Method for evaluating the quality of tobacco cut surface compacted by cutter

Abstract

The present invention provides a method for evaluating the quality of a tobacco section compacted by a shredder, comprising: 1) acquiring a color image and a depth image of the tobacco section compacted by the shredder; 2) performing field of view registration on the color image and the depth image to obtain a color point cloud image (RGB-D); 3) segmenting the color image to acquire a tobacco image; 4) extracting data representing image texture and data representing image smoothness from the tobacco image; 5) performing image matching on the tobacco image and a standard non-defective tobacco image to locate defect information in the tobacco image; 6) extracting data representing defect depth from the concave information; and 7) evaluating the quality of the tobacco section compacted by the shredder according to the data representing image texture, the data representing image smoothness and the data representing defect depth.

Description

Technical Field

[0001] The invention belongs to the field of tobacco, and in particular relates to a method for evaluating the quality of a cut surface of tobacco compacted by a shredder. Background Art

[0002] Tobacco cutter is one of the key main equipment for tobacco industry to make tobacco wire. Its main function is to cut tobacco into wire. Cutters can be divided into leaf cutters and stem cutters. They have similar structural principles and are used for cutting tobacco leaves and stems respectively.

[0003] For tobacco stems, the compacted tobacco section at the cutter's blade has a very important influence on the tobacco stem shredding effect. Compared with leaf blades, tobacco stems are affected by their own morphology and mechanical properties, and are not easy to compact at the blade, showing several depressions on the compacted tobacco section. In the depressions and their surroundings on the section, the mutual clamping force between the tobacco stems is insufficient. Under the action of the blades of the cutting system, the tobacco stems in the depressions and their surroundings are prone to displacement, resulting in inconsistent thickness of the tobacco stems, and even causing the tobacco stems to fall off, affecting the thickness uniformity of the tobacco stem shredding. Therefore, it is necessary to evaluate the compacted tobacco section of the cut tobacco shredding to predict the shredding effect of the stem cutter and make corresponding adjustments.

[0004] Traditionally, the evaluation of the tobacco cut surface compacted by the shredder mainly relies on manual visual inspection, and the evaluation results are only qualitatively described as good, average or poor. The evaluation results lack unified standards and are not scientific.

[0005] Therefore, it is urgent to provide a quantitative evaluation method for the tobacco cut surface compacted by the cutter to increase the scientificity and uniformity of the evaluation and better guide the equipment adjustment of the stem cutter.

[0006] It should be noted that the information disclosed in the background technology section of the present invention is only intended to increase the understanding of the overall background of the present invention, and should not be regarded as acknowledging or suggesting in any form that the information constitutes prior art already known to those skilled in the art. Summary of the invention

[0007] In view of the above problems, the purpose of the present invention is to provide a method for quantitatively evaluating the compacted section of a shredder. The method quantitatively analyzes the actual section, extracts quantitative analysis indicators of the section, and performs weighted calculation on the extracted indicators. The method can be used to quantitatively evaluate the compacted section of a shredder.

[0008] In some aspects, the present disclosure provides a method for evaluating the quality of a cut surface of tobacco compacted by a shredder, characterized in that it comprises:

[0009] 1) Obtaining a color image and a depth image of the tobacco section compacted by the cutter;

[0010] 2) performing field of view registration on the color image and the depth image to obtain a color point cloud image (RGB-D);

[0011] 3) Segment the color image to obtain the tobacco image;

[0012] 4) extracting data representing image texture and data representing image smoothness from the tobacco image;

[0013] 5) matching the tobacco image with a standard non-defective tobacco image to locate defect information in the tobacco image;

[0014] 6) extracting data representing the defect depth from the concave information;

[0015] 7) Evaluating the quality of the tobacco cut surface compacted by the cutter based on the data characterizing the image texture, the data characterizing the image smoothness and the data characterizing the defect depth.

[0016] In some embodiments, step 1) includes: using an RGB-D camera to photograph the cut surface of the tobacco compacted by the cutter to obtain a color image and a depth image.

