WO2023226103A1 - 一种打叶复烤风分过程中叶片结构测量方法及风分器出叶量测量方法 - Google Patents

一种打叶复烤风分过程中叶片结构测量方法及风分器出叶量测量方法 Download PDF

Info

Publication number
WO2023226103A1
WO2023226103A1 PCT/CN2022/098572 CN2022098572W WO2023226103A1 WO 2023226103 A1 WO2023226103 A1 WO 2023226103A1 CN 2022098572 W CN2022098572 W CN 2022098572W WO 2023226103 A1 WO2023226103 A1 WO 2023226103A1
Authority
WO
WIPO (PCT)
Prior art keywords
image
pixel area
leaf
fragments
medium
Prior art date
Application number
PCT/CN2022/098572
Other languages
English (en)
French (fr)
Inventor
武凯
熊文
蔡炳彪
朱保昆
岳衡
蔡昊城
陆舍铭
卢婷
许健
倪朝敏
秦文平
苏毅
赵辉
周海东
贾蔚
汪海
于春霞
王明敬
乔丹娜
吴亿勤
倪旭东
李韶阳
彭彦华
杨钰婷
王鹏
伊奥尔
Original Assignee
云南中烟工业有限责任公司
Priority date (The priority date is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the date listed.)
Filing date
Publication date
Priority claimed from CN202210560853.5A external-priority patent/CN115018861A/zh
Priority claimed from CN202210561024.9A external-priority patent/CN115035048A/zh
Application filed by 云南中烟工业有限责任公司 filed Critical 云南中烟工业有限责任公司
Priority to EP22859572.4A priority Critical patent/EP4307225A1/en
Publication of WO2023226103A1 publication Critical patent/WO2023226103A1/zh

Links

Images

Classifications

    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
    • G06T7/00—Image analysis
    • G06T7/0002—Inspection of images, e.g. flaw detection
    • G06T7/0004—Industrial image inspection
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
    • G06T7/00—Image analysis
    • G06T7/10—Segmentation; Edge detection
    • G06T7/11—Region-based segmentation
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
    • G06T7/00—Image analysis
    • G06T7/10—Segmentation; Edge detection
    • G06T7/136—Segmentation; Edge detection involving thresholding
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
    • G06T7/00—Image analysis
    • G06T7/10—Segmentation; Edge detection
    • G06T7/174—Segmentation; Edge detection involving the use of two or more images
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
    • G06T7/00—Image analysis
    • G06T7/10—Segmentation; Edge detection
    • G06T7/187—Segmentation; Edge detection involving region growing; involving region merging; involving connected component labelling
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
    • G06T7/00—Image analysis
    • G06T7/10—Segmentation; Edge detection
    • G06T7/194—Segmentation; Edge detection involving foreground-background segmentation
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
    • G06T7/00—Image analysis
    • G06T7/60—Analysis of geometric attributes
    • G06T7/62—Analysis of geometric attributes of area, perimeter, diameter or volume
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
    • G06T2207/00—Indexing scheme for image analysis or image enhancement
    • G06T2207/30—Subject of image; Context of image processing
    • G06T2207/30108—Industrial image inspection
    • G06T2207/30128—Food products

