Feature fusion classification method for multiple types of packaging bags

Abstract

The invention discloses a feature fusion classification method for multiple types of packaging bags, and relates to the technical field of classification of multiple types of packaging bags. Based on a random forest thought, a feature fusion classification algorithm for various types of packaging bags is given; through three different classification methods, namely a support vector machine, template matching and a neural network, various packaging bag images transmitted by a camera are denoised by using a median filter, and the denoised images are enhanced based on a homomorphic filtering algorithm. And classifying the images by using a support vector machine model, a template matching algorithm and a neural network model, and performing a voting mechanism that minority obeys majority on prediction results of the three methods to obtain a final result. The feature fusion classification algorithm for the multiple types of packaging bags based on the voting mechanism is high in accuracy, errors are reduced, and a more accurate result is obtained.

Description

Technical Field

[0001] The present invention belongs to the technical field of classification of various packaging bags, and in particular relates to a feature fusion classification method for various types of packaging bags. Background Art

[0002] In factory automated production, timely classification of various packaging bags plays an important role. By classifying packaging bags, the factory can improve the level of automated intelligence and production efficiency, thereby improving the quality and market competitiveness of the factory's products.

[0003] The existing packaging bag classification technology is lagging behind and cannot meet the needs of efficient and high-precision detection, which affects the accuracy and efficiency of detection; the traditional algorithms corresponding to the existing packaging bag classification technology are fast but have poor robustness and are easily interfered with; the deep learning method is slow and does not meet the requirements of timely classification; random forest is a common method of deep learning, which is a classifier composed of multiple decision trees, but there is no association between each decision tree; whenever a new input sample enters the forest, each decision tree will classify the sample, and then obtain the final classification result by voting or averaging. Random forest has high accuracy and strong interpretability. It performs well in many fields including classification, regression and feature selection. It is a powerful machine learning algorithm.

[0004] In order to reduce classification errors and improve system robustness, this technical solution proposes a method based on the idea of ​​random forest to classify packaging bags using multiple classification methods, and obtains the final result through a minority-subordinate voting mechanism. Summary of the invention

[0005] The present invention provides a feature fusion classification method for various types of packaging bags, which solves the above problems.

[0006] In order to solve the above technical problems, the present invention is achieved through the following technical solutions:

[0007] The classification robot applied to the feature fusion classification method includes packaging bags, conveyor belts, camera modules, mechanical arms and palletizing platforms; the packaging bags are transported to the camera module via the conveyor belt, images of the packaging bags to be inspected are collected, and then the classification module is used to classify the images collected by the camera module to determine the final category, and finally the mechanical arm places the packaging bags on the palletizing platform of the corresponding category; the classification module stores a computer program of the feature fusion classification method for multiple types of packaging bags;

[0008] A feature fusion classification method for multiple types of packaging bags of the present invention comprises the following steps:

[0009] S1, the conveyor belt transports the packaging bag to the camera module, and the camera module installed on the packaging machine collects images of the packaging bag to be inspected;

[0010] S2. Process the collected packaging bag images using an image denoising module to improve image quality and reduce interference;

[0011] S3, enhancing the packaging bag image through an image enhancement module;

[0012] S4, constructing a support vector machine model and solving it; this step includes the following sub-steps:

[0013] S41, collect multiple types of packaging bag images, take a part of them to form a sample training set, and take the other part to form a test set;

[0014] S42, constructing a support vector machine model, and training the support vector machine model using the training set;

[0015] Among them, building a support vector machine model includes the following steps:

[0016] Establish the objective function with the shortest distance between feature points and hyperplane

[0017]

[0018] sty i (w T x i +b)≥1,i=1,2…n. (2)

[0019] Where w represents an n-dimensional vector, b represents a real number, and st represents a constraint condition; y i represents the label of the i-th sample, which is 1 or -1, representing positive and negative samples respectively; x i The feature vector representing the i-th sample is an n-dimensional vector;

[0020] In order to solve the objective function, the Lagrange function is introduced. The Lagrange function of the above formula can be written as:

[0021]

[0022] Among them, α=(α1+α2,…,α n ) T ,w represents the normal vector of the partitioning hyperplane, x i represents the i-th feature attribute of a single sample, and b represents the distance between the origin and the hyperplane;