[0017] In some embodiments, step 2) includes: performing field of view registration based on the acquired color image and depth image, registering the field of view of the depth image to the field of view of the color image, and making the depth information correspond to the color information.

[0018] In some embodiments, step 3) includes: removing non-tobacco images from the registered images, and only retaining tobacco images.

[0019] In some embodiments, in step 4), the indicator characterizing the texture information of the image is the entropy of the image, and the indicator characterizing the smoothness information of the image is the total variation of the image on the vertical axis.

[0020] In some embodiments, step 5) includes: taking a tobacco area without depressions as a standard section, performing image matching between the acquired tobacco image and the standard section, and acquiring depression information of the tobacco area.

[0021] In some embodiments, in step 6), the data characterizing the defect depth includes: an average value of the pit depth, a variance of the pit depth, or a combination thereof.

[0022] In some embodiments, step 7) includes assigning weight coefficients to data characterizing image texture, data characterizing image smoothness, and data characterizing defect depth, and using a weighted calculation method to quantitatively evaluate the quality of the cut surface of tobacco compacted by the cutter.

[0023] In some embodiments, the entropy of an image is calculated by the following formula:

[0024]

[0025] Where H represents the entropy value, and the tobacco area of ​​the tobacco section compacted by the cutter is an image with k depth values, and the probability of each value appearing is p1, p2, ..., p k .

[0026] In some embodiments, the total variation of the image in the longitudinal direction (ie, the depth direction) is calculated by the following formula:

[0027]

[0028] In the formula, x ij The coordinate value data of each point on the image.

[0029] In some embodiments, the average value of the depression is calculated as follows:

[0030]

[0031] In the formula, the tobacco depression information is defined as an image with an M×N resolution, and f(i, j) is the coordinate value data of each pixel point.

[0032] In some embodiments, the calculation formula for calculating the variance of the concavity is as follows:

[0033]

[0034] In the formula, the tobacco depression information is defined as an image with an M×N resolution, f(i,j) is the coordinate value data of each pixel point (i and j are the horizontal and vertical coordinate values ​​respectively), is the average value of the depression.

[0035] Beneficial effects of the present invention:

[0036] One or more technical solutions disclosed herein have one or more of the following beneficial effects:

[0037] 1) This method uses machine vision to sample and quantitatively evaluate the compacted section, thereby replacing manual subjective evaluation, and the quantitative evaluation results are more unified.

[0038] 2) The compacted cross section is used to extract evaluation indicators such as image entropy, total variation, mean value of concavity, and variance of concavity. Compared with manual subjective evaluation, the quantitative evaluation results are more scientific.

[0039] 3) This method is not only used for quantitative evaluation of the compacted section of the wire cutter, but is also applicable to quantitative evaluation of similar planes or curved surfaces with abnormal deformation. BRIEF DESCRIPTION OF THE DRAWINGS

[0040] Figure 1 The structural principle diagram of the stem cutting machine

[0041] Figure 2 A flow chart of a method provided in one embodiment of the present invention

[0042] Figure 3 Original image of the compacted section of the shredder

[0043] Figure 4 A color point cloud image of the compacted section after registration according to an embodiment of the present invention

[0044] Figure 5 A color point cloud image of the tobacco area after image segmentation according to an embodiment of the present invention

[0045] Figure 6 The figure is a flowchart of image registration based on feature points according to an embodiment of the present invention. DETAILED DESCRIPTION

[0046] In order to make the purpose, technical solution and advantages of the present invention clearer, the technical solution of the present invention will be clearly and completely described below in conjunction with the accompanying drawings. Obviously, the described embodiments are part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making creative work are within the scope of protection of the present invention.

[0047] Figure 1 The schematic diagram of the structure of a stem cutting machine is shown. As shown in the figure, tobacco stems 2 enter a wedge-shaped channel formed by an upper copper chain 3 and a lower copper chain 4 through a feeding device 1. The upper copper chain 3 and the lower copper chain 4 move to convey and squeeze the material, so that it gradually forms a compact "tobacco cake" and is sent to a rectangular knife gate 5 at the small end of the wedge-shaped channel, where the tobacco cake forms a compacted cut surface.