Definitions

  • the invention relates to the technical field of blade threshing and re-bake detection, specifically a machine vision-based method for segmenting blade shapes and counting during the air separation process, and a method for measuring the amount of leaves from an air separator based on the blade structure.
  • the blade structure In the threshing and re-curing process of flue-cured tobacco, the blade structure (referred to as "blade structure”, also often referred to as "leaf shape segmentation and statistics" in the tobacco industry, the two are synonyms and can be used interchangeably) is a An important production quality evaluation index.
  • the defoliating air separation unit is basically a black box.
  • the current blade structure requires sampling after the blades have been defoliated by the defoliating air separation unit.
  • the screening method is used for offline testing. There are problems such as serious lag in detection results and poor real-time performance, and it cannot be timely.
  • the process parameters are adjusted based on the test results, and the sheet structure cannot be effectively controlled and improved.
  • the amount of leaves produced by the air separator is also an important indicator to judge the level of the air separation process. Analysis and knowledge of the working status and working quality of each leaf breaker at all levels of air separation can provide It provides a strong basis for debugging equipment parameters.
  • the air separator uses the principle of wind flotation to separate the stem and leaf mixture output from the topper into pure leaves and tobacco stems. In production practice, the amount of leaves output from the air separator (that is, the output of tobacco leaves per unit time) is an important production indicator that most directly reflects the output efficiency of the air separation section.
  • the air separation efficiency is determined by many factors such as the structural parameters of the air separator itself and process condition parameters.
  • the value of the air separation efficiency C needs to be calibrated according to the actual process parameters. And it often needs to be recalibrated regularly to adapt to changes in the actual process parameters of the air separator, which brings a lot of trouble to the production monitoring of the air separator, because such a calibration process is very cumbersome and requires a large number of production line experiments.
  • Technical personnel have always hoped to have A more direct and convenient method to determine the amount of leaves produced by the air separator.
  • the present invention aims to solve the above problems.
  • the first aspect of the present invention proposes a machine vision-based blade structure measurement method during the threshing and re-baking process, which includes:
  • Step 1 Obtain the background image without blades and the real-time image with blades in the wind separator in the wind separation state through the machine vision imaging system;
  • Step 2 Perform the same denoising processing and image enhancement processing on the background image and the real-time image respectively, perform a difference operation on the processed images, and obtain a difference image;
  • Step 3 Grayscale the difference image, and then use the threshold method of global fixed leaf grayscale threshold segmentation to segment the leaf image to obtain the binary image of the leaf;
  • Step 4 According to the preset pixel area division standards of ultra-large, large, medium, small, and fragments, use the connected domain analysis method to count the pixel areas of ultra-large, large, medium, small, and fragments.
  • Step 5 Extract blade duty cycle data and blade structure distribution data by calculating the pixel area.
  • the machine vision imaging system in step 1 includes an air separator, a camera, and a camera bracket.
  • the camera is installed on the camera bracket to take pictures of the interior of the air separator to collect images.
  • the threshold method for global fixed leaf grayscale threshold segmentation includes: setting the grayscale interval [Tmin, Tmax] for image extraction, setting a threshold T in this interval, and Tmin ⁇ T ⁇ Tmax, dividing the grayscale into Pixels with a grayscale value greater than T are set to white, and pixels less than or equal to T are set to black; or conversely, pixels with a grayscale value greater than T are set to black, and pixels less than or equal to T are set to white.
  • the pixel area demarcation standard the pixel area range of super large films: 25000-500000, the pixel area range of large films: 1500-25000, the pixel area range of medium films: 600-1500, the pixel area range of small films: 100 -600, pixel area range of fragments: 50-100.
  • the blade duty cycle data is obtained by calculating the ratio of the blade area on the image to the total area of a single image.
  • the blade structure distribution data is obtained by calculating the percentage of the pixel area of a certain sheet type to the sum of the pixel areas of all sheet types.
  • a second aspect of the present invention proposes a method for measuring the amount of leaves produced by an air separator based on the blade structure, including:
  • Step 1 Obtain real-time images of the blades in the air separator in the air separation state through the machine vision imaging system
  • Step 2 Process the real-time image
  • Step 3 According to the preset pixel area division standards of large films, medium films, small films and fragments, obtain the area ratios of large films, medium films, small films and fragments on the processed image;
  • Step 4 Measure the leaf output of the air separator according to the following preset functional relationship:
  • Amount of leaves 1776.44-5.77*X 1 -185.71*X 4 -0.65*X 2 *X 2 -0.30*X 3 *X 3 +56.44*X 4 *X 4 , where X 1 is the area rate of large tracts, X 2 is the area ratio % of the medium flakes, X 3 is the area ratio % of the small flakes, and X 4 is the area ratio % of the fragments.
  • the machine vision imaging system in step 1 includes an air separator, an industrial camera, two LED lights, and an electrical control cabinet.
  • the industrial camera is located directly in front of the upwind bin of the air separator, and the LED light is in front of the industrial camera.
  • the electric control cabinet is connected to industrial cameras and LED lights through data lines.
  • the electric control cabinet includes industrial computers and monitors. The industrial computers and monitors are connected through data lines.
  • the image processing in step 2 includes grayscale, binarization, and denoising;
  • the binarization includes: setting the grayscale interval [Tmin, Tmax] for image extraction, setting a threshold T in this interval, and Tmin ⁇ T ⁇ Tmax, and setting the pixels with a grayscale value greater than T as White, pixels less than or equal to T are set to black; or conversely, pixels with a grayscale value greater than T are set to black, and pixels less than or equal to T are set to white.
  • the pixel area delimitation standard in step 3 is: the pixel area range of large pieces: 1500-25000, the pixel area range of medium pieces: 600-1500, the pixel area range of small pieces: 100-600, the pixel area of fragments Range: 50-100.
  • the step 3 to obtain the area ratios of large sheets, medium sheets, small sheets and fragments includes obtaining the areas of large sheets, medium sheets, small pieces and fragments through connected domain analysis method, and then dividing them respectively by the horizontal area in the air separation chamber of the air separator.
  • the cross-sectional area gives the area ratio of the large, medium, small and fragmented pieces.
  • the blade structure measured in the first aspect of the present invention is used in the method of measuring the amount of leaves from an air separator based on the blade structure in the second aspect of the present invention.
  • Blade structure data provides support for feedback control, real-time evaluation, and detection methods of the quality of the defoliation process.
  • the present invention uses machine vision and algorithms to instantly measure the blade structure, that is, the proportion of large blades, medium blades, small blades and fragments, and finds a method to calculate the amount of leaves from the air separator through a self-created function. It is not only simple and practical, but also has real-time capabilities. sex.