[0023] Analyze the corresponding dual problem:

[0024]

[0025] Among them, α i and α j is the Lagrange multiplier, used to solve the dual problem, and is a vector, each element of which corresponds to a sample; i and j Represents the labels of the i-th and j-th samples, indicating 1 or -1, representing positive and negative samples respectively; x i T The feature vector representing the i-th sample is an n-dimensional vector, where T represents transposition; x j : The feature vector of the jth sample, also an n-dimensional vector;

[0026] Solve the optimal solution

[0027] Find the optimal classification function:

[0028]

[0029] Among them, Sgn() represents a sign function, which outputs 1 when the input is greater than 0, -1 when it is less than 0, and 0 when it is equal to 0; f(x) represents a classification function, which is used to calculate which category the sample x belongs to;

[0030] The support vector machine model adopts a linear kernel function, and the formula is as follows:

[0031] K(x i ,x j )=(x i *x j ) (7)

[0032] Among them, K(x i ,x j ) represents the kernel function, representing the sample x i and sample x j The similarity between i and x j is the input vector;

[0033] S43, using a testing machine to test the performance of the support vector machine model to achieve the required accuracy;

[0034] S44, input the packaging bag image to be detected, perform category judgment, and obtain the classification result;

[0035] S5, constructing a template matching model and determining the category; specifically including the following steps:

[0036] S51, constructing a plurality of template images according to the packaging bags to be identified, converting them into grayscale images, and obtaining grayscale values ​​of various regions of the template images;

[0037] S52, converting the collected image into a grayscale image, identifying the grayscale image of the collected image with the grayscale image of each template image using a template matching algorithm, setting a recognition threshold, and when the correlation exceeds the threshold, determining that the packaging bag to be identified may be the category of the template image, and the template matching algorithm formula used is as follows:

[0038]

[0039] Wherein, R(x,y) represents the correlation between the area centered at (x,y) in the image to be identified and the template. The larger R(x,y) is, the higher the similarity is, and the higher the possibility that the packaging bag to be identified is the template category is. T is the template image, I is the matching image, (x',y') represents all pixel coordinates of the traversed template image, T'(x',y') represents the pixel value at the position (x',y') of the template image, and I'(x+x',y+y') represents the pixel value at the position (x+x',y+y') of the image to be identified.

[0040] S53, comparing the matching degrees of each template of the packaging bag to be identified, and taking the category with the highest matching degree as the final classification result;

[0041] S6. Constructing a neural network model and solving it, specifically including the following steps:

[0042] S61, collect multiple types of packaging bag images, take a part of them to form a sample training set, and take the other part to form a test set;

[0043] S62. Initialize the neural network parameters. The BP neural network consists of three layers: input layer, hidden layer and output layer. Each layer contains several neurons. The formula for the number of neurons is as follows:

[0044]

[0045] Where n is the number of neurons in the input layer, m is the number of neurons in the output layer, and a is a constant between [1,10];

[0046] The weight update formula is as follows:

[0047]

[0048] Among them, W (k+1) represents the K+1th iteration, that is, the updated weight; W (k) represents the Kth iteration, that is, the weight before update; μ represents the learning rate, which controls the step size of each update; represents the gradient of weight W, that is, the direction of weight change; L represents the loss function;

[0049] The bias update formula is as follows:

[0050]

[0051] Among them, b (k+1) represents the bias value after the k+1th iteration, that is, the updated bias value; b (k) Represents the bias value of the kth iteration, that is, the bias value before the update; represents the gradient of bias b, that is, the derivative of the loss function with respect to bias b,

[0052] S63, select the Sigmoid function as the activation function, the formula is as follows:

[0053]

[0054] S64, forward propagation calculation, the forward propagation function formula is as follows:

[0055] h (l) =f(W (l) h (l-1) +b (l) ) (13)

[0056] Among them, h (l-1) Represents the output vector of the previous layer, which is also the input vector of the current layer, h (l) represents the output vector of the current layer, f represents the activation function of the current layer, W (l) Represents the weight matrix of the previous layer and the current layer, b (l) Represents the bias of the current layer;