[0048] The main motor drives the knife roller 6 to rotate through the transmission belt. The blades on the knife roller 6 are continuously fed quantitatively by the knife pusher device and sharpened by the knife sharpener 7 to form a cutting cylinder with a specified diameter and sharp edges. The knife roller 6 cuts the "tobacco cake" continuously fed from the knife door 5 into tobacco shreds with a width that meets the process requirements and outputs them from the drop hopper 8.

[0049] Figure 2 A flow chart of a method provided by an embodiment of the present invention is shown as follows: Figure 2 As shown, the quantitative evaluation method of the compacted section of a shredder provided in this embodiment comprises the following steps:

[0050] 101 obtaining a color image and a depth image of a cut surface of tobacco compacted by a shredder;

[0051] 102 performing field of view registration on the color image and the depth image to obtain a color point cloud image (RGB-D);

[0052] 103 segmenting the color image to obtain a tobacco image;

[0053] 104 extracting data representing image texture and data representing image smoothness from the tobacco image;

[0054] 105 performing image matching between the tobacco image and a standard non-defective tobacco image to locate defect information in the tobacco image;

[0055] 106 extracting data representing the defect depth from the concave information;

[0056] 107 The quality of the cut surface of the tobacco compacted by the shredder is evaluated according to the data representing the image texture, the data representing the image smoothness and the data representing the defect depth.

[0057] The quantitative evaluation method of the compacted section of a shredder is further described below through specific examples.

[0058] 1. Obtain color image and depth image of the compacted section of the shredder;

[0059] In this embodiment, the color image and depth image of the compacted section of the shredder are obtained by using an RGB-D camera (such as Microsoft's developer toolkit Azure Kinect DK hardware). Specifically:

[0060] 1) During the production process, the blade cover of the stem cutter is manually opened to display the compacted cut surface, such as Figure 3 As shown;

[0061] 2) Use the RGB-D camera to shoot a video of a certain length and obtain the video stream information of the compacted section (usually in .mkv format). The obtained video stream saves multiple frames of color images and corresponding depth images, as well as several camera parameters;

[0062] 3) Since the captured video scene is fixed, only one frame of color image and depth image needs to be extracted from the video stream for processing.

[0063] Second, the registration of the color image and depth image of the compacted section of the shredder;

[0064] According to the acquired RGB image and depth image, the depth image is registered to the field of view of the RGB image, and the depth information and RGB color information are matched. Then a color point cloud image is obtained, such as Figure 4 .

[0065] The conversion of color images and depth images into point clouds is a transformation of coordinate systems: image coordinate system → world coordinate system. The constraints of the transformation are the internal parameters of the camera, which can be obtained by reading the configuration information of the video. The conversion formula is as follows:

[0066]

[0067] Among them, x, y, z are the point cloud coordinate system, x', y' are the image coordinate system, and D is the depth value.

[0068] 3. Perform image segmentation on the color image to obtain the tobacco area section;

[0069] Since the compacted section area obtained in step 1 contains a large number of background pixels, such as knife gate information, in order to facilitate subsequent processing, the tobacco area needs to be located, that is, the image including only the tobacco area needs to be segmented, such as Figure 5 .

[0070] In this embodiment, there is an obvious color difference between the tobacco area and the non-tobacco area (such as the knife gate, etc.), so the GrabCut algorithm in the traditional learning method is used to utilize the color and boundary information in the image, and a relatively good segmentation effect can be obtained with only a small amount of user interaction.

[0071] The following is the execution flow of the GrabCut algorithm in this embodiment.

[0072] 1) Interactive input. The operator directly selects the tobacco stem section to provide an initial operating condition for the algorithm. The algorithm will treat all pixels outside the box as background pixels, and all pixels inside the box as pixels that may be tobacco areas.