  • the present invention breaks through the traditional cognition and directly relates the relative proportion of the blade structure to the amount of leaf extraction using a function, and the quantification is accurate. It can be seen from the functional relationship expression of the present invention that there is no term such as the amount of leaf insertion A.
  • Figure 1 is the background image without blades.
  • Figure 2 is a real-time blade diagram when there are blades.
  • Figure 3 is the difference image between the background image and the real-time blade image.
  • Figure 4 is a binary image of leaves based on a global fixed gray threshold.
  • Figure 5 is a pixel image of a very large blade.
  • Figure 6 is a pixel image of a large leaf.
  • Figure 7 is a pixel image of the middle blade.
  • Figure 8 is a pixel image of a small leaf.
  • Figure 9 is a pixel image of a debris blade.
  • Figure 10 is a schematic diagram of the machine vision device.
  • 1-air separator 2-LED light
  • 3-industrial camera 3-electric control cabinet.
  • This embodiment relates to a machine vision-based method for measuring the blade structure during the threshing and re-baking process, which includes:
  • Step 1 Obtain the background image without blades and the real-time image with blades in the air separation state in the air separator through the machine vision imaging system.
  • the machine vision imaging system includes a wind separator, a camera, and a camera stand.
  • the camera is installed on the camera stand to take pictures of the inside of the wind separator to collect images.
  • Figure 1 is the background image without blades.
  • Figure 2 is a real-time leaf image when there are leaves. The time interval for the camera to collect images is 50ms.
  • Step 2 Perform the same denoising processing and image enhancement processing on the background image and the real-time image respectively, and perform a difference operation on the processed images to obtain a difference image.
  • the denoising process is implemented through the Gaussian filtering algorithm, and the image enhancement process is implemented using the image enhancement algorithm. After these two steps of processing, the image feature information will be enhanced. Then the processed background image and the real-time image are compared to obtain a difference image, as shown in Figure 3.
  • Step 3 Grayscale the difference image, and then use the threshold method of global fixed leaf grayscale threshold segmentation to segment the leaf image to obtain a binary image of the leaf.
  • the threshold method for global fixed leaf grayscale threshold segmentation includes: setting the grayscale interval [Tmin, Tmax] for image extraction, setting a threshold T in this interval, and Tmin ⁇ T ⁇ Tmax, and selecting pixels whose grayscale value is greater than T Set it to white, and set the pixels less than or equal to T to black; or conversely, set the pixels with a grayscale value greater than T to black, and set the pixels less than or equal to T to white.
  • the binary image of the blade is finally obtained, as shown in Figure 4.
  • Step 4 According to the preset pixel area division standards of ultra-large, large, medium, small, and fragments, use the connected domain analysis method to count the pixel areas of ultra-large, large, medium, small, and fragments.
  • the pixel area range of super large films 25000-500000
  • the pixel area range of large films 1500-25000
  • the pixel area range of medium films 600-1500
  • the pixel area range of small films 100-600
  • the pixel area range of fragments 50-100.
  • Figures 5-9 show the pixel images of super large films, large films, medium films, small films and fragments respectively.
  • Step 5 Extract blade duty cycle data and blade structure distribution data by calculating the pixel area.
  • the blade duty cycle data is obtained by calculating the ratio of the blade area on the image to the total area of a single image;
  • the blade structure distribution data is obtained by calculating the percentage of the pixel area of a certain blade type to the total pixel area of all blade types.
  • this machine vision imaging system platform can obtain and analyze the video image of the air separation status in the air separator, so that the corresponding duty cycle data and blade structure data can be quickly extracted to realize the defoliation process.
  • Quality feedback control, real-time evaluation, and detection methods provide support.
  • This embodiment relates to a method for measuring the amount of leaves produced by an air separator based on the blade structure, including:
  • Step 1 Obtain real-time images of blades in the air separation state in the air separator through the machine vision imaging system.
  • the machine vision imaging system builds a visual imaging platform outside the glass window of the upwind bin of the air separator, including an industrial camera and two LED lights, which are used for real-time imaging and detection of high-speed moving tobacco leaves inside the upwind bin. Lighting inside the wind separation bin.
  • the industrial camera used for image collection is located directly in front of the transparent plexiglass of the upwind bin of the wind separator. The camera placement distance needs to be determined according to the camera's field of view to ensure that the picture covers the complete upwind bin.
  • the selected camera resolution is 1280 ⁇ 1024, and the shooting time interval is 50ms.
  • FIG. 10 represents the LED lights, which are located on the left and right sides of the camera.
  • the electronic control cabinet is next to the air distributor.
  • the cabinet includes industrial computers for data transmission and image processing and LCD monitors for image display.
  • the electrical cabinet is connected to cameras and LED light sources respectively.
  • Step 2 Process the real-time image.
  • the method is:
  • the R channel Since the color of tobacco leaves tends to be yellow, among the three RGB channels, the R channel is the most sensitive. The R channel in the tobacco leaf image is extracted for subsequent analysis.
  • Tmin, Tmax Set the grayscale interval [Tmin, Tmax] for image extraction, set a threshold T in this interval, and Tmin ⁇ T ⁇ Tmax, set the pixels with a grayscale value greater than T as white, and set the pixels with a grayscale value less than or equal to T as white. Black; or conversely, set the pixels with a grayscale value greater than T to black, and set the pixels with a grayscale value less than or equal to T to white.
  • the method used in this invention is to open the image to denoise:
  • S(x,y) represents the structural element used in the opening operation.
  • the structural element used in the present invention is a 3 ⁇ 3 circular structure. After the denoising is completed, the final processed result can be obtained, as shown in Figure 4 Show.
  • Step 3 According to the preset pixel area division standards of large films, medium films, small films and fragments, obtain the area ratios of large films, medium films, small films and fragments on the processed image. Use connected domain analysis to obtain the areas of large, medium, small and fragmented areas.
  • Pixel area demarcation standards the pixel area range of large films: 1500-25000, the pixel area range of medium films: 600-1500, the pixel area range of small films: 100-600, and the pixel area range of fragments: 50-100. Since the cross-sectional area of the wind compartment on the air separator is known, the corresponding area ratio can be obtained by dividing the area of the large, medium, small and fragmented pieces by the cross-sectional area.
  • Figures 6-9 show the pixel images of very large, large, medium, small and fragmented films respectively.
  • Step 4 Measure the amount of leaves produced by the air separator according to the preset functional relationship.
  • the functional relationship is:
  • Amount of leaves 1776.44-5.77*X 1 -185.71*X 4 -0.65*X 2 *X 2 -0.30*X 3 *X 3 +56.44*X 4 *X 4 , where X 1 is the area rate of large tracts, X 2 is the area ratio % of the medium flakes, X 3 is the area ratio % of the small flakes, and X 4 is the area ratio % of the fragments.

