A method for identifying types of cigarette auxiliary materials for storage
The deep neural network is used to identify the types of cigarette auxiliary materials, which solves the problem of lack of intelligence and automation of the auxiliary materials storage process, and achieves efficient and accurate identification and inventory of auxiliary materials, improving production efficiency.
Patent Information
- Application Number
- CN202411832107.2
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2024-12-12
- Publication Date
- 2025-04-01
AI Technical Summary
The cigarette industry's auxiliary material storage process lacks intelligence and automation, resulting in low counting efficiency, high labor intensity, low accuracy, energy waste and error-prone problems.
Deep neural network is used as the auxiliary material type recognition model, and convolutional calculation is performed using the RGB three-dimensional image of auxiliary material taken by the camera, and intelligent recognition and judgment of auxiliary material types are achieved through cascade operation and matrix calculation.
It realizes intelligent identification and judgment of auxiliary material types, optimizes the storage process of auxiliary material, improves inventory efficiency, reduces non-essential working hours, reduces labor costs, and improves production supply efficiency.
Smart Images
Figure CN119741549A_ABST
Abstract
Description
Technical Field
[0001] The present invention relates to the technical field of auxiliary material storage, and in particular to a type identification method for auxiliary material storage of cigarettes. Background Art
[0002] The distribution of auxiliary materials in the cigarette industry is divided into two forms: manual and automatic. However, the automatic auxiliary material return process has not been fully intelligent and automated, and still uses manual counting, which has the characteristics of low counting efficiency, high labor intensity, low accuracy, energy waste, and easy errors.
[0003] According to the previous survey, there are very few companies in the industry that carry out intelligent inventory of auxiliary materials. The existing system is only applicable to the inventory of coil paper and tipping paper, and cannot cover the intelligent identification and determination of auxiliary material types. The existing system requires manual operation of the controller and lacks automation. There is a certain degree of human interference and error in the inventory process. Although it can improve the efficiency of warehousing, it still has problems such as long inventory time and low efficiency in large-scale production environments. Summary of the invention
[0004] The purpose of the present invention is to provide a method for identifying the types of cigarette auxiliary materials entering storage, so as to solve the problem.
[0005] To achieve this object, the present invention adopts the following technical solutions:
[0006] A method for identifying the type of cigarette auxiliary materials entering storage, the method comprising the following steps:
[0007] A deep neural network is used as the auxiliary material type recognition model, and the auxiliary material RGB three-dimensional image taken by the camera is used as input and convolution calculation is performed;
[0008] The convolutional layers of each level of the deep neural network are cascaded, and the obtained image data is used to obtain a prediction frame and the auxiliary material category of the prediction frame through matrix calculation.
[0009] Preferably, the convolution calculation uses a maximum pooling layer as a standard for pixel value extraction.
[0010] Preferably, a dynamic allocation strategy is performed using a Task-Aligned Assigner for the deep neural network, and the weights are dynamically adjusted according to the progress of training and the characteristics of the samples.
[0011] Preferably, the dynamic adjustment weight is adjusted according to the training loss or index, and the degree of alignment is achieved by controlling the optimization of the classification score and the alignment score IoU.
[0012]
[0013] Among them, α and β are weight hyperparameters, s is the predicted score corresponding to the labeled category, and u is the IoU between the predicted box and the actual box. The sum of the two can measure the degree of alignment; when the score is higher and the IoU is higher, the value of t is closer to 1.
[0014] Preferably, the alignment step comprises:
[0015] Based on the classification score and the IoU between the predicted box and the actual box, a loss function of the IoU of the associated classification and regression is obtained by weighting. The IoU loss function is as follows:
[0016]
[0017] Among them, IoU represents the intersection-over-union ratio of the actual box and the predicted box; A and B represent the actual box and the predicted box respectively:
[0018]
[0019] , Represent the center points of the predicted box and the actual box respectively. represents the Euclidean distance between the center points of the predicted box and the actual box, c represents the diagonal distance of the minimum closure area that can contain both the predicted box and the actual box, and v is used to measure the similarity of the aspect ratio, which is defined as:
[0020]
[0021]
[0022] , Represent the width of the predicted box and the width of the actual box respectively. , Represent the height value of the predicted box and the height value of the actual box respectively;
[0023] Calculate whether the center point of the anchor point is within the current actual box. Only the anchor point value within the actual box can be used as a positive sample.
[0024] Preferably, the alignment score IoU in the actual box is used as a positive sample, and the rest are used as negative samples for training.