[0073] 2) Assign the pixels that may be tobacco areas to the corresponding initialization label an, obtain some pixels belonging to the foreground (an=1), and the rest are pixels belonging to the background (an=0).

[0074] 3) Use a Gaussian mixture model (GMM) to model the foreground and background.

[0075] 4) Based on the input, GMM learns and creates a new pixel distribution. The pixels inside the rectangular box (which may be foreground or background) can be classified according to their relationship with the pixels outside the box (background).

[0076] 5) Create an image based on the distribution of pixels. The nodes in the image are pixels. In addition to the pixels as nodes, there are two other nodes: Source_node and Sink_node. All foreground pixels are connected to Source_node. All background pixels are connected to Sink_node.

[0077] 6) The weight of the edge connecting a pixel to Source_node / end_node is determined by the probability that they belong to the same class. The weight between two pixels is determined by the information of the edge or the similarity of the two pixels. If the colors of two pixels are very different, the weight of the edge between them will be small.

[0078] 7) Use mincut algorithm to segment the above image. It will divide the image into Source_node and Sink_node according to the minimum cost equation. The cost equation is the sum of the weights of all the edges that are cut. After cutting, all pixels connected to Source_node are considered foreground, and all pixels connected to Sink_node are considered background.

[0079] 8) Continue this process until the classification converges.

[0080] 4. Extracting data representing image texture and data representing image smoothness from tobacco images;

[0081] For the tobacco area section, the amount of information that the image can convey can be expressed by the entropy of the image, which reflects the richness of information such as the texture of the image; the total variation of the image in the vertical direction represents the smoothness of the tobacco area section, specifically:

[0082] 1) Calculation of image entropy

[0083] The calculation formula is as follows:

[0084]

[0085] In the formula, the tobacco area of ​​the compacted cut surface of the cutter is defined as an image with k depth values, and the probability of each value appearing is p1, p2, ..., p k .

[0086] 2) Calculation of total variation of images

[0087] The calculation formula is as follows:

[0088]

[0089] In the formula, x ij The coordinate data of each point on the image

[0090] 5. Match the tobacco image with a standard defect-free tobacco image to locate defect information in the tobacco image.

[0091] The image matching process is as follows Figure 6 As shown, the step includes:

[0092] Input the image to be registered and perform preprocessing 601;

[0093] Extract feature points 602 from the simple processed graphics obtained by preprocessing;

[0094] Perform feature point matching 603 on a feature point set obtained by extracting feature points;

[0095] Calculate the transformation model of N pairs of corresponding feature points obtained by feature point matching 604;

[0096] Resampling 605 the spatial transformation of the two images calculated by the transformation model;

[0097] Output the registered image after resampling.

[0098] Image matching can be defined as the position transformation and grayscale transformation of two images, that is, mapping the pixel coordinate X = (x, y) of one image to a coordinate X^' = (x^', y^') in another new coordinate system, and then resampling the pixel values ​​of the image.

[0099] In this embodiment, a tobacco region without depressions is used as a standard section, and the obtained tobacco region section is image matched with the standard section to obtain the depression information of the tobacco region.

[0100] There are always unique pixels in an image, which can be considered as the features of the image and called feature points. These feature points can be used for image feature matching. Each key point is represented by a descriptor (a feature vector containing the essential characteristics of the key point). The descriptor should be robust to image transformations (such as position transformation, scaling transformation, brightness transformation, etc.).

[0101] There are four main steps in feature point-based image matching: extraction of feature points, correspondence of points with the same name (i.e., matching of feature points), calculation of image transformation model, and image resampling.