Landscapes

  • Engineering & Computer Science (AREA)
  • Physics & Mathematics (AREA)
  • Computer Vision & Pattern Recognition (AREA)
  • General Physics & Mathematics (AREA)
  • Theoretical Computer Science (AREA)
  • Quality & Reliability (AREA)
  • Geometry (AREA)
  • Image Analysis (AREA)

Abstract

一种基于机器视觉的打叶复烤风分过程叶片机构测量方法，通过机器视觉成像系统获取风分器（1）内风分状态下的无叶片的背景图像与有叶片的实时图像，对其分别进行去噪处理和增强处理然后做差，对差值图像进行灰度化，再运用全局固定叶片灰度阈值分割的threshold方法对叶片图像进行分割，获取叶片的二值化图像，按照预设的超大片、大片、中片、小片和碎片的像素面积划分标准，采用连通域分析法统计超大片、大片、中片、小片和碎片的像素面积，通过对像素面积的计算来提取叶片占空比数据和叶片结构分布数据。基于从处理后的图像上得到大片、中片、小片和碎片的面积率，还可以依据预设函数关系式测量风分器的出叶量。

Description

一种打叶复烤风分过程中叶片结构测量方法及风分器出叶量测量方法 技术领域
本发明涉及叶片打叶复烤检测技术领域，具体为一种基于机器视觉的打叶复烤风分过程叶片形态分割及统计方法，以及根据叶片结构测量风分器出叶量的方法。
背景技术
在烤烟打叶复烤工艺段，叶片的片型结构(简称“叶片结构”，在烟草行业内也通常被称为“叶片形态分割及统计”，二者是同义词，可互换使用)是一项重要的生产质量评价指标。打叶风分机组基本上是一个黑匣子，目前片型结构需要待叶片经打叶风分机组打叶后取样，采用筛分法进行离线检测，存在检测结果滞后严重、实时性差等问题，不能及时根据检测结果调整工艺参数，片型结构无法得到有效的控制和提高。
机器视觉随着人工智能技术在工业领域落地，而逐渐深入到工业生产的各种场景之中。但目前国内外的打叶复烤企业，尚没有对风分器进行可视化成像的研究。
此外，在烟草加工的风分工艺中，风分器出叶量也是判断风分工艺工水平的一个重要指标，分析获知每一个打叶机以各级风分的工作状态和工作质量,可以为设备参数的调试提供了有力的依据。风分器是一种利用风力浮选原理将打叶机输出的梗、叶混合物分离成单纯的叶片和烟梗。在生产实践中，风分器出叶量(即单位时间内烟叶产出量)是最直接体现风分工段产出效率的重要生产指标，技术人员通常认为风分器出叶量Y＝进叶量A*叶片占比B(叶片占比即叶片质量占叶片和烟梗总质量的比例)*风分效率C，其中叶片占比通过对进料进行筛分后分别测量烟叶重量和烟梗重量而确定，风分效率C依赖于风分器本身的设备参数和工作参数而标定，叶梗完全分离则风分效率C为100％，叶梗不完全分离则风分效率C为介于0％和100％之间的某个值。原料波动会导致叶片占比经常变化，需要针对每一批次原料进行测量。而风分效率是由风分器自身结构参数和工艺条件参数等多方面因素影响而确定的，对于某台具体的风分器，该风分效率C的值需要根据实际工艺参数标定后得到，且经常需要定期重新标定，以适应风分器实际工艺参数的变化，这给风分器生产监控带来很多麻烦，因为这样的标定工序很繁琐，需要大量生产线实验才行，技术人员一直希望有更直接、更方便地确定风分器出叶量的方法。
本发明旨在解决以上问题。
发明内容
本发明第一方面提出了一种基于机器视觉的打叶复烤风分过程叶片结构测量方法，包括：
步骤1:通过机器视觉成像系统获取风分器内风分状态下的无叶片的背景图像与有叶片的 实时图像；
步骤2:对所述的背景图像和实时图像分别进行相同的去噪处理和图像增强处理，对处理后的图像进行做差运算，得到差值图像；
步骤3:对差值图像进行灰度化，再运用全局固定叶片灰度阈值分割的threshold方法对叶片图像进行分割，获取叶片的二值化图像；
步骤4:按照预设的超大片、大片、中片、小片和碎片的像素面积划分标准，采用连通域分析法统计超大片、大片、中片、小片和碎片的像素面积。
步骤5:通过对像素面积的计算来提取叶片占空比数据和叶片结构分布数据。
优选的：所述步骤1中的机器视觉成像系统包括风分器、相机，相机支架，所述相机安装在相机支架上用以对风分器内部拍照以采集图像。
优选的：所述全局固定叶片灰度阈值分割的threshold方法包括：设定图像提取的灰度区间[Tmin,Tmax],在该区间内设置一个阈值T，且Tmin＜T＜Tmax，将灰度值大于T的像素设为白色，小于或者等于T的像素设为黑色；或者反过来，将灰度值大于T的像素设为黑色，小于或者等于T的像素设为白色。
