Method for identifying stem slivers in tobacco shreds based on color and morphological characteristics
By using a color and morphological feature-based identification method, and employing an industrial camera and multilayer filters to separate tobacco shreds and stems, the problem of long identification time and poor adaptability in existing technologies is solved, achieving high-precision stem identification and detection.
Patent Information
- Application Number
- CN202511085025.0
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2025-08-04
- Publication Date
- 2025-11-18
AI Technical Summary
In existing technologies, the identification of tobacco stems relies on manual identification, which is time-consuming and inefficient. Image processing methods rely on large sample sizes, have poor adaptability, and unstable identification accuracy.
A color and morphological feature-based recognition method is adopted. Images are acquired through an industrial camera, and color feature filters and multi-layer shape feature filters are used to separate tobacco shreds and stems. Combined with quantitative recognition indicators such as gray value, contour area, aspect ratio, rectangularity and group, length and width, the stem targets are gradually filtered out.
It enables accurate identification of stems in different tobacco processing stages, with high accuracy and strong adaptability, providing a scientific basis for the detection of stems in tobacco and improving identification efficiency and accuracy.
Smart Images
Figure CN120976540A_ABST
Abstract
Description
TECHNICAL FIELD
[0001] The present application relates to the technical field of tobacco detection, in particular to a tobacco stem signature recognition method based on color and morphological features. BACKGROUND
[0002] Stem signatures are inevitably produced during the processing of tobacco leaves and stems. Research has shown that stem signatures can easily cause cigarette punctures, air leaks, and burning end blowout, which are important factors affecting cigarette quality, draw stability, combustion, and sensory quality.
[0003] Traditional tobacco stem signature recognition usually uses manual picking and visual identification methods, but these methods are time-consuming, inefficient, and have large differences in recognition results between different personnel.
[0004] Current image processing recognition techniques, such as the invention patent CN113888468, disclose a stem signature recognition and detection method based on image processing, which mainly extracts the shape, color, and texture features of the image and uses random forest deep learning for recognition. Although this method can identify tobacco stem signatures to some extent, it does not involve the specific recognition mechanism and quantitative determination basis of stem signatures, and the recognition accuracy is overly dependent on sample size and training models. Moreover, if the tobacco or stem signature environment changes, the stem signature recognition effect will be greatly reduced, requiring re-sampling, training, modeling, and verification.
[0005] To solve the above problems, the present application is proposed. SUMMARY
[0006] The present application aims to address the shortcomings of the prior art by providing a tobacco stem signature recognition method based on color and morphological features. This method can accurately identify tobacco stem signatures and adapt to different processing steps in the tobacco processing process, providing a scientific basis for online detection of tobacco stem signature content and accurate removal of stem signatures.
[0007] To solve the above technical problems, the present application adopts the following technical solutions:
[0008] A tobacco stem signature recognition method based on color and morphological features, comprising:
[0009] S1, obtaining image features of a tobacco and stem mixture, and converting the image features into a single-channel gray-scale image H component map;
[0010] S2, using a color feature filter to separate the image of the tobacco and the background in the H component map;
[0011] S3, using a first shape feature filter to further separate the tobacco image in the S2 image features;
[0012] S4. The tobacco image in the image features of S3 is further separated using the second shape feature filter;
[0013] S5. Perform connected component processing and labeling on the image features separated in S4 to determine the boundaries and coordinates of the stems, thereby identifying the stems and corresponding information in the tobacco.
[0014] Furthermore, in step S1, an industrial camera is used to acquire image features of the mixture of tobacco shreds and stems, and a blue background and white light source are used for illumination during the image acquisition process.
[0015] Furthermore, in step S1, the H component image of a single-channel grayscale image is obtained by extracting or combining the RGB three colors of the image features separately.
[0016] Furthermore, in step S2, separating the tobacco image and background in the H component image using a color feature filter includes:
[0017] S201. Binarize the H component and set the gray value threshold range of the pixel to [0, 255].
[0018] S202. Design a color feature filter, using grayscale value Gray = 80 as the boundary condition;
[0019] S203. Gray value judgment: Filter pixels with a gray value threshold range of [0,80] to separate some tobacco from the background.