[0102] 6. extracting data representing the depth of the defect from the concave information;

[0103] For the depression information of the tobacco area after matching with the standard section, the depression degree can be expressed by the average depression value, which is used to reflect the mean depression depth of the tobacco area of ​​the compacted section of the shredder; the variance of the depression is used to reflect the degree of dispersion of the depression depth of the tobacco area of ​​the compacted section of the shredder and the mean, that is, the data fluctuation range of the depression depth value, specifically:

[0104] 1) Calculation of the average value of the depression

[0105] The calculation formula is as follows:

[0106]

[0107] In the formula, the tobacco depression information is defined as an image with M×N resolution, and f(i,j) is the coordinate value data of each pixel point

[0108] 2) Calculation of the variance of the concavity

[0109] The calculation formula is as follows:

[0110]

[0111] In the formula, the tobacco depression information is defined as an image with M×N resolution, f(i,j) is the coordinate value data of each pixel point, is the average value of the depression.

[0112] 7. Evaluate the quality of the tobacco cut surface compacted by the shredder based on the data characterizing the image texture, the data characterizing the image smoothness and the data characterizing the defect depth.

[0113] In this embodiment, step 4 obtains the image entropy H and total variation And other evaluation indicators, step six obtained the average value of the concavity Based on these indicators, the quantitative evaluation of the compacted section of the wire cutter can be carried out according to actual needs; different weighted ratios can also be set for different parameters to conduct comprehensive calculation and evaluation.

[0114] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, rather than to limit it. Although the present invention has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that they can still modify the technical solutions described in the aforementioned embodiments, or make equivalent replacements for some of the technical features therein. However, these modifications or replacements do not deviate the essence of the corresponding technical solutions from the spirit and scope of the technical solutions of the embodiments of the present invention.

Claims

1. A method for evaluating the quality of the cut surface of tobacco compacted by a shredder, characterized in that: include 1) Obtaining a color image and a depth image of the tobacco section compacted by the cutter; 2) performing field of view registration on the color image and the depth image to obtain a color point cloud image; 3) Segment the color point cloud image to obtain the tobacco image; 4) extracting data representing image texture and data representing image smoothness from the tobacco image; wherein the index representing the image texture information is the entropy of the image, and the index representing the image smoothness information is the total variation of the image on the vertical axis; 5) Taking the tobacco area without depression as the standard section, matching the acquired tobacco image with the standard section to obtain depression information of the tobacco area; 6) extracting data representing the defect depth from the concave information, wherein the data representing the defect depth includes: an average value of the concave depth, a variance of the concave depth, or a combination thereof; 7) Evaluating the quality of the tobacco cut surface compacted by the cutter based on the data characterizing the image texture, the data characterizing the image smoothness and the data characterizing the defect depth.

2. The method according to claim 1, characterized in that Step 1) includes: using an RGB-D camera to photograph the cut surface of the tobacco compacted by the shredder to obtain a color image and a depth image.

3. The method according to claim 2, characterized in that Step 2) includes: performing field of view registration according to the acquired color image and depth image, registering the field of view of the depth image to the field of view of the color image, and making the depth information correspond to the color information.

4. The method according to claim 3, characterized in that Step 3) includes: removing non-tobacco images from the registered images and retaining only tobacco images.

5. The method according to claim 1, characterized in that Step 7) includes assigning weight coefficients to the data representing the image texture, the data representing the image smoothness, and the data representing the defect depth, and using a weighted calculation method to quantitatively evaluate the quality of the tobacco cut surface compacted by the shredder.

6. The method according to claim 1, characterized in that The entropy of an image is calculated using the following formula: In the formula, H represents the entropy value, the tobacco area of ​​the tobacco section compacted by the cutter is an image with k depth values, and the probability of each value appearing is , , ..., .

7. The method according to claim 1, characterized in that The total variation of the image in the vertical direction is calculated by the following formula: In the formula, The coordinate value data of each point on the image.

8. The method according to claim 1, characterized in that The calculation formula for calculating the average value of the depression depth is as follows: In the formula, the tobacco depression information is defined as an image with M×N resolution, The coordinate value data for each pixel point.

9. The method according to claim 1, characterized in that: The calculation formula for calculating the variance of the depression depth is as follows: In the formula, the tobacco depression information is defined as an image with M×N resolution, For each pixel point’s coordinate value data, is the average value of the depression depth.

Images