优选的：所述像素面积划定标准：超大片的像素面积范围：25000-500000，大片的像素面积范围：1500-25000，中片的像素面积范围：600-1500，小片的像素面积范围：100-600，碎片的像素面积范围：50-100。
优选的：通过计算图像上的叶片面积与单张图像总面积的比值来获得叶片占空比数据。
优选的：通过求算某一片型像素面积占所有片型像素面积总和的百分比来获得叶片结构分布数据。
本发明第二方面提出一种根据叶片结构测量风分器出叶量的方法，包括：
步骤1:通过机器视觉成像系统获取风分器内风分状态下的有叶片的实时图像；
步骤2:对实时图像进行处理；
步骤3:按照预设的大片、中片、小片和碎片的像素面积划分标准，在处理后的图像上得到大片、中片、小片和碎片的面积率；
步骤4:依据以下预设函数关系式测量风分器的出叶量：
出叶量＝1776.44-5.77*X 1-185.71*X 4-0.65*X 2*X 2-0.30*X 3*X 3+56.44*X 4*X 4，其中X 1是大片面积率，X 2是中片面积率％，X 3是小片面积率％，X 4是碎片面积率％。
优选的，所述步骤1中的机器视觉成像系统包括风分器、工业相机、两个LED灯、一个 电控柜，工业相机位于风分器的上风分仓的正前方，LED灯在工业相机两侧，电控柜与工业相机和LED灯通过数据线连接，电控柜包括工控机和显示器，工控机与显示器通过数据线连接。
优选的，所述步骤2对图像进行处理包括灰度化、二值化、去噪；
优选的，所述二值化包括：设定图像提取的灰度区间[Tmin,Tmax],在该区间内设置一个阈值T，且Tmin＜T＜Tmax，将灰度值大于T的像素设为白色，小于或者等于T的像素设为黑色；或者反过来，将灰度值大于T的像素设为黑色，小于或者等于T的像素设为白色。
优选的，所述步骤3中的像素面积划定标准：大片的像素面积范围：1500-25000，中片的像素面积范围：600-1500，小片的像素面积范围：100-600，碎片的像素面积范围：50-100。
优选的，所述步骤3得到大片、中片、小片和碎片的面积率包括通过连通域分析法得到大片、中片、小片和碎片的面积，再分别除以风分器风分腔体内的横截面面积得到所述大片、中片、小片和碎片的面积率。
优选地，将本发明第一方面测得的叶片结构用在本发明第二方面所述的基于叶片结构测量风分器出叶量的方法中。
本发明的有益效果为：
1、提出了一种基于机器视觉的打叶复烤风分过程叶片形态分割及统计方法，能够得到并分析风分器内风分状态视频图像，从而可以迅速提取出相应的占空比数据、叶片结构数据，为实现打叶工艺过程质量的反馈控制、实时评价，检测方法提供支撑。
2、本发明通过机器视觉和算法即时测量叶片结构即大片、中片、小片和碎片的占比，来通过自创函数找到了测算风分器出叶量的方法，不仅简单实用，而且具有即时性。此外，本发明突破了传统认知，将叶片结构这种相对比例与出叶量直接用函数关联起来，且定量精准，从本发明的函数关系式可见，其中完全没有进叶量A这一项，而是仅以叶片面积率作为自变量，因此可以通过叶片面积率就能直接测量出叶量，通过叶片的结构测量风分器的出叶量，给生产监管提供了新的角度，此外，也省略了用筛分称重法对每一批次原料测量叶片占比的麻烦。
附图说明
图1是无叶片时的背景图。
图2是有叶片时的实时叶片图。
图3是背景图与实时叶片图的差值图像。
图4是基于全局固定灰度阈值的叶片二值图。
图5是超大片叶片的像素图。
图6是大片叶片的像素图。
图7是中片叶片的像素图。
图8是小片叶片的像素图。
图9是碎片叶片的像素图。
图10是机器视觉装置的示意图。
其中：1-风分器、2-LED灯、3-工业相机、4-电控柜。
具体实施方式
为了使本发明的目的、技术方案和有益效果更加清楚，下面将对本发明的优选实施例进行详细的说明，以方便技术人员理解。
实施例1
本实施例涉及一种基于机器视觉的打叶复烤风分过程叶片结构测量方法，包括：
步骤1:通过机器视觉成像系统获取风分器内风分状态下的无叶片的背景图像与有叶片的实时图像。机器视觉成像系统包括风分器、相机，相机支架，相机安装在相机支架上用以对风分器内部拍照以采集图像。图1是无叶片时的背景图。图2是有叶片时的实时叶片图，相机采集图像的时间间隔是50ms。
步骤2:对所述的背景图像和实时图像分别进行相同的去噪处理和图像增强处理，对处理后的图像进行做差运算，得到差值图像。其中，去噪处理是通过高斯滤波算法，图像增强处理是采用图像增强算法来实现，经过这两步处理后，图像特征信息会增强。然后将处理后的背景图像和实时图像做差，得到差值图像，如图3。
