Cigarette end face defect detection method and detection system based on three-dimensional data

Abstract

The invention discloses a cigarette end surface defect detection method and detection system based on three-dimensional data, and the main design concept is to provide an end surface defect identification and measurement improvement solution based on a three-dimensional point cloud processing technology for cigarette end surface defect on-line real-time detection demand and a field production scene. By performing targeted preprocessing on the three-dimensional point cloud data of the end face of the cigarette and adopting geometric feature extraction and point cloud segmentation methods, defects such as loose ends, loose ends and sinking on the end face of the cigarette can be efficiently and accurately identified and positioned. Compared with a traditional detection method based on a two-dimensional image, the depth information of the three-dimensional point cloud data is effectively utilized, the three-dimensional features of the end face defects can be reflected more accurately, and effective recognition of the tiny defects is ensured; in addition, through optimization processing, noise suppression and the like of the point cloud data, the detection precision and stability can be further improved, and the requirements of a production line for efficient and real-time detection of the end face defects of the cigarettes are met.

Description

Technical Field

[0001] The present invention relates to the field of cigarette product detection, and particularly to a method and a detection system for detecting defects on the end face of cigarette sticks based on three-dimensional data. Background Art

[0002] Traditional methods for detecting defects on the end face of cigarette sticks usually rely on two-dimensional detection techniques based on images. By capturing images of the end face of cigarette sticks and extracting features, combined with classical machine learning algorithms such as support vector machines (SVM) or decision trees, etc., to identify defects. However, due to the influence of factors such as illumination, morphological changes, and viewing angles on the end face of cigarette sticks, with the continuous improvement of the quality requirements for cigarette production, product quality is gradually developing towards refined and intelligent control. Traditional two-dimensional detection methods based on images have great limitations when dealing with complex defect types, especially when it is difficult to accurately capture and distinguish small or irregular defects. For example: Among the appearance quality requirements for the end face of cigarettes, there are two types of defects, namely "hollow end" and "sinking". The "hollow end" is defined as "the end face of the cigarette stick should not be loose, that is, there should not be both a hollow depth > 1.0 mm and a hollow cross-sectional ratio > 2 / 3 on its end face", and the "sinking" is defined as "the overall sinking of the end face of the cigarette stick with a depth ≥ 0.5 mm but < 1.0 mm". According to the definition, there are differences between the two types of defects, namely "hollow end" and "sinking". However, in the on-line quality inspection and control during the production process, cigarette sticks with the two types of defects, namely "hollow end" and "sinking", are usually removed as "hollow end". However, the causes of the two types of defects are different, and it is very difficult for traditional detection methods to distinguish them, thus affecting the efficiency of cigarette quality control and the consumption control level during the production process. In addition, fluctuations in image quality and background noise will also lead to a decrease in the accuracy of defect recognition, affecting the stability and reliability of detection.

[0003] Specifically, on the one hand, due to the three-dimensional sense and depth changes of the shape and defects of cigarette sticks, two-dimensional images cannot fully reflect these features, resulting in a significant reduction in detection accuracy in complex environments; on the other hand, traditional image processing methods usually cannot effectively cope with challenges such as uneven illumination and complex textures on the surface of cigarette sticks. Therefore, the present invention believes that to solve such problems, more depth information needs to be introduced to more accurately identify and locate end face defects from a three-dimensional perspective. Summary of the Invention

[0004] In view of the above, the present invention aims to provide a method and a detection system for detecting defects on the end face of cigarette sticks based on three-dimensional data to solve the aforementioned technical problems.

[0005] The technical solution adopted by the present invention is as follows:

[0006] In the first aspect, the present invention provides a method for detecting defects on the end face of cigarette sticks based on three-dimensional data, which includes:

[0007] Obtain the point cloud data of the end face of the cigarette rod during the movement of the cigarette rod;

[0008] Preprocess the point cloud data;

[0009] Segment the preprocessed point cloud data based on geometric features, and extract the to-be-determined defect area of the end face of the cigarette rod;

[0010] Perform quantitative analysis on the to-be-determined defect area of the end face of the cigarette rod to obtain quantitative defect features;

[0011] Use the quantitative defect features to determine whether it belongs to the target area with real defects.

