Seal stamp-pad ink early warning detection system based on deep learning

Abstract

The invention discloses a seal and stamp-pad ink early-warning detection system based on deep learning, and relates to the field of electronic information, the seal and stamp-pad ink early-warning detection system comprises a hardware module and a detection module, and the detection module comprises the following steps: S1, data acquisition and labeling: acquiring seal images in different ink states through a high-definition camera, and recording application printing times through a counter; s2, preprocessing the collected data, including de-noising, graying and binaryzation, so as to improve the image quality; s3, constructing a deep learning model, inputting the preprocessed data into the model, and performing feature extraction on the image by using a deep learning algorithm, the extracted features including the contour, texture and color of the seal; and S4, stamp-pad ink state evaluation: according to the extracted features, evaluating the stamp-pad ink state through a deep learning model. The method has the advantages that the image features can be automatically extracted and evaluated, the real-time monitoring and early warning of the stamp-pad ink state are realized, and the working efficiency is improved.

Description

Technical Field

[0001] The present invention relates to the field of electronic information, and in particular to a seal ink warning detection system based on deep learning. Background Art

[0002] In seal management, the number of seal uses and the state of the ink are two important monitoring indicators. The number of seal uses reflects the usage frequency of the seal, while the state of the ink directly affects the clarity and recognizability of the seal. The following problems exist in traditional seal management methods:

[0003] 1. Dependence on manual inspection: It is necessary to visually judge the ink state regularly, which is prone to missed inspection or misjudgment;

[0004] 2. Inaccurate seal use records: Mechanical counters are susceptible to false triggering or missed triggering;

[0005] 3. Delayed maintenance: The exhaustion of the ink leads to unclear seals, affecting the legal effect of documents;

[0006] 4. Lack of intelligent analysis: It is impossible to predict the remaining number of seal uses and the maintenance cycle

[0007] Therefore, it is of great significance to utilize advanced technologies such as deep learning to realize the automation and intelligence of seal management. The present invention designs a seal ink warning detection system based on deep learning. Summary of the Invention

[0008] The purpose of the present invention is to solve the problems in the prior art, and a seal ink warning detection system based on deep learning is proposed.

[0009] A seal ink warning detection system based on deep learning includes the following parts:

[0010] Hardware module: The hardware module includes a high-definition camera, a counter, and an alarm;

[0011] Detection module, the detection module includes the following steps:

[0012] S1: Data acquisition and annotation: Collect seal images under different ink states through a high-definition camera, and record the corresponding number of seal uses through a counter;

[0013] S2: Preprocess the collected data, including denoising, grayscale conversion, and binarization, to improve the image quality;

[0014] S3: Build a deep learning model, input the preprocessed data into the model, and use deep learning algorithms to extract features from the image. The extracted features include the outline, texture, and color of the seal;

[0015] S4: Ink state evaluation. Based on the extracted features, the deep learning model evaluates the ink state.

[0016] In the above-mentioned seal ink warning detection system based on deep learning, the feature extraction uses the convolutional neural network CNN algorithm, and the formula used is:

[0017]

[0018] Among them, f(a,b) represents the convolutional kernel, g(i+a,j+b) represents the image, and h(i,j) realizes the multiplication of the convolutional kernel and the corresponding original image and then summation.

[0019] In the above-mentioned seal ink warning detection system based on deep learning, the ink state evaluation uses a binary classification model for numerical evaluation. Among them, the steps include:

[0020] S1: Input the image after feature extraction into the input layer;

[0021] S2: Set the weights and biases of the input nodes;

[0022] S3: Use the sigmoid function classifier to perform binary classification on the image;

[0023] S4: Obtain after comparing the labeled result with the predicted result:

[0024] True positive TP: The number of positive classes correctly predicted by the model;

[0025] False negative FN: The number of positive classes wrongly predicted as negative classes by the model;

[0026] False positive FP: The number of negative classes wrongly predicted as positive classes by the model;

[0027] True negative TN: The number of negative classes correctly predicted by the model;

[0028] Among them, accuracy:

[0029]

[0030] Precision:

[0031]

[0032] Recall:

[0033]

[0034] The evaluation result numerically represents the remaining amount of ink.

[0035] In the above-mentioned seal ink warning detection system based on deep learning, the system also sets a warning threshold. When the evaluation result of the ink state is lower than the threshold, the warning mechanism is triggered, and the information is transmitted to the alarm for alarming. The alarm also includes warning information, where the warning information includes the ink state, the remaining number of uses, and the content of the recommended ink replenishment.

[0036] In the above-mentioned seal ink warning detection system based on deep learning, the system also includes an adaptive learning mechanism that corrects misjudged samples according to user feedback and updates the model through online incremental learning.

[0037] In the above-mentioned seal ink warning detection system based on deep learning, the system is regularly monitored and maintained to ensure its normal operation, and the model is regularly updated and optimized to adapt to the changing seal and ink states.

[0038] In the above-mentioned seal ink warning detection system based on deep learning, in step S1, when collecting the seal image, annotation is carried out simultaneously, including the annotation of the ink state, the type of ink, and the number of ink uses.

[0039] Compared with the existing technologies, the advantages of the present invention are as follows:

[0040] 1. The deep learning algorithm can automatically extract image features and evaluate them without manual intervention. It realizes the real-time monitoring and warning of the ink state, improving work efficiency.