步骤3:对差值图像进行灰度化，再运用全局固定叶片灰度阈值分割的threshold方法对叶片图像进行分割，获取叶片的二值化图像。全局固定叶片灰度阈值分割的threshold方法包括：设定图像提取的灰度区间[Tmin,Tmax],在该区间内设置一个阈值T，且Tmin＜T＜Tmax，将灰度值大于T的像素设为白色，小于或者等于T的像素设为黑色；或者反过来，将灰度值大于T的像素设为黑色，小于或者等于T的像素设为白色。经过此步处理，最后得到叶片的二值化图像，如图4。
步骤4:按照预设的超大片、大片、中片、小片和碎片的像素面积划分标准，采用连通域分析法统计超大片、大片、中片、小片和碎片的像素面积。其中，超大片的像素面积范围：25000-500000，大片的像素面积范围：1500-25000，中片的像素面积范围：600-1500，小片的像素面积范围：100-600，碎片的像素面积范围：50-100。其中，图5-9分别展示了超大片、 大片、中片、小片和碎片的像素图。
步骤5:通过对像素面积的计算来提取叶片占空比数据和叶片结构分布数据。其中，通过计算图像上的叶片面积与单张图像总面积的比值来获得叶片占空比数据；通过求算某一片型像素面积占所有片型像素面积总和的百分比来获得叶片结构分布数据。
经过步骤1-5处理之后，本机器视觉成像系统平台能够得到并分析风分器内风分状态视频图像，从而可以迅速提取出相应的占空比数据、叶片结构数据，为实现打叶工艺过程质量的反馈控制、实时评价，检测方法提供支撑。
实施例2
本实施例涉及一种根据叶片结构测量风分器出叶量的方法，包括：
步骤1:通过机器视觉成像系统获取风分器内风分状态下的有叶片的实时图像。机器视觉成像系统具体来说在风分器的上风分仓的玻璃窗口外部搭建视觉成像平台，包括一个工业相机和两个LED灯，分别用于对上风分仓内部高速运动的烟叶实时成像和对风分仓内部的照明。用于图像采集的工业相机位于风分器上风分仓的透明有机玻璃的正前方，相机放置距离需要根据相机的视场范围决定，要保证画面覆盖完整的上风分仓，为了兼顾图像传输和图像处理的实时性，选取的相机分辨率为1280×1024，拍摄的时间间隔为50ms，图10的2表示LED灯，分别位于相机的左右两侧，电控柜在风分器的一旁，电控柜内包括用于数据传输和图像处理的工控机以及用于图像显示的液晶显示器，电柜分别连接相机和LED光源。
步骤2:对实时图像进行处理。方法为：
1：对彩色图像进行灰度化。
由于烟叶的颜色偏向黄色，因此RGB三通道中，对于R通道最敏感，将烟叶图像中的R通道提取出来用于后续的分析。
2：对灰度化的图像进行二值化。
设定图像提取的灰度区间[Tmin,Tmax],在该区间内设置一个阈值T，且Tmin＜T＜Tmax，将灰度值大于T的像素设为白色，小于或者等于T的像素设为黑色；或者反过来，将灰度值大于T的像素设为黑色，小于或者等于T的像素设为白色。
3：最后进行去噪处理。
由于风分仓长时间的运行，四周的仓体和透明玻璃上会留下灰尘，因此还需要对采集的图片进行去噪处理，对灰尘等一些微小颗粒进行去除。本发明采用的方法是图像进行开运算进行去噪：
Figure PCTCN2022098572-appb-000001
其中,S(x,y)表示开运算使用的结构元素，本发明使用的结构元素是3×3大小的圆形结构，去噪完成后，即可得到最终处理好的结果，如图4所示。
步骤3:按照预设的大片、中片、小片和碎片的像素面积划分标准，在处理后的图像上得到大片、中片、小片和碎片的面积率。用连通域分析法得到大片、中片、小片和碎片的面积。像素面积划定标准：大片的像素面积范围：1500-25000，中片的像素面积范围：600-1500，小片的像素面积范围：100-600，碎片的像素面积范围：50-100。由于风分器上风分仓的横截面面积是已知的，将大片、中片、小片和碎片的面积除以横截面面积即得到对应的面积率。图6-9分别展示了超大片、大片、中片、小片和碎片的像素图。
步骤4:依据预设函数关系式测量风分器的出叶量。函数关系式为：
出叶量＝1776.44-5.77*X 1-185.71*X 4-0.65*X 2*X 2-0.30*X 3*X 3+56.44*X 4*X 4，其中X 1是大片面积率，X 2是中片面积率％，X 3是小片面积率％，X 4是碎片面积率％。
经过对函数的实际检测得知：训练样本最大相对误差％＝7.46％，训练样本最小相对误差％＝5.73％，训练样本平均相对误差％＝1.98％，本检测采用多因子及平方项逐步回归方法建立的模型较稳定，模型的平均相对误差在5.2％以内波动。
最后说明的是，以上优选实施例仅用于说明本发明的技术方案而非限制，尽管通过上述优选实施例已经对本实用新型进行了详细的描述，但本领域技术人员应当理解，可以在形式上和细节上对其作出各种各样的改变，而不偏离本实用新型权利要求书所限定的范围