[0057] S65, calculate the error, the error loss objective function formula is as follows:

[0058] CE(y,y')=-ylog(y')-(1-y)log(1-y') (14)

[0059] Among them, y represents the true label, and y' represents the predicted value of the model;

[0060] S66. Update the weights and biases in the neural network through the back propagation algorithm to reduce the loss function;

[0061] S67, repeating steps S64-S66, training the neural network for multiple iterations until convergence or the maximum number of iterations is reached;

[0062] S68. Use the test set to evaluate the performance of the trained neural network model;

[0063] S69, using the trained neural network model to classify the packaging bags to be inspected;

[0064] S7, the results obtained by the three classification methods are voted on by the majority voting mechanism to obtain the final classification result;

[0065] S8. Classify and palletize the packaging bags;

[0066] Furthermore, the classification robot applied by the feature fusion classification method includes packaging bags, conveyor belts, camera modules, robotic arms and palletizing tables; the packaging bags are transported to the bottom of the camera module by means of the conveyor belt, images of the packaging bags to be inspected are captured, and then the classification module is used to classify the images captured by the camera module to determine the final category, and finally the robotic arm places the packaging bags on the palletizing table of the corresponding category; the classification module stores a computer program for the feature fusion classification method for multiple types of packaging bags.

[0067] Compared with the prior art, the present invention has the following beneficial effects:

[0068] (1) This solution is based on the classification method of support vector machine, template matching and random forest of neural network, which can improve the classification accuracy of packaging bags and reduce the occurrence of misidentification;

[0069] (2) This solution can improve the factory’s automation intelligence level and production efficiency, thereby enhancing the quality and market competitiveness of the factory’s products.

[0070] Of course, any product implementing the present invention does not necessarily need to achieve all of the advantages described above at the same time. BRIEF DESCRIPTION OF THE DRAWINGS

[0071] In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the accompanying drawings required for describing the embodiments will be briefly introduced below. Obviously, the accompanying drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other accompanying drawings can be obtained based on these accompanying drawings without paying creative work.

[0072] Figure 1 The present invention is a flow chart of the method.

[0073] Figure 2 This is a schematic diagram of packaging bag identification according to the present invention.

[0074] Figure 3 The figure is a schematic diagram of the support vector machine method flow of the present invention.

[0075] Figure 4 The figure is a schematic diagram of the flow chart of the template matching method of the present invention.

[0076] Figure 5 It is a schematic diagram of the BP neural network method flow of the present invention;

[0077] In the accompanying drawings, the components represented by the reference numerals are listed as follows:

[0078] 1-packaging bag, 2-conveyor belt, 3-camera module, 4-robotic arm, 5-palletizing platform. DETAILED DESCRIPTION

[0079] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only 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 creative work are within the scope of protection of the present invention.

[0080] like Figure 2 As shown, the classification robot corresponding to the feature fusion classification algorithm for multiple types of packaging bags in this technical solution includes the following parts: 1-packaging bags, 2-conveyor belt, 3-camera module, 4-manipulator and 5-palletizing platform. First, the packaging bags are transported to the camera module through the conveyor belt, and the image of the packaging bags to be inspected is collected. Then, the classification module is used to classify the images through three methods to determine the final category. Finally, the robot arm puts the packaging bags on the palletizing platform of the corresponding category.

[0081] like Figure 1 As shown in FIG. 1 , a feature fusion classification algorithm for multiple types of packaging bags includes the following steps:

[0082] S1: The conveyor belt transports the packaging bag to the camera module, and the camera module installed on the packaging machine collects images of the packaging bag to be inspected.

[0083] S2: Use the image denoising module to process the collected packaging bag images to improve image quality and reduce interference.

[0084] S3: Enhance the packaging bag image through the image enhancement module.

[0085] S4: Build a support vector machine model and solve it, such as Figure 3 As shown, the method specifically includes the following steps.

[0086] S41: Collect images of various types of packaging bags, take a part of them to construct a sample training set, and take the other part to construct a test set.