[0012] In at least one possible implementation manner, the obtaining of the quantitative defect features includes calculating the vertical height difference of each point in the to-be-determined defect area, and obtaining the defect depth value based on the vertical height difference, at least including: taking the maximum vertical height difference among all points in the same to-be-determined defect area as the defect depth value.

[0013] In at least one possible implementation manner, the obtaining of the quantitative defect features further includes:

[0014] Calculate the projected area of the point cloud in the to-be-determined defect area to obtain the defect two-dimensional feature;

[0015] Alternatively, combine the convex hull algorithm to extract the boundary of the to-be-determined defect area, and use the boundary to obtain the area of the to-be-determined defect area.

[0016] In at least one possible implementation manner, the determining whether it belongs to the target area with real defects includes:

[0017] Use the threshold segmentation method to discriminate the quantitative defect features of each single point in the to-be-determined defect area, and based on the statistical results after discrimination, determine whether the current to-be-determined defect area is the target area;

[0018] Alternatively, directly discriminate whether the current to-be-determined defect area is the target area based on the relationship between the maximum vertical height difference and the preset corresponding threshold.

[0019] In at least one possible implementation manner, the preprocessing includes:

[0020] Use a filtering algorithm to remove the noise in the point cloud data;

[0021] Perform downsampling on the noise-reduced point cloud data;

[0022] Perform registration and alignment processing on the downsampled point cloud data;

[0023] Also, when the cigarette end faces are collected by multiple devices or at different times, the point cloud data after registration and alignment processing is normalized.

[0024] In at least one possible implementation manner, the downsampling includes: reducing the amount of data by using a Voxel grid; and the registration and alignment processing includes: performing the registration and alignment processing by using an ICP algorithm.

[0025] In a second aspect, the present invention provides a cigarette end surface defect detection system based on three-dimensional data, comprising:

[0026] Point cloud acquisition module, used to obtain point cloud data of the end face of the cigarette while it is moving;

[0027] Point cloud preprocessing module, used to preprocess point cloud data;

[0028] The defect area extraction module is used to segment the pre-processed point cloud data based on geometric features and extract the pending defect area of the cigarette end face;

[0029] The feature quantification module is used to perform quantitative analysis on the undetermined defect area of the cigarette end face to obtain quantitative defect features;

[0030] The target defect determination module is used to determine whether the target area is a target area with real defects by using quantitative defect features.

[0031] Compared with the prior art, the main design concept of the present invention is to propose an improved solution for end face defect identification and measurement based on three-dimensional point cloud processing technology in response to the needs of online real-time detection of cigarette end face defects and on-site production scenarios. By performing targeted preprocessing on the three-dimensional point cloud data of the cigarette end face, and using geometric feature extraction and point cloud segmentation methods, it is possible to efficiently and accurately identify and locate defects such as empty heads, thin heads, and sunken ends on the cigarette end faces. Compared with traditional detection methods based on two-dimensional images, the present invention effectively utilizes the depth information of three-dimensional point cloud data, can more accurately reflect the three-dimensional features of end face defects, and ensure the effective identification of minor defects; in addition, through the optimization processing and noise suppression of point cloud data, the detection accuracy and stability can be further improved, meeting the production line's needs for efficient and real-time detection of cigarette end face defects. BRIEF DESCRIPTION OF THE DRAWINGS

[0032] In order to make the purpose, technical solutions and advantages of the present invention more clear, the present invention will be further described below with reference to the accompanying drawings, in which:

[0033] Figure 1 A schematic diagram of a cigarette end face defect detection method based on three-dimensional data provided by an embodiment of the present invention;

[0034] Figure 2Schematic diagram of a cigarette end defect detection system based on three-dimensional data provided by an embodiment of the present invention. Detailed implementation manners

[0035] The embodiments of the present invention will be described in detail below. The examples of the embodiments are shown in the accompanying drawings, where the same or similar reference numerals denote the same or similar elements or elements with the same or similar functions throughout. The embodiments described below by referring to the accompanying drawings are exemplary and are only used to explain the present invention, and should not be construed as limiting the present invention.

[0036] An embodiment of a method for detecting cigarette end defects based on three-dimensional data is proposed by the present invention. Specifically, as Figure 1 shown, which includes:

[0037] Step S1: Obtain the point cloud data of the cigarette end during the movement of the cigarette.