Claims (10)

  1. 一种基于机器视觉的打叶复烤风分过程叶片结构测量方法，其特征在于：包括：
    步骤1:通过机器视觉成像系统获取风分器内风分状态下的无叶片的背景图像与有叶片的实时图像；
    步骤2:对所述的背景图像和实时图像分别进行相同的去噪处理和图像增强处理，对处理后的图像进行做差运算，得到差值图像；
    步骤3:对差值图像进行灰度化，再运用全局固定叶片灰度阈值分割的threshold方法对叶片图像进行分割，获取叶片的二值化图像；
    步骤4:按照预设的超大片、大片、中片、小片和碎片的像素面积划分标准，采用连通域分析法统计超大片、大片、中片、小片和碎片的像素面积。
    步骤5:通过对像素面积的计算来提取叶片占空比数据和叶片结构分布数据。
  2. 根据权利要求1所述的基于机器视觉的打叶复烤风分过程叶片结构测量方法，其特征在于：所述步骤1中的机器视觉成像系统包括风分器、相机，相机支架，所述相机安装在相机支架上用以对风分器内部拍照以采集图像。
  3. 根据权利要求1所述的基于机器视觉的打叶复烤风分过程叶片结构测量方法，其特征在于：所述全局固定叶片灰度阈值分割的threshold方法包括：设定图像提取的灰度区间[Tmin,Tmax],在该区间内设置一个阈值T，且Tmin＜T＜Tmax，将灰度值大于T的像素设为白色，小于或者等于T的像素设为黑色；或者反过来，将灰度值大于T的像素设为黑色，小于或者等于T的像素设为白色。
  4. 根据权利要求1所述的基于机器视觉的打叶复烤风分过程叶片结构测量方法，其特征在于：所述像素面积划定标准：超大片的像素面积范围：25000-500000，大片的像素面积范围：1500-25000，中片的像素面积范围：600-1500，小片的像素面积范围：100-600，碎片的像素面积范围：50-100。
  5. 根据权利要求1所述的基于机器视觉的打叶复烤风分过程叶片结构测量方法，其特征在于：通过计算图像上的叶片面积与单张图像总面积的比值来获得叶片占空比数据。
  6. 根据权利要求1所述的基于机器视觉的打叶复烤风分过程叶片结构测量方法，其特征在于：通过求算某一片型像素面积占所有片型像素面积总和的百分比来获得叶片结构分布数据。
  7. 一种根据叶片结构测量风分器出叶量的方法，其特征在于：包括：
    步骤1:通过机器视觉成像系统获取风分器内风分状态下的有叶片的实时图像；
    步骤2:对实时图像进行处理；
    步骤3:按照预设的大片、中片、小片和碎片的像素面积划分标准，在处理后的图像上得到大片、中片、小片和碎片的面积率；
    步骤4:依据以下预设函数关系式测量风分器的出叶量：
    出叶量＝1776.44-5.77*X 1-185.71*X 4-0.65*X 2*X 2-0.30*X 3*X 3+56.44*X 4*X 4，其中X 1是大片面积率，X 2是中片面积率％，X 3是小片面积率％，X 4是碎片面积率％。
  8. 根据权利要求7所述的测量风分器出叶量的方法，其特征在于：所述步骤2对图像进行处理包括灰度化、二值化、去噪；其中所述二值化包括：设定图像提取的灰度区间[Tmin,Tmax],在该区间内设置一个阈值T，且Tmin＜T＜Tmax，将灰度值大于T的像素设为白色，小于或者等于T的像素设为黑色；或者反过来，将灰度值大于T的像素设为黑色，小于或者等于T的像素设为白色。
  9. 根据权利要求7所述的测量风分器出叶量的方法，其特征在于：所述步骤3中的像素面积划定标准：大片的像素面积范围：1500-25000，中片的像素面积范围：600-1500，小片的像素面积范围：100-600，碎片的像素面积范围：50-100。
  10. 根据权利要求1所述的测量风分器出叶量的方法，其特征在于：所述步骤3得到大片、中片、小片和碎片的面积率包括通过连通域分析法得到大片、中片、小片和碎片的面积，再分别除以风分器风分腔体内的横截面面积得到所述大片、中片、小片和碎片的面积率。
PCT/CN2022/098572 2022-05-23 2022-06-14 一种打叶复烤风分过程中叶片结构测量方法及风分器出叶量测量方法 WO2023226103A1 (zh)

Priority Applications (1)

Application Number Priority Date Filing Date Title
EP22859572.4A EP4307225A1 (en) 2022-05-23 2022-06-14 Leaf structure measurement method during threshing, redrying and air separation process and leaf outlet amount measurement method of air separator

Applications Claiming Priority (4)

Application Number Priority Date Filing Date Title
CN202210560853.5A CN115018861A (zh) 2022-05-23 2022-05-23 一种基于机器视觉的打叶复烤风分过程叶片形态分割及统计方法
CN202210560853.5 2022-05-23
CN202210561024.9A CN115035048A (zh) 2022-05-23 2022-05-23 一种根据叶片结构测量风分器出叶量的方法
CN202210561024.9 2022-05-23

Publications (1)

Publication Number Publication Date
WO2023226103A1 true WO2023226103A1 (zh) 2023-11-30

Family

ID=87863244

Family Applications (1)

Application Number Title Priority Date Filing Date
PCT/CN2022/098572 WO2023226103A1 (zh) 2022-05-23 2022-06-14 一种打叶复烤风分过程中叶片结构测量方法及风分器出叶量测量方法

