

4. The cigarette burning image data acquisition and panoramic stitching method according to claim 3, characterized in that: The dynamic exposure adjustment includes adaptively adjusting exposure parameters according to brightness changes in different image areas during the combustion process based on an automatic brightness control algorithm.

5. The cigarette burning image data acquisition and panoramic stitching method according to claim 3, characterized in that: The grayscale balancing includes: The grayscale of each of the images is adjusted by grayscale histogram equalization so that the brightness of each of the images is evenly distributed.

6. The cigarette burning image data acquisition and panoramic stitching method according to claim 3, characterized in that: The color calibration includes The hue and saturation of each of the images are adjusted by a color calibration algorithm to correct the color differences of each of the images caused by changes in viewing angle and the influence of light source.

7. The cigarette burning image data acquisition and panoramic stitching method according to claim 3, characterized in that: The multi-level feature point screening and weighting strategy is used to extract and match the feature points of each image, including Perform multi-scale feature extraction on the image using a SIFT algorithm, a SURF algorithm or an ORB algorithm, and obtain key feature points of the image; Perform multi-level screening of key feature points according to confidence, remove low-confidence noise points, and retain high-confidence key feature points; Using the feature points after multi-level screening, matching is performed based on a weighted strategy. By calculating the Euclidean distance or Hamming distance of the feature points, point pairs with smaller distances are identified as matching points and marked at the corresponding positions. Different weights are assigned to the selected key feature points based on their confidence.

8. The cigarette burning image data acquisition and panoramic stitching method according to claim 1, characterized in that: Multi-scale smoothing of the splicing edges, including, By adjusting the smoothness of each layer of features on the stitching path, edge breakage or ghosting caused by angle differences can be eliminated.

9. The cigarette burning image data acquisition and panoramic stitching method according to claim 1, characterized in that: The cigarette burning image data collection and panoramic stitching method also includes: The stitched image data generated after the panoramic stitching is transmitted to the data platform to form a cigarette burning image database, so that the database supports real-time query, data sharing and data analysis.

10. The cigarette burning image data acquisition and panoramic stitching method according to claim 9, characterized in that: Multi-dimensional annotation is performed on key features in each of the spliced ​​image data in the cigarette burning image database.

11. A cigarette burning image data acquisition and panoramic stitching system, used to implement the method according to any one of claims 1 to 10, characterized in that: The cigarette burning image data acquisition and panoramic stitching system comprises: Multi-angle image acquisition module, used to acquire multi-angle images of the cigarette burning process in real time; Data pre-processing module, including dynamic exposure adjustment, grayscale equalization, color calibration and noise removal to optimize image quality; Image stitching module, which achieves high-precision stitching based on posture estimation algorithm, multi-level feature point screening and weighting strategy, depth estimation and seamless fusion technology; The data storage and analysis module is used to store the generated panoramic image data set and support real-time query, feature annotation and analysis of dynamic combustion changes.

12. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the steps of the method according to any one of claims 1 to 10 are implemented.

13. A computer device comprising a memory and a processor, wherein a computer program capable of running on the processor is stored in the memory, characterized in that: When the processor executes the computer program, the steps of the method according to any one of claims 1 to 10 are implemented.

Images