[0020] Furthermore, in step S3, the first shape feature filter uses contour area and aspect ratio shape features to separate the tobacco image, specifically including:
[0021] S301. Calculate the outline area and aspect ratio lwRatio of each tobacco shred or stem object in the image of S2.
[0022] S302. Design the first shape feature filter, setting the threshold range of the contour area Area to [300mm, 12000mm], and the threshold range of the aspect ratio lwRatio to [3, 100].
[0023] S303. Filter out objects that do not meet the outline area threshold [300mm, 12000mm] and aspect ratio threshold [3, 100], thereby further separating some tobacco from the background.
[0024] Furthermore, in step S4, the second shape feature filter uses rectangularity and perimeter ratio morphological features to separate the tobacco shreds image, specifically including:
[0025] S401. Calculate the outline rectangle Rect, outline perimeter c1 and convex polygon perimeter c2 of each tobacco shred or stem object in the image of S3. The ratio of the two perimeters is denoted as the perimeter ratio Cr, and its formula is: Cr = c1 / c2.
[0026] S402. Design a second shape feature filter, setting the threshold range for contour rectangle (Rect) to [0.35, 1] and the threshold range for perimeter ratio (Cr) to [0.75, 1].
[0027] S403. Filter out objects that do not meet the Rect threshold [0.35, 1] and Cr threshold [0.75, 1], thereby separating the remaining tobacco from the background.
[0028] Compared with the prior art, the beneficial effects of the present invention are:
[0029] The tobacco stem tag identification mechanism based on color and morphological features of the present invention clarifies the specific identification conditions and order, and determines the quantitative identification index based on color, outline area, aspect ratio, rectangularity and perimeter ratio features, which breaks through the limitations of the current stem tag image recognition, such as relying on a large number of samples for training and narrow range of application.
[0030] The tobacco stem identification mechanism based on color and morphological characteristics of this invention can accurately identify tobacco stem targets. It can also be widely applied to tobacco stem identification in different tobacco processing stages, providing technical support for accurate detection of tobacco stem content and stem removal. It has good application prospects and promotion value in the entire tobacco industry. Attached Figure Description
[0031] Figure 1 The flowchart shows the method for identifying tobacco stem tags based on color and morphological features provided by the present invention.
[0032] Figure 2 This is a flowchart illustrating the specific process of using a color feature filter to separate tobacco shreds.
[0033] Figure 3 The flowchart shows the specific process of separating tobacco shreds using the first shape feature filter.
[0034] Figure 4 The flowchart shows the specific process of using a second shape feature filter to separate tobacco shreds.
[0035] Figure 5 Image for identifying stem tags in tobacco shreds after thin-plate drying at a cigarette factory. Detailed Implementation
[0036] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, the technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. The components of the embodiments of the present invention described and shown in the accompanying drawings can generally be arranged and designed in various different configurations.
[0037] Please see Figure 1 This invention provides a method for identifying tobacco stem tags based on color and morphological features, comprising:
[0038] S1. Obtain the image features of the mixture of tobacco shreds and stems, and convert the image features into a single-channel grayscale image H component map;
[0039] Specifically, in this step, an industrial camera is used to acquire image features of the mixture of tobacco shreds and stems. During the image acquisition process, a blue background and white light source are used for illumination. For the H component image, in this embodiment, the image features are processed by extracting or combining the RGB three colors separately to obtain a single-channel grayscale image H component image.
[0040] S2. Use a color feature filter to separate the tobacco image from the background in the H component image;
[0041] refer to Figure 2 As shown, the specific separation process includes:
[0042] S201. Binarize the H component and set the gray value threshold range of the pixel to [0, 255].
[0043] S202. Design a color feature filter, using grayscale value Gray = 80 as the boundary condition;
[0044] S203. Gray value judgment: Filter pixels with a gray value threshold range of [0,80] to separate some tobacco from the background.