Country Status (2)

Country Link
EP (1) EP4307225A1 (zh)
WO (1) WO2023226103A1 (zh)

Cited By (1)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN118037821A (zh) * 2024-03-14 2024-05-14 湖北麦麦农业科技有限公司 一种生菜叶片面积检测方法及系统

Citations (6)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN104914105A (zh) * 2015-06-09 2015-09-16 中国农业科学院茶叶研究所 一种基于图像识别技术的茶叶等级鉴定方法
CN106446968A (zh) * 2016-11-22 2017-02-22 湖北民族学院 基于fpga的烟叶自动分级系统
CN109829943A (zh) * 2018-11-13 2019-05-31 上海烟草集团有限责任公司 基于机器视觉的叶片结构检测方法、系统、介质及设备
CN112700488A (zh) * 2020-12-11 2021-04-23 浙江托普云农科技股份有限公司 基于图像拼接的活体长叶片叶面积分析方法、系统及装置
WO2022049549A1 (en) * 2020-09-04 2022-03-10 Itc Limited Artificial intelligence based tobacco particle measurement system
CN114304698A (zh) * 2022-01-24 2022-04-12 红云红河烟草(集团)有限责任公司 一种打叶复烤分切打叶物料混合的动态比例掺配方法

Patent Citations (6)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN104914105A (zh) * 2015-06-09 2015-09-16 中国农业科学院茶叶研究所 一种基于图像识别技术的茶叶等级鉴定方法
CN106446968A (zh) * 2016-11-22 2017-02-22 湖北民族学院 基于fpga的烟叶自动分级系统
CN109829943A (zh) * 2018-11-13 2019-05-31 上海烟草集团有限责任公司 基于机器视觉的叶片结构检测方法、系统、介质及设备
WO2022049549A1 (en) * 2020-09-04 2022-03-10 Itc Limited Artificial intelligence based tobacco particle measurement system
CN112700488A (zh) * 2020-12-11 2021-04-23 浙江托普云农科技股份有限公司 基于图像拼接的活体长叶片叶面积分析方法、系统及装置
CN114304698A (zh) * 2022-01-24 2022-04-12 红云红河烟草(集团)有限责任公司 一种打叶复烤分切打叶物料混合的动态比例掺配方法

Non-Patent Citations (1)

* Cited by examiner, † Cited by third party
Title
CHENGUANG ZHANG, TANG XIANGYANG;CHI YILIN: "Machine Vision Test System on Stripped Tobacco Structure", MECHATRONICS, vol. 14, no. 11, 20 November 2008 (2008-11-20), pages 82 - 84, 92, XP093112500 *

Cited By (1)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN118037821A (zh) * 2024-03-14 2024-05-14 湖北麦麦农业科技有限公司 一种生菜叶片面积检测方法及系统

Also Published As

Publication number Publication date
EP4307225A1 (en) 2024-01-17

Similar Documents

Publication Publication Date Title
CN106780537B (zh) 一种基于图像处理的方格蔟蚕茧筛选装置及方法
CN106918594B (zh) 一种在线分析籽棉色特征和杂质的方法
CN101944231B (zh) 小麦穗部形态参数提取方法
CN104198324B (zh) 基于计算机视觉的烟丝中叶丝比例测定方法
CN104297160A (zh) 一种工夫红茶发酵适度判别方法和装置
CN103076288B (zh) 一种基于计算机视觉的鱼肉自动分级装置和方法
CN102508110A (zh) 基于纹理的绝缘子缺陷诊断方法
WO2023226103A1 (zh) 一种打叶复烤风分过程中叶片结构测量方法及风分器出叶量测量方法
CN104256882A (zh) 基于计算机视觉的烟丝中再造烟叶比例测定方法
CN104198325A (zh) 基于计算机视觉的烟丝中梗丝比例测定方法
CN107490583A (zh) 一种基于机器视觉的夹片缺陷检测方法
CN108198176A (zh) 一种基于图像分析烟草成熟度的判别方法
CN110310275A (zh) 一种基于图像处理的链式传送带缺陷检测方法
CN109472779A (zh) 一种基于形态结构的纱线外观特征参数提取与分析方法
CN108007945A (zh) 一种基于x射线透射图像定量检测烟梗中粗梗率和长短梗率的测定方法
CN112505039B (zh) 一种烟用梗丝复切理论成丝效果评价方法
CN115018722A (zh) 一种基于风分过程稳态占空比预测风分器瞬时出叶量的方法
CN112213320A (zh) 一种基于机器视觉的饼干检测装置及饼干检测方法
CN104778709B (zh) 一种基于纱线序列图像的电子黑板的构建方法
CN106530248A (zh) 一种变电站场景视频噪声的智能检测方法
CN111199344B (zh) 一种控制烟叶分选过程中纯度的方法
CN213275384U (zh) 一种基于机器视觉的饼干检测装置
CN115015245A (zh) 一种根据风分器内动态大片面积率来测量风分器瞬时出叶量的方法
CN104198491B (zh) 基于计算机视觉的烟丝中膨胀叶丝比例测定方法
CN114486768A (zh) 用于籽棉收购环节的籽棉颜色级检测装置及其分级方法

Legal Events

Date Code Title Description
ENP Entry into the national phase

Ref document number: 2022859572

Country of ref document: EP

Effective date: 20230303