[0041] 2. The deep learning algorithm has high precision and robustness in the field of image recognition. Even in the face of complex seal images and different usage scenarios, it can accurately evaluate the ink state and trigger a warning.

[0042] 3. The deep learning model can be trained and optimized according to actual needs. It can adapt to different types of seals and ink warning requirements, and has high scalability and flexibility. BRIEF DESCRIPTION OF THE DRAWINGS

[0043] Figure 1 It is a flowchart of a seal ink warning detection system based on deep learning proposed by the present invention. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0044] Refer to Figure 1 , a seal ink warning detection system based on deep learning, including the following steps

[0045] Data preparation: Collect seal images in different ink states through a high-definition camera, record the corresponding number of uses with a counter, and at the same time collect seal images in different ink states and annotate the data, including the annotation of the ink state and the annotation of the seal type, etc.;

[0046] Build a model and use the Convolutional Neural Network (CNN) algorithm for feature extraction.

[0047] The formulas used for feature extraction include:

[0048]

[0049] Among them, f(a, b) represents the convolutional kernel, g(i + a, j + b) represents the image, and h(i, j) realizes the multiplication of the convolutional kernel and the corresponding original image and then summation.

[0050] Model training: Input the labeled images into the model for training, and then use a binary classification model for numerical evaluation according to the results. The steps include:

[0051] S1: Input the images after feature extraction into the input layer;

[0052] S2: Set the weights and biases of the input nodes;

[0053] S3: Use the sigmoid function classifier to perform binary classification on the images;

[0054] S4: Obtain the following after comparing the labeled results with the predicted results:

[0055] True Positive (TP): The number of positive classes correctly predicted by the model;

[0056] False Negative (FN): The number of positive classes mispredicted as negative classes by the model;

[0057] False Positive (FP): The number of negative classes mispredicted as positive classes by the model;

[0058] True Negative (TN): The number of negative classes correctly predicted by the model;

[0059] Among them, accuracy:

[0060]

[0061] Precision:

[0062]

[0063] Recall:

[0064]

[0065] F1 is the evaluation result and represents the remaining amount of ink numerically. Ink warning work is realized through the predicted remaining amount.

[0066] Model Deployment: Deploy the trained model to the actual seal ink warning system, test and optimize the system to ensure its stability and accuracy. The system also includes an adaptive learning mechanism that corrects misjudged samples based on user feedback and updates the model through online incremental learning.

[0067] System Monitoring and Maintenance: Regularly monitor and maintain the system to ensure its normal operation. Update and optimize the model according to actual needs to adapt to the changing states of seals and inks.

[0068] As is known by common technical knowledge, the present invention can be implemented by other embodiments that do not depart from its spiritual essence or essential features. Therefore, the above-disclosed embodiments are illustrative in all aspects and not exclusive. All changes within the scope of the present invention or within the scope equivalent to the present invention are encompassed by the present invention.

Claims

1. A seal ink warning detection system based on deep learning, characterized in that, It includes the following parts: Hardware module: The hardware module includes a high-definition camera, a counter, and an alarm. Detection module, and the detection module includes the following steps: S1: Data acquisition and annotation: Collect seal images under different ink states through a high-definition camera, and record the corresponding number of seal impressions through a counter. S2: Preprocess the collected data, including denoising, grayscale conversion, and binarization, to improve the image quality. S3: Build a deep learning model, input the preprocessed data into the model, and use deep learning algorithms to extract features from the image. The extracted features include the outline, texture, and color of the seal. S4: Ink state evaluation. According to the extracted features, evaluate the ink state through a deep learning model.

2. The seal ink warning detection system based on deep learning according to claim 1, characterized in that: In step S1, annotation is performed while collecting the seal image, including ink state, ink type, and number of ink applications.

3. The seal ink warning detection system based on deep learning according to claim 1, wherein: Feature extraction uses the convolutional neural network CNN algorithm, and the formula used is: Among them, f(a,b) represents the convolutional kernel, g(i+a,j+b) represents the picture, and h(i,j) realizes the multiplication of the convolutional kernel and the original picture and then summation.

4. The seal ink warning detection system based on deep learning according to claim 2, characterized in that: The ink state evaluation is trained using a binary classification model, and the steps include: S1: Input the picture after feature extraction into the input layer. S2: Set the weights and biases of the input nodes. S3: Use the sigmoid function classifier to perform binary classification on the picture. S4: Obtain after comparing the annotation result with the prediction result: True positive TP: The number of positive classes correctly predicted by the model. False negative FN: The number of positive classes wrongly predicted as negative classes by the model. False positive FP: The number of negative classes wrongly predicted as positive classes by the model. True negative TN: The number of negative classes correctly predicted by the model. Among them, accuracy: Precision: Recall: F1 is the evaluation result and represents the remaining amount of ink numerically.

5. The seal ink warning detection system based on deep learning according to claim 1, wherein: The system also sets a warning threshold. When the ink state evaluation result is lower than the threshold, the warning mechanism is triggered, and the information is transmitted to the alarm for alarming. The alarm also includes warning information, where the warning information includes ink state, remaining usage times, and content suggesting ink replenishment.

6. The seal ink warning detection system based on deep learning according to claim 1, characterized in that: The system also includes an adaptive learning mechanism, which corrects misjudged samples according to user feedback and updates the model through online incremental learning.

7. A stamp ink warning detection system based on deep learning according to claim 1, characterized in that: Regularly monitor and maintain the system, and regularly update and optimize the model to adapt to the changing seal and ink states.

Images