[0087] S42: constructing a support vector machine model, and using the training set to train the support vector machine model, wherein constructing the support vector machine includes the following steps:

[0088] Establish the objective function with the shortest distance between feature points and hyperplane

[0089]

[0090] sty i (w T x i +b)≥1,i=1,2…n. (2)

[0091] Where w represents an n-dimensional vector, b represents a real number, and st represents a constraint condition. i Represents the label of the i-th sample, which is 1 or -1, representing positive and negative samples respectively. i The feature vector representing the i-th sample is an n-dimensional vector.

[0092] In order to solve the objective function, the Lagrange function is introduced. The Lagrange function of the above formula can be written as:

[0093]

[0094] Among them, α=(α1+α2,…,α n ) T ,w represents the normal vector of the partitioning hyperplane, x i represents the i-th feature attribute of a single sample, and b represents the distance between the origin and the hyperplane.

[0095] Analyze the corresponding dual problem:

[0096]

[0097] Among them, α i and α j is the Lagrange multiplier, used to solve the dual problem, and is a vector, each element of which corresponds to a sample. i and j Represents the labels of the i-th and j-th samples, which are 1 or -1, representing positive and negative samples respectively. i T The feature vector representing the i-th sample is an n-dimensional vector, where T represents transposition. j : The feature vector of the jth sample, also an n-dimensional vector.

[0098] Solve the optimal solution

[0099] Find the optimal classification function:

[0100]

[0101] Among them, Sgn() represents a sign function, which outputs 1 when the input is greater than 0, -1 when it is less than 0, and 0 when it is equal to 0. f(x) represents a classification function, which is used to calculate which category the sample x belongs to.

[0102] The support vector machine model adopts a linear kernel function, and the formula is as follows:

[0103] K(x i ,x j )=(x i *x j ) (7)

[0104] Among them, K(x i ,x j ) represents the kernel function, representing the sample x i and sample x j The similarity between i and x j is the input vector.

[0105] S43: Use the test machine to test the performance of the support vector machine model to achieve the required accuracy.

[0106] S44: Input the packaging bag image to be detected, perform category judgment, and obtain a classification result.

[0107] S5: Build a template matching model and determine the category, such as Figure 4 As shown, it specifically includes the following steps:

[0108] S51: constructing a plurality of template images according to the packaging bags to be identified, converting them into grayscale images, and obtaining grayscale values ​​of various regions of the template images.

[0109] S52: Convert the collected image into a grayscale image, and use the template matching algorithm to identify the grayscale image of the collected image and the grayscale image of each template image. Set a recognition threshold. When the correlation exceeds the threshold, it is determined that the packaging bag to be identified may be the category of the template image. The template matching algorithm formula used is as follows:

[0110]

[0111] Among them, R(x,y) represents the correlation between the area centered at (x,y) in the image to be identified and the template. The larger R(x,y) is, the higher the similarity is, and the higher the possibility that the packaging bag to be identified is the template category. T is the template image, I is the matching image, (x',y') represents all pixel coordinates of the traversed template image, T'(x',y') represents the pixel value of the template image at position (x',y'), and I'(x+x',y+y') represents the pixel value of the image to be identified at position (x+x',y+y').

[0112] S53: Compare the matching degrees of each template of the packaging bag to be identified, and take the category with the highest matching degree as the final classification result.

[0113] S6: Build a neural network model and solve it; Figure 5 As shown, it specifically includes the following steps:

[0114] S61: Collect images of various types of packaging bags, take a part of them to construct a sample training set, and take the other part to construct a test set.

[0115] S62: Initialize the neural network parameters. The BP neural network consists of three layers: input layer, hidden layer and output layer. Each layer contains several neurons. The formula for the number of neurons is as follows:

[0116]

[0117] Where n is the number of neurons in the input layer, m is the number of neurons in the output layer, and a is a constant between [1,10].

[0118] The weight update formula is as follows:

[0119]

[0120] Among them, W (k+1) represents the K+1th iteration, that is, the updated weight. (k) represents the Kth iteration, that is, the weight before update. μ represents the learning rate, which controls the step size of each update. It represents the gradient of weight W, that is, the direction of weight change. L represents the loss function.