[0045] S3. The tobacco image in the image features of S2 is further separated using the first shape feature filter;
[0046] In this embodiment, the first shape feature filter uses contour area and aspect ratio shape features to separate tobacco shred images, referring to... Figure 3 As shown, the separation process specifically includes: S301, calculating the outline area and aspect ratio lwRatio of each tobacco shred or stem object in the image in S2;
[0047] S302. Design the first shape feature filter, setting the threshold range of the contour area Area to [300mm, 12000mm], and the threshold range of the aspect ratio lwRatio to [3, 100].
[0048] S303. Filter out objects that do not meet the outline area threshold [300mm, 12000mm] and aspect ratio threshold [3, 100], thereby further separating some tobacco from the background.
[0049] Some tobacco shreds cannot be filtered by the color feature filter in S2, but can be filtered by another method in S303, thereby expanding the range of tobacco shred filtration.
[0050] S4. The tobacco image in the image features of S3 is further separated using the second shape feature filter;
[0051] In this embodiment, the second shape feature filter uses rectangularity and perimeter ratio morphological features to separate tobacco shred images, referencing... Figure 4 As shown, the separation process specifically includes:
[0052] S401. Calculate the outline rectangle Rect, outline perimeter c1 and convex polygon perimeter c2 of each tobacco shred or stem object in the image of S3. The ratio of the two perimeters is denoted as the perimeter ratio Cr, and its formula is: Cr = c1 / c2.
[0053] S402. Design a second shape feature filter, setting the threshold range for contour rectangle (Rect) to [0.35, 1] and the threshold range for perimeter ratio (Cr) to [0.75, 1].
[0054] S403. Filter out objects that do not meet the Rect threshold [0.35, 1] and Cr threshold [0.75, 1], thereby separating the remaining tobacco from the background.
[0055] In S2, some tobacco shred features are filtered by color. In S303, some tobacco shred features are filtered by contour area and aspect ratio. However, some tobacco shred features still exist. In order to further expand the tobacco shred filtering range, S403 is used to further filter the remaining tobacco shred features, thereby improving the stem tag recognition accuracy.
[0056] S5. Perform connected component processing and labeling on the image features separated in S4, that is, identify and number the boundaries of the stem objects, determine the boundaries and coordinates of the stems, and thus identify the stems and corresponding information in the tobacco.
[0057] It is worth noting that the color feature filter, the first shape feature filter, and the second shape feature filter described above can be independent device structures or algorithms set in a computer.
[0058] In this application, the image features of tobacco shreds are gradually filtered out by quantitative identification indicators such as color features, contour area and aspect ratio, rectangularity and perimeter ratio. This filtering and separation process is a data processing process rather than directly eliminating tobacco shreds from the image. Through this step-by-step filtering process, the stems and other targets in the tobacco shreds can be accurately identified.
[0059] To demonstrate the accuracy of this identification method, after the thin-plate drying process on a cigarette manufacturing production line, 1 kg of dried tobacco shreds was taken, thoroughly loosened and evenly divided into 5 portions, each weighing 0.2 kg. Each portion was then used to identify the stem tags on a static image recognition device according to the identification method of this invention. The static image recognition device was configured as follows: a Hikvision MV-CL022-40GC industrial camera with a resolution of 2592×2048, mounted perpendicular to the platform surface; a white artificial LED point light source (BC-DW30-W) and a white light source controller (BC-DPS1Z-8A4C) with a light intensity of 7660–7830 lux on the platform surface. The camera was connected to a laptop computer, which displayed the captured images and stored them as a photo set. These images were then transferred to another laptop computer for stem tag image recognition processing.
[0060] The identified meme tags are marked using the identification method of this application, wherein the meme tag identification image is as follows: Figure 5 As shown, the identification of jokes is then checked manually. The purposes are threefold: first, to determine whether identified jokes are genuine or misidentified; second, to locate unidentified jokes in the image; and third, to statistically analyze the joke recognition accuracy and unidentified rate. Based on this, the joke recognition accuracy and unidentified rate are calculated using the following formulas:
[0061] Joke recognition accuracy = (Number of jokes confirmed by manual judgment / Number of jokes recognized by jokes) × 100%;
[0062] The unidentified tag rate = number of tags that were not identified by manual judgment / (number of tags that were not identified by manual judgment + number of tags that were confirmed to be tags by manual judgment); The tag identification accuracy and tag unidentified rate data obtained for these 5 samples are shown in Table 1.