[0038] In the application of detecting defects on the end face of a cigarette (which may include the end face of the tobacco section or the filter section), data acquisition and preprocessing are crucial steps. To ensure the accuracy and high quality of the point cloud data, a high-precision 3D scanner or laser scanning device needs to be used to scan the end face of the cigarette to obtain 3D point cloud data with rich spatial information. Through these devices, the details of the cigarette end face can be captured, including surface irregularities and possible defects.

[0039] Laser scanning (line scan laser device) emits laser pulses and receives the signals reflected from the cigarette surface, calculates the time difference of the reflected signals, and then determines the 3D position and shape of the cigarette. These data form a dense point cloud, which can accurately and comprehensively represent the end face characteristics and contours of the cigarette. In actual operation, the scanning frequency and laser pulse frequency determine the accuracy and density of the point cloud. Especially on a high-speed conveyor belt (moving platform), a higher scanning frequency and pulse rate ensure that a fast-moving cigarette can be captured, meeting the requirement of online acquisition of the point cloud data of the cigarette end face.

[0040] Step S2: Preprocess the point cloud data.

[0041] In the detection of cigarette end face defects, the purpose of the preprocessing stage is to convert the original point cloud data into a format suitable for subsequent analysis and defect detection. First, it is inevitable that the original three-dimensional point cloud data obtained by the laser scanner contains noise or irregular information. Therefore, a filtering algorithm can be used to remove the noise in the point cloud data to ensure the accuracy of subsequent analysis. Then, downsampling can be performed on the denoised point cloud data. For example, but not limited to, using a Voxel grid to reduce the data volume, thereby improving the processing speed and reducing the computational burden. After that, the downsampled point cloud data is registered and aligned through, for example, but not limited to, the ICP algorithm to ensure that the data under different perspectives, acquisition devices, and acquisition conditions can be unified in a coordinate system. Finally, the processed point cloud data is normalized to unify its scale and position to ensure the consistency and reliability of the data. Especially when the cigarette end face is collected by multiple scanners or at different time points, this step is more crucial and can effectively avoid analysis deviations caused by scale differences. The above specific preprocessing steps provide clear, clean, and standardized point cloud data for subsequent defect detection, ensuring the accuracy and efficiency of defect detection.

[0042] Due to the importance of the preprocessing optimization stage, in combination with the foregoing embodiments, the key links involved are further elaborated as follows:

[0043] (1) Denoising of cigarette end face point cloud data

[0044] During the acquisition process, the point cloud data often contains noise points, which will affect the accuracy of subsequent processing. To reduce this noise, in some preferred embodiments of the present invention, the Statistical Outlier Removal (SOR) algorithm is adopted. It removes the outlier points with large distance deviations by calculating the distance between each point and its neighborhood points. The specific method is to calculate the average distance between the neighborhood points of each point in the point cloud. If the distance of a certain point is significantly greater than the average value, it is regarded as a noise point. The SOR algorithm formula is:

[0045] (1)

[0046] and respectively represent two points in the point cloud, usually any two different points in the point cloud;

[0047] , , respectively represent the in three-dimensional space , , coordinates;

[0048] , , respectively represent points in three-dimensional space 、 、 coordinates;

[0049] represents the Euclidean distance between point and point in three-dimensional space, that is, their distance in three-dimensional space.

[0050] (2) Regarding the downsampling of the point cloud data of the cigarette end face

[0051] Point clouds usually contain a large number of data points, which will lead to a very large amount of computational processing. Therefore, it is necessary to downsample the point cloud to reduce the amount of data. In some preferred embodiments of the present invention, the voxel grid filtering algorithm is used to divide the point cloud data space into voxel units of equal size, and a representative point within each voxel is used to replace all the points in the entire voxel, thereby reducing the number of points; the principle of voxel grid filtering refers to the following formula: the point cloud space is divided into voxels, and a representative point (such as the average value or the center point) is selected within each voxel to achieve the effect of reducing the amount of data.

[0052] (2)

[0053] : represents the region of the point cloud space after voxel grid filtering, which is a set containing all voxels;

[0054] : represents the coordinate axes in the point cloud space, usually the coordinates in three-dimensional space, respectively representing the coordinates on the axis;

[0055] : represents the side length of each voxel unit. The sizes of all voxels are the same, and the range of the voxels is within interval;

[0056] : represents all coordinates belong to the interval , that is to say, the coordinates of all point clouds are within this range.