[0025] Beneficial effects of the invention: The invention realizes intelligent identification and determination of auxiliary material types. Compared with the traditional system, the invention optimizes the auxiliary material storage process, improves inventory efficiency, reduces unnecessary working time, reduces labor costs, improves production and supply efficiency, and provides strong support for the intelligent management of the packaging workshop. BRIEF DESCRIPTION OF THE DRAWINGS
[0026] In order to more clearly illustrate the technical solutions in the embodiments of the present invention, the following briefly introduces the drawings required for use in the description of the embodiments of the present invention. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on the contents of the embodiments of the present invention and these drawings without paying any creative work.
[0027] Figure 1 The invention discloses a method for identifying types of cigarette auxiliary materials for storage. DETAILED DESCRIPTION
[0028] The present invention will be further described in detail below in conjunction with the accompanying drawings and embodiments. It is to be understood that the specific embodiments described herein are only used to explain the present invention, rather than to limit the present invention. It should also be noted that, for ease of description, only parts related to the present invention, rather than all structures, are shown in the accompanying drawings.
[0029] In the description of the present invention, unless otherwise clearly specified and limited, the terms "connected", "connected", and "fixed" should be understood in a broad sense, for example, it can be a fixed connection, a detachable connection, or an integral connection; it can be a mechanical connection or an electrical connection; it can be a direct connection or an indirect connection through an intermediate medium, it can be the internal connection of two elements or the interaction relationship between two elements. For ordinary technicians in this field, the specific meanings of the above terms in the present invention can be understood according to specific circumstances.
[0030] In the present invention, unless otherwise clearly specified and limited, a first feature being "above" or "below" a second feature may include that the first and second features are in direct contact, or may include that the first and second features are not in direct contact but are in contact through another feature between them. Moreover, a first feature being "above", "above" and "above" a second feature includes that the first feature is directly above and obliquely above the second feature, or simply indicates that the first feature is higher in level than the second feature. A first feature being "below", "below" and "below" a second feature includes that the first feature is directly below and obliquely below the second feature, or simply indicates that the first feature is lower in level than the second feature.
[0031] In the description of this embodiment, the terms "upper", "lower", "left", "right" and other directions or positional relationships are based on the directions or positional relationships shown in the drawings, and are only for the convenience of description and simplification of operation, rather than indicating or implying that the device or element referred to must have a specific direction, be constructed and operate in a specific direction, and therefore cannot be understood as limiting the present invention. In addition, the terms "first" and "second" are only used to distinguish in the description and have no special meaning.
[0032] Embodiments of the present invention are described in detail below, examples of which are shown in the accompanying drawings, wherein the same or similar reference numerals throughout represent the same or similar elements or elements having the same or similar functions. The embodiments described below with reference to the accompanying drawings are exemplary and are intended to be used to explain the present invention, and should not be construed as limiting the present invention.
[0033] The method for identifying the types of cigarette auxiliary materials entering the warehouse is proposed to be implemented by using a target detection method based on a deep neural network. The deep learning framework used includes but is not limited to one of the network models such as HRNet, YOLO, and CenterNet. These deep learning frameworks use the idea of regression to directly detect the target position and classify the target object through a one-stage deep neural network, that is, based on a single end-to-end network, from the input of the original image to the output of the object position and category in one go.
[0034] This method uses the RGB three-dimensional image of auxiliary materials taken by the camera as input, and uses the trained deep neural network to achieve fast and accurate positioning and detection of auxiliary material targets. Through this detection result, various auxiliary materials such as coil paper and tipping paper can be accurately located in the image, and the coordinate positions and approximate outlines of these auxiliary materials in the image can be obtained, laying the foundation for subsequent counting work. The specific implementation structure diagram is shown in the figure. Figure 1 As shown, the present invention provides a method for identifying the types of cigarette auxiliary materials entering storage, which specifically includes:
[0035] Step 1: Input and convolution operation of auxiliary RGB 3D image data. The convolution operation uses the maximum pooling layer as the standard for pixel value extraction, which can maximize the retention of important features in the convolution area and ensure the accuracy of subsequent calculations.
[0036] Step 2: Cascade operation and matrix calculation. The cascade operation is to comprehensively consider the features of each convolution layer, rather than only calculating based on the final convolution layer results. This effectively avoids the problem of information loss during the convolution process. Matrix calculation is to calculate the predicted box in the image data, generally represented by (x, y, w, h) and the category of the actual box.
[0037] In addition, in order to make auxiliary material identification more robust, during the training process, we used the Task-Aligned Assigner to perform a dynamic allocation strategy, dynamically adjusting the weights according to the progress of training and the characteristics of the samples. Because it may be difficult to distinguish between positive and negative samples in the early stages of training, more attention should be paid to samples that are easily misclassified. As training progresses, the samples are better distinguished, so the weights of difficult samples should be gradually reduced, while the weights of easy-to-classify samples should be increased. The dynamic allocation strategy can be adjusted according to training loss or other indicators to better adapt to different data sets and models. The formula is as follows:
[0038]
[0039] Among them, α and β are weight hyperparameters. s is the predicted score corresponding to the labeled category, and u is the IoU between the predicted box and the actual box. The sum of the two can measure the degree of alignment. t can simultaneously control the optimization of the classification score and IoU to achieve the degree of alignment, thereby guiding the network to dynamically focus on high-quality anchors. The higher the category score and the higher the IoU, the closer the value of t is to 1.
[0040] The specific steps are as follows:
[0041] 1. Based on the classification score and the IoU between the predicted box and the actual box, a loss function of the IoU of the associated classification and regression is obtained by weighting. The IoU loss function here is as follows:
[0042]
[0043] Among them, IoU represents the intersection-over-union ratio of the actual box and the predicted box, which is calculated as follows, where A and B represent the actual box and the predicted box respectively:
[0044]
[0045] , Represent the center points of the predicted box and the actual box respectively. represents the Euclidean distance between two center points. c represents the diagonal distance of the minimum enclosing area that can contain both the predicted box and the actual box. α is the weight function, and v is used to measure the similarity of aspect ratios, defined as:
[0046]
[0047]
[0048] , Represent the predicted width value and the actual width value respectively. , They represent the predicted height and the true height respectively.
[0049] 2. Calculate whether the center point of the anchor point is within the current actual box. Only the anchor point value within the current actual box can be used as a positive sample.
[0050] 3. Under the premise of meeting 2, the ones in the actual frame based on the alignment score IoU are used as positive samples, and the rest are used as negative samples for training.
[0051] Obviously, the above embodiments of the present invention are merely examples for clearly illustrating the present invention, and are not intended to limit the embodiments of the present invention. For those skilled in the art, various obvious changes, readjustments and substitutions can be made without departing from the protection scope of the present invention. It is not necessary and impossible to list all the embodiments here. Any modifications, equivalent substitutions and improvements made within the spirit and principles of the present invention should be included in the protection scope of the claims of the present invention.
Claims
1. A method for identifying the types of cigarette auxiliary materials entering storage, characterized in that: The type identification method comprises the following steps: A deep neural network is used as the auxiliary material type recognition model, and the auxiliary material RGB three-dimensional image taken by the camera is used as input and convolution calculation is performed; The convolutional layers of each level of the deep neural network are cascaded, and the obtained image data is used to obtain a prediction frame and the auxiliary material category of the prediction frame through matrix calculation.
2. A method for identifying types of cigarette auxiliary materials for storage according to claim 1, characterized in that: The convolution calculation uses the maximum pooling layer as the standard for pixel value extraction.
3. A method for identifying types of cigarette auxiliary materials for storage according to claim 1, characterized in that: A Task-Aligned Assigner is used to perform a dynamic allocation strategy for the deep neural network, and the weights are dynamically adjusted according to the progress of training and the characteristics of the samples.
4. A method for identifying types of cigarette auxiliary materials for storage according to claim 3, characterized in that: The dynamic adjustment weight is adjusted according to the training loss or index, and the degree of alignment is achieved by controlling the optimization of the classification score and the alignment score IoU. Among them, α and β are weight hyperparameters, s is the predicted score corresponding to the labeled category, and u is the IoU between the predicted box and the actual box. The sum of the two can measure the degree of alignment; when the score is higher and the IoU is higher, the value of t is closer to 1.
5. A method for identifying types of cigarette auxiliary materials for storage according to claim 4, characterized in that: The alignment steps include: Based on the classification score and the IoU between the predicted box and the actual box, a loss function of the IoU of the associated classification and regression is obtained by weighting. The IoU loss function is as follows: Among them, IoU represents the intersection-over-union ratio of the actual box and the predicted box; A and B represent the actual box and the predicted box respectively: , Represent the center points of the predicted box and the actual box respectively. represents the Euclidean distance between the center points of the predicted box and the actual box, c represents the diagonal distance of the minimum closure area that can contain both the predicted box and the actual box, and v is used to measure the similarity of the aspect ratio, which is defined as: , Represent the width of the predicted box and the width of the actual box respectively. , Represent the height value of the predicted box and the height value of the actual box respectively; Calculate whether the center point of the anchor point is within the current actual box. Only the anchor point value within the actual box can be used as a positive sample.
6. A method for identifying types of cigarette auxiliary materials for storage according to claim 5, characterized in that: Based on the alignment score IoU, the actual box is used as a positive sample, and the rest are used as negative samples for training.