[0063] Table 1. Accuracy and non-identification rates of stem sticks in tobacco shreds after drying of conventional cigarette sheets.
[0064]
[0065] As shown in Table 1, the stem identification mechanism of this invention achieves an accuracy of 98.1% in identifying stems in tobacco shreds after the thin-plate drying process in a conventional cigarette manufacturing process, with a stem non-identification rate of only 6.5%. This demonstrates that the identification mechanism of this invention can accurately identify stems in tobacco shreds of different specifications and at different processing stages, with a wide range of applications. It provides technical support for the accurate detection of stem content and stem removal in tobacco shreds, and has good application prospects and promotional value.
[0066] Although the invention has been described herein with reference to several illustrative embodiments, it should be understood that many other modifications and implementations can be devised by those skilled in the art, which will fall within the scope and spirit of the principles disclosed herein. More specifically, various modifications and improvements can be made to the components or layout of the subject matter arrangement within the scope of the disclosure, drawings, and claims. Besides modifications and improvements to the components or layout, other uses will be apparent to those skilled in the art.
Claims
1. A method for identifying tobacco stem tags based on color and morphological features, characterized in that, include: S1. Obtain the image features of the mixture of tobacco shreds and stems, and convert the image features into a single-channel grayscale image H component map; S2. Use a color feature filter to separate the tobacco image from the background in the H component image; S3. The tobacco image in the image features of S2 is further separated using the first shape feature filter; S4. The tobacco image in the image features of S3 is further separated using the second shape feature filter; S5. Perform connected component processing and labeling on the image features separated in S4 to determine the boundaries and coordinates of the stems, thereby identifying the stems and corresponding information in the tobacco.
2. The method for identifying tobacco stem tags based on color and morphological features according to claim 1, characterized in that: In step S1, an industrial camera is used to acquire image features of the mixture of tobacco shreds and stems, and a blue background and white light source are used for illumination during the image acquisition process.
3. The method for identifying tobacco stem tags based on color and morphological features according to claim 1, characterized in that: In step S1, the H component image of a single-channel grayscale image is obtained by extracting or combining the RGB three colors of the image features separately.
4. The method for identifying tobacco stem tags based on color and morphological features according to claim 1, characterized in that: In step S2, separating the tobacco image and background in the H component image using a color feature filter includes: S201. Binarize the H component and set the gray value threshold range of the pixel to [0, 255]. S202. Design a color feature filter, using grayscale value Gray = 80 as the boundary condition; S203. Gray value judgment: Filter pixels with a gray value threshold range of [0,80] to separate some tobacco from the background.
5. The method for identifying tobacco stem tags based on color and morphological features according to claim 1, characterized in that: In step S3, the first shape feature filter uses contour area and aspect ratio shape features to separate the tobacco image, specifically including: S301. Calculate the outline area and aspect ratio lwRatio of each tobacco shred or stem object in the image of S2. S302. Design the first shape feature filter, setting the threshold range of the contour area Area to [300mm, 12000mm], and the threshold range of the aspect ratio lwRatio to [3, 100]. S303. Filter out objects that do not meet the outline area threshold [300mm, 12000mm] and aspect ratio threshold [3, 100], thereby further separating some tobacco from the background.
6. The method for identifying tobacco stem tags based on color and morphological features according to claim 1, characterized in that: In step S4, the second shape feature filter uses rectangularity and perimeter ratio morphological features to separate the tobacco shreds image, specifically including: S401. Calculate the outline rectangle Rect, outline perimeter c1 and convex polygon perimeter c2 of each tobacco shred or stem object in the image of S3. The ratio of the two perimeters is denoted as the perimeter ratio Cr, and its formula is: Cr = c1 / c2. S402. Design a second shape feature filter, setting the threshold range for contour rectangle (Rect) to [0.35, 1] and the threshold range for perimeter ratio (Cr) to [0.75, 1]. S403. Filter out objects that do not meet the Rect threshold [0.35, 1] and Cr threshold [0.75, 1], thereby separating the remaining tobacco from the background.