[0057] (3) Regarding the registration and alignment of the point cloud data of the cigarette end face

[0058] When actually collecting the end - face point cloud of a moving cigarette, it can be obtained from different perspectives or at different time points. Therefore, there may be offsets between the point clouds, resulting in the inability to directly fuse the data. In some preferred embodiments of the present invention, the Iterative Closest Point algorithm (ICP) is used to align point clouds from different sources. ICP minimizes the Euclidean distance between two sets of points through iterative optimization, enabling the alignment of the two sets of point clouds. Suppose there are two sets of point clouds and . The algorithm adjusts the pose (position and orientation) of one point cloud by continuously calculating the nearest - neighbor matching of the two sets of points and gradually minimizes the distance error between the point clouds.

[0059] (3)

[0060] : Represents the error function in the ICP algorithm. Specifically, it is the total error between the two sets of point clouds. The goal is to align the point clouds by minimizing this error. Here, the error represents the difference between the points in the point cloud after applying the transformation to the point cloud and the points in the point cloud

[0061] : Represents the th point in the point cloud . The point cloud is the reference point cloud, usually the target to be aligned to another point cloud;

[0062] : Represents the th point in the point cloud . The point cloud is the point cloud that needs to be adjusted (through the transformation ) and aligned with ;

[0063] : Represents the new position obtained by applying the transformation to the point . The transformation usually consists of translation and rotation, and is used to adjust the points of the point cloud to the corresponding positions of the point cloud ;

[0064] : Represents the square of the Euclidean distance between the point and the transformed point . Calculate the error between each pair of corresponding points and sum them. The smaller the error value, the better the alignment of the point clouds; that is, it represents the sum of the errors for all pairs of points (i.e., and The sum of the errors between all corresponding points is calculated by summing each pair of corresponding points in .

[0065] Step S3, segmenting the preprocessed point cloud data based on geometric features to extract the pending defect area of the cigarette end face;

[0066] After the preprocessing stage is completed, the next key task is to segment the point cloud and extract the defective areas that need to be further determined. In order to achieve this goal, the present invention adopts a point cloud segmentation algorithm based on geometric features. Therefore, in the region extraction and segmentation stage of point cloud data processing, the point cloud data can be segmented into multiple regions through the region growing algorithm and the K-means clustering algorithm, that is, according to the geometric features such as the spatial position and normal vector of the point, the point cloud is segmented into several regions with similar properties, so as to identify the part containing defects. Among them, the region growing algorithm starts from the seed point (initial point) and gradually adds adjacent points to the region according to the geometric similarity between the points until certain conditions (such as distance thresholds) are met; K-means groups similar points together through a clustering mechanism. For defect detection on the end face of cigarettes, these segmentation methods can effectively extract areas where defects may exist. Through precise regional division, the accuracy and efficiency of defect identification can be improved, providing reliable data support for subsequent cigarette quality analysis.

[0067] Here is a description of one of the segmentation and extraction methods:

[0068] Starting from a seed point, expand to the points in its neighborhood. If the distance between the extended point and the current point meets the threshold condition, it will be added to the region, otherwise stop expanding.

[0069] (4)

[0070] : The current point, usually the seed point, can also be the point being expanded;

[0071] :With the current point Any point in the set of points to be compared;

[0072] : indicates a point and Point The distance between

[0073] : Distance threshold, used to determine whether the points are in the same cluster (region). , then it is considered that point Can be added to the current area;

[0074] : Indicates that the point is included The area (i.e. the extracted pending defect area) is the set of all points that meet the above distance conditions.

[0075] Step S4, performing quantitative analysis on the undetermined defect area of the cigarette end face to obtain quantitative defect characteristics;

[0076] Defect area quantification refers to the calculation and analysis of quantitative indicators of defect areas based on point cloud data, so as to describe and evaluate defects from multiple dimensions. Usually, these quantitative indicators can include information such as defect depth, area, boundary curve, etc.