[0121] The bias update formula is as follows:

[0122]

[0123] Among them, b (k+1) Indicates the bias value after the k+1th iteration, that is, the updated bias value. (k) Represents the bias value of the kth iteration, that is, the bias value before the update. represents the gradient of bias b, that is, the derivative of the loss function with respect to bias b,

[0124] S63: The activation function selects the Sigmoid function, and the formula is as follows:

[0125]

[0126] S64: Forward propagation calculation, the forward propagation function formula is as follows:

[0127] h (l) =y(W (l) h (l-1) +b(l) ) (13)

[0128] Among them, h (l-1) Represents the output vector of the previous layer, which is also the input vector of the current layer, h (l) represents the output vector of the current layer, f represents the activation function of the current layer, W (l) Represents the weight matrix of the previous layer and the current layer, b (l) Represents the bias of the current layer.

[0129] S65: Calculate the error. The error loss objective function formula is as follows:

[0130] CE(y,y')=-ylog(y')-(1-y)log(1-y') (14)

[0131] Among them, y represents the true label and y' represents the predicted value of the model.

[0132] S66: Update the weights and biases in the neural network through the back-propagation algorithm to reduce the loss function.

[0133] S67: Repeat steps S64-S66, training the neural network for multiple iterations until convergence or the maximum number of iterations is reached.

[0134] S68: Use the test set to evaluate the performance of the trained neural network model.

[0135] S69: Use the trained neural network model to classify the packaging bags to be inspected.

[0136] S7: The results obtained by the three classification methods are subjected to majority voting mechanism to obtain the final classification result.

[0137] S8: Classify and palletize the packaging bags.

[0138] The preferred embodiments of the present invention disclosed above are only used to help illustrate the present invention. The preferred embodiments do not describe all the details in detail, nor do they limit the invention to the specific implementation methods described. Obviously, many modifications and changes can be made according to the content of this specification. This specification selects and specifically describes these embodiments in order to better explain the principles and practical applications of the present invention, so that those skilled in the art can understand and use the present invention well. The present invention is limited only by the claims and their full scope and equivalents.

