Data security monitoring method and monitoring system based on large model

Abstract

The present invention relates to a data security monitoring method and a monitoring system based on a large model, which belongs to the field of monitoring technology and includes the following steps: S1, sensing whether the monitor body is located in a protective shell, if so, proceeding to step S2; S2, sensing whether the notch of the positioning slot is in contact with the positioning plate, if so, proceeding to step S4, otherwise proceeding to step S3; S3, detecting whether the positioning plate is damaged; S4, detecting whether the card plate moves, if so, proceeding to step S5; S5, detecting and judging whether the movement process of the card plate is completed, if so, proceeding to step S6; S6, detecting whether the card plate is completely separated from the monitor body, if so, the monitor body can be taken out, otherwise, the monitor body cannot be taken out temporarily and manual troubleshooting is required. An automated monitoring solution is designed for inspection and maintenance of the monitoring device, so that when the monitor body is removed, the monitoring device will not be damaged due to improper removal.

Description

Technical Field

[0001] The present invention belongs to the field of monitoring technology, and in particular relates to a data security monitoring method and a monitoring system based on a large model. Background Art

[0002] A monitoring device based on network data security is commonly used at present, which includes a base, the upper surface of which is fixedly connected to a protective shell, the upper surface of which is clamped with a monitor body, a mounting slot provided in the base, a transmission slot provided on the lower surface of the inner wall of the mounting slot, and a screw rod hinged in the transmission slot through a bearing. This monitoring device based on network data security is provided with a screw rod, a screw rod nut, a vertical rod, a moving block, a clamping plate and a clamping slot, so that the monitor body in this device is fixed with the base by clamping, and the monitor body can be quickly disassembled and assembled by simply turning the handle, which can not only effectively ensure the installation stability of the monitor body, but also can be quickly removed when the monitor body needs to be inspected and maintained, thereby effectively shortening the time required for a single inspection and improving the inspection and maintenance efficiency of the monitor body.

[0003] However, there is currently no automated monitoring solution for the inspection and maintenance of monitoring devices. For example, when the monitor body is removed, it is very likely to be damaged if the removal is improper.

[0004] Therefore, at this stage, it is necessary to design data security monitoring methods and monitoring systems based on large models to solve the above problems. Summary of the Invention

[0005] The purpose of the present invention is to provide a data security monitoring method and monitoring system based on a large model to solve the technical problems existing in the above-mentioned prior art. At present, there is no automated monitoring solution for the inspection and maintenance of monitoring devices. For example, when the monitor body is removed, if the removal is improper, it is very likely to damage the monitoring device.

[0006] To achieve the above object, the technical solution of the present invention is:

[0007] The data security monitoring method based on a large model includes the following steps:

[0008] S1, detecting whether the monitoring body is located in the protective shell, if so, proceeding to step S2;

[0009] S2, detecting whether the notch of the positioning groove is in contact with the positioning plate, if so, proceeding to step S4, otherwise proceeding to step S3;

[0010] S3. Check whether the positioning plate is damaged;

[0011] S4, check whether the card board moves, if so, proceed to step S5;

[0012] S5, detecting and judging whether the moving process of the card board is completed, if so, proceeding to step S6;

[0013] S6. Check whether the card board is completely separated from the monitor body. If so, the monitor body can be taken out. Otherwise, the monitor body cannot be taken out temporarily and manual troubleshooting is required.

[0014] Furthermore, in step S3, the specific position where the notch of the positioning groove does not contact the positioning plate is first determined, and then whether the positioning plate at the specific position is damaged is detected. If not damaged, whether the positioning plate is worn is detected.

[0015] Furthermore, the large model is combined with the sound to determine whether the movement process of the pallet is completed, as follows:

[0016] Data collection: Collect sound samples when the card is moving and when it stops. Ensure that a preset number of samples are collected in a real environment, including movement sounds at different speeds and positions, as well as silence or background noise when the card is stopped.

[0017] Data preprocessing: converting sound signals into a format suitable for model input; extracting sound features to capture the spectral characteristics of the sound;

[0018] Dataset construction: Pair the processed sound features with labels and divide them into training, validation, and test sets. Labels include: moving and stopping.

[0019] Model selection: RNN, CNN, CRNN or Transformer;

[0020] Model training: Train the model on the training set and adjust the hyperparameters on the validation set;

[0021] Model evaluation: Evaluate model performance on the test set to ensure that the model can accurately determine whether the card has stopped;

[0022] Deployment: Deploy the trained model to the actual environment to monitor the sound in real time and determine the status of the card board.

[0023] Furthermore, combined with the large model, image recognition is used to detect whether the card board and the monitor body are completely separated, as follows:

[0024] Image acquisition:

[0025] Ensure a stable lighting environment to avoid reflections and shadows;

[0026] Use a fixed-position image acquisition unit to ensure consistent shooting angle and distance each time;

[0027] Image preprocessing:

[0028] Denoising: Use Gaussian filtering, median filtering, etc. to remove image noise;

[0029] Grayscale: Convert a color image to a grayscale image;

[0030] Contrast enhancement: Use methods such as histogram equalization to enhance the contrast of the edges of the card, the main structure, and potential gaps;

[0031] ROI selection: focus on the key area of ​​the card insertion interface;

[0032] Feature extraction and separation judgment:

[0033] Use one or more of the following methods in combination:

[0034] a. Edge detection and gap analysis:

[0035] Use edge detection algorithm to extract significant edges of ROI area;

[0036] Objective: Detect whether there is a continuous gap edge between the edge of the card board and the edge of the monitor body, with a width exceeding the threshold;

[0037] method:

[0038] In the reference image, the edge of the cardboard and the edge of the body are closely fitted or partially overlapped;

[0039] In the image to be tested, if the card board is separated, a new, relatively parallel gap edge line will be generated between its edge and the edge of the body;

[0040] The algorithm detects the existence, continuity, and average width of this gap line;

[0041] If the gap is continuous and the width is greater than the preset threshold, it is determined to be separated;

[0042] b. Feature point matching and displacement detection:

[0043] Extract stable feature points in the ROI area of ​​the reference image and the corresponding ROI area of ​​the image to be tested;

[0044] Perform feature point matching;

[0045] analyze:

[0046] If the pallet is not separated, the position deviation between the matching point pairs is within the set range;

[0047] If the pallet is completely separated, the feature points of the pallet area will undergo a preset displacement of integrity and consistency relative to the feature points of the body area;

[0048] Calculation: Calculate the average displacement vector or affine transformation matrix of the matching point pair; if the displacement exceeds the threshold or the transformation matrix shows obvious translation / rotation, it is judged as separation;

[0049] c. Contour analysis / Blob analysis:

[0050] Perform binary segmentation on the pre-processed image to separate the card board and the body area;

[0051] Perform morphological operations to optimize the segmentation results, remove noise, and connect broken edges;

[0052] Extract the outline of the pallet;

[0053] analyze:

[0054] Calculate the overlapping area or Hausdorff distance between the cardboard contour and the reference contour of the body;

[0055] If the card is inserted correctly, the overlapping area should meet a certain value;