[0077] For example, after completing the segmentation of point cloud data and the extraction of defective areas, the severity and scope of the defect can be evaluated by quantitatively expressing the depth of the defective area (usually the vertical height difference as the point cloud depth value, i.e., the z coordinate). In defect detection based on point cloud data, depth quantification helps to accurately evaluate the geometric form, shape, and spatial distribution of defects, thereby providing more refined quality control and analysis.

[0078] As mentioned above, the severity of the defect can be measured by calculating the z coordinates (depth value, point cloud height difference) of each point in the defect area. For example, the depth of the short area can be represented by the point cloud height difference in the pending defect area, and the depth value of the defect can be obtained by calculating the maximum depth difference of all points in the same pending area:

[0079] (5)

[0080] : Indicates the maximum depth difference within a certain undetermined defect area.

[0081] , , : The three-dimensional coordinates of a point in the point cloud, and is the horizontal coordinate, is the vertical coordinate (which represents the depth).

[0082] In addition, the two-dimensional features of the defect can be obtained by calculating the projected area of the point cloud in the pending defect area, or the boundary of the pending defect area can be extracted in combination with the convex hull algorithm, and then the area of the pending defect area can be calculated based on the boundary.

[0083] Step S5: Use the quantitative defect features to determine whether the target area is a target area where real defects exist.

[0084] In actual operation, the threshold segmentation method can be used to discriminate the features of each single point in a certain area to be determined, and based on the statistical results after discrimination, determine whether the current area to be determined with defects is the target area; or it can also directly determine whether the current area to be determined with defects is the target area based on the relationship between the maximum depth difference and the threshold in the foregoing example.

[0085] Corresponding to the above-mentioned embodiments of the method, the present invention also provides an embodiment of a cigarette end face defect detection system based on three-dimensional data, as Figure 2 shown, which includes:

[0086] A point cloud acquisition module 201, configured to acquire point cloud data of the cigarette end face during the movement of the cigarette;

[0087] A point cloud preprocessing module 202, configured to preprocess the point cloud data;

[0088] A defect area extraction module 203, configured to segment the preprocessed point cloud data based on geometric features, and extract the area to be determined with defects on the cigarette end face;

[0089] A feature quantization module 204, configured to perform quantitative analysis on the area to be determined with defects on the cigarette end face to obtain quantitative defect features;

[0090] A target defect determination module 205, configured to use the quantitative defect features to determine whether it belongs to the target area with real defects.

[0091] It should be understood that the division of each component in the above-mentioned Figure 2 shown cigarette end face defect detection system based on three-dimensional data is only a logical function division. In actual implementation, it can be fully or partially integrated into a physical entity, or physically separated. And these components can all be implemented in the form of software called by a processing element; some components can also be implemented in the form of software called by a processing element, and some components can be implemented in the form of hardware. For example, a certain above-mentioned module can be a separately established processing element, or can be integrated in a certain chip of an electronic device. The implementation of other components is similar. In addition, all or part of these components can be integrated together or can be independently implemented. During the implementation process, each step of the above method or each of the above components can be completed by the integrated logic circuit in the processor element or the instruction in the form of software.

[0092] For example, the above components may be one or more integrated circuits configured to implement the above methods, such as one or more application specific integrated circuits (ASIC), or one or more microprocessors (DSP), or one or more field programmable gate arrays (FPGA). For another example, these components may be integrated together to be implemented in the form of a system-on-a-chip (SOC).

[0093] In summary, the main design concept of the present invention is to propose an improved solution for end face defect identification and measurement based on three-dimensional point cloud processing technology in response to the needs of online real-time detection of cigarette end face defects and on-site production scenarios. By performing targeted preprocessing on the three-dimensional point cloud data of the cigarette end face, and using geometric feature extraction and point cloud segmentation methods, it is possible to efficiently and accurately identify and locate the empty heads, thin heads, and sunken ends on the cigarette end faces. Compared with traditional detection methods based on two-dimensional images, the present invention effectively utilizes the depth information of three-dimensional point cloud data, can more accurately reflect the three-dimensional features of end face defects, and ensure the effective identification of minor defects; in addition, through the optimization processing and noise suppression of point cloud data, the detection accuracy and stability can be further improved, meeting the production line's needs for efficient and real-time detection of cigarette end face defects.