Claims

1. A feature fusion classification method for multiple types of packaging bags, characterized in that: The steps include: S1, the conveyor belt transports the packaging bag to the camera module, and the camera module installed on the packaging machine collects images of the packaging bag to be inspected; S2. Process the collected packaging bag images using an image denoising module to improve image quality and reduce interference; S3, enhancing the packaging bag image through an image enhancement module; S4, constructing a support vector machine model and solving it; this step includes the following sub-steps: S41, collect multiple types of packaging bag images, take a part of them to form a sample training set, and take the other part to form a test set; S42, constructing a support vector machine model, and training the support vector machine model using the training set; Among them, building a support vector machine model includes the following steps: Establish the objective function with the shortest distance between feature points and hyperplane s.t.y i (w T x i +b)≥1,i=1,2…n. (2) Where w represents an n-dimensional vector, b represents a real number, and st represents a constraint condition; y i represents the label of the i-th sample, which is 1 or -1, representing positive and negative samples respectively; x i The feature vector representing the i-th sample is an n-dimensional vector; In order to solve the objective function, the Lagrange function is introduced. The Lagrange function of the above formula is written as: Among them, α=(α1+α2,…,α n ) T ,w represents the normal vector of the partitioning hyperplane, x i represents the i-th feature attribute of a single sample, and b represents the distance between the origin and the hyperplane; Analyze the corresponding dual problem: Among them, α i and α j is the Lagrange multiplier, used to solve the dual problem, and is a vector, each element of which corresponds to a sample; i and j Represents the labels of the i-th and j-th samples, indicating 1 or -1, representing positive and negative samples respectively; x i T The feature vector representing the i-th sample is an n-dimensional vector, where T represents transposition; x j : The feature vector of the jth sample, also an n-dimensional vector; Solve the optimal solution Find the optimal classification function: Among them, Sgn() represents a sign function, which outputs 1 when the input is greater than 0, -1 when it is less than 0, and 0 when it is equal to 0; f(x) represents a classification function, which is used to calculate which category the sample x belongs to; The support vector machine model adopts a linear kernel function, and the formula is as follows: K(x i ,x j )=(x i *x j ) (7) Among them, K(x i ,x j ) represents the kernel function, representing the sample x i and sample x j The similarity between i and x j is the input vector; S43, using a testing machine to test the performance of the support vector machine model to achieve the required accuracy; S44, input the packaging bag image to be detected, perform category judgment, and obtain the classification result; S5, constructing a template matching model and determining the category; specifically including the following steps: S51, constructing a plurality of template images according to the packaging bags to be identified, converting them into grayscale images, and obtaining grayscale values ​​of various regions of the template images; S52, converting the collected image into a grayscale image, identifying the grayscale image of the collected image with the grayscale image of each template image using a template matching algorithm, setting a recognition threshold, and when the correlation exceeds the threshold, determining that the packaging bag to be identified may be the category of the template image, and the template matching algorithm formula used is as follows: Wherein, R(x,y) represents the correlation between the area centered at (x,y) in the image to be identified and the template. The larger R(x,y) is, the higher the similarity is, and the higher the possibility that the packaging bag to be identified is the template category is. T is the template image, I is the matching image, (x',y') represents all pixel coordinates of the traversed template image, T'(x',y') represents the pixel value at the position (x',y') of the template image, and I'(x+x',y+y') represents the pixel value at the position (x+x',y+y') of the image to be identified. S53, comparing the matching degrees of each template of the packaging bag to be identified, and taking the category with the highest matching degree as the final classification result; S6. Constructing a neural network model and solving it, specifically including the following steps: S61, collect multiple types of packaging bag images, take a part of them to form a sample training set, and take the other part to form a test set; S62. Initialize the neural network parameters. The BP neural network consists of three layers: input layer, hidden layer and output layer. Each layer contains several neurons. The formula for the number of neurons is as follows: Where n is the number of neurons in the input layer, m is the number of neurons in the output layer, and a is a constant between [1,10]; The weight update formula is as follows: Among them, W (k+1) represents the K+1th iteration, that is, the updated weight; W (k) represents the Kth iteration, that is, the weight before update; μ represents the learning rate, which controls the step size of each update; represents the gradient of weight W, that is, the direction of weight change; L represents the loss function; The bias update formula is as follows: Among them, v (k+1) represents the bias value after the k+1th iteration, that is, the updated bias value; b (k) Represents the bias value of the kth iteration, that is, the bias value before the update; represents the gradient of bias b, that is, the derivative of the loss function with respect to bias b, S63, select the Sigmoid function as the activation function, the formula is as follows: S64, forward propagation calculation, the forward propagation function formula is as follows: h (l) =f(W (l) h (l-1) +b (l) ) (13) Among them, h (l-1) Represents the output vector of the previous layer, which is also the input vector of the current layer, h (l) represents the output vector of the current layer, f represents the activation function of the current layer, W (l) Represents the weight matrix of the previous layer and the current layer, b (l) Represents the bias of the current layer; S65, calculate the error, the error loss objective function formula is as follows: CE(y,y')=-ylog(y')-(1-y)log(1-y') (14) Among them, y represents the true label, and y' represents the predicted value of the model; S66. Update the weights and biases in the neural network through the back propagation algorithm to reduce the loss function; S67, repeating steps S64-S66, training the neural network for multiple iterations until convergence or the maximum number of iterations is reached; S68. Use the test set to evaluate the performance of the trained neural network model; S69, using the trained neural network model to classify the packaging bags to be inspected; S7, the results obtained by the three classification methods are voted on by the majority voting mechanism to obtain the final classification result; S8. Classify and palletize the packaging bags.

2. The feature fusion classification method for multiple types of packaging bags according to claim 1 is characterized in that: The classification robot used in the feature fusion classification method comprises a packaging bag (1), a conveyor belt (2), a camera module (3), a mechanical arm (4) and a stacking platform (5); the packaging bag (1) is transported to the camera module (3) via the conveyor belt (2), an image of the packaging bag (1) to be inspected is captured, and then the classification module is used to classify the image captured by the camera module (3) to determine the final category, and finally the mechanical arm (4) places the packaging bag (1) on a stacking platform (5) of the corresponding category; the classification module stores a computer program for the feature fusion classification method for multiple types of packaging bags.

Images