[0094] If the expressions expressing orientation are mentioned in the embodiments of the present invention, they are relative concepts based on the embodiments. In addition, "at least one" means one or more, and "plurality" means two or more. "And / or" describes the association relationship of associated objects, indicating that three relationships may exist. For example, A and / or B can represent the existence of A alone, the existence of A and B at the same time, and the existence of B alone. Among them, A and B can be singular or plural. The character " / " generally indicates that the previous and next associated objects are in an "or" relationship. "At least one of the following" and similar expressions refer to any combination of these items, including any combination of single or plural items. For example, at least one of a, b and c can be represented by: a, b, c, a and b, a and c, b and c, or a and b and c, wherein a, b, c can be single or multiple.

[0095] The structure, features, and effects of the present invention have been described in detail based on the embodiments shown in the drawings. However, the above are only the preferred embodiments of the present invention. It should be noted that for the technical features involved in the above embodiments and their preferred modes, those skilled in the art can reasonably combine and match them into various equivalent solutions without departing from and without changing the design concept and technical effects of the present invention. Therefore, the scope of implementation of the present invention is not limited by the drawings shown. Any changes made in accordance with the concept of the present invention, or equivalent embodiments modified to equivalent changes, should still be within the protection scope of the present invention as long as they do not exceed the spirit covered by the specification and the drawings.

Claims

1. A method for detecting defects on the end face of cigarette rods based on three-dimensional data, characterized in that, Comprising: Obtaining point cloud data of the end face of a cigarette during the movement of the cigarette; Preprocessing the point cloud data; Segmenting the preprocessed point cloud data based on geometric features to extract a to-be-determined defect area of the end face of the cigarette; Performing quantitative analysis on the to-be-determined defect area of the end face of the cigarette to obtain quantitative defect features; Using the quantitative defect features to determine whether it belongs to a target area with a real defect.

2. The method for detecting defects on the end face of cigarette rods based on three-dimensional data according to claim 1, wherein The obtaining of the quantitative defect features includes calculating the vertical height difference of each point in the to-be-determined defect area and obtaining a defect depth value based on the vertical height difference, at least including: taking the maximum vertical height difference among all points in the same to-be-determined defect area as the defect depth value.

3. The method for detecting cigarette end face defects based on three-dimensional data according to claim 2, wherein, The obtaining of the quantitative defect features further includes: Calculating the projected area of the point cloud in the to-be-determined defect area to obtain a defect two-dimensional feature; Or, combining a convex hull algorithm to extract the boundary of the to-be-determined defect area and using the boundary to obtain the area of the to-be-determined defect area.

4. The method for detecting defects on the end face of cigarette rods based on three-dimensional data according to claim 2, characterized in that, The determining whether it belongs to a target area with a real defect includes: Using a threshold segmentation method to discriminate the quantitative defect features of each single point in the to-be-determined defect area, and determining whether the current to-be-determined defect area is a target area according to the statistical result after discrimination; Or, directly discriminating whether the current to-be-determined defect area is a target area based on the relationship between the maximum vertical height difference and a preset corresponding threshold.

5. The method for detecting defects on the end face of cigarette rods based on three-dimensional data according to any one of claims 1 to 4, characterized in that, The preprocessing includes: Using a filtering algorithm to remove noise in the point cloud data; Performing downsampling on the noise-reduced point cloud data; Performing registration and alignment processing on the downsampled point cloud data; And, when the end face of the cigarette is collected by multiple devices or at different times, normalizing the point cloud data after the registration and alignment processing.

6. The method for detecting defects on the end face of a cigarette based on three-dimensional data according to claim 5, wherein The downsampling includes: reducing the data volume using a Voxel grid; the registration and alignment processing includes: performing registration and alignment processing using an ICP algorithm.

7. A cigarette end face defect detection system based on three-dimensional data, characterized in that, Comprising: A point cloud acquisition module for obtaining point cloud data of the end face of a cigarette during the movement of the cigarette; A point cloud preprocessing module for preprocessing the point cloud data; A defect area extraction module for segmenting the preprocessed point cloud data based on geometric features to extract a to-be-determined defect area of the end face of the cigarette; A feature quantification module for performing quantitative analysis on the to-be-determined defect area of the end face of the cigarette to obtain quantitative defect features; A target defect determination module for using the quantitative defect features to determine whether it belongs to a target area with a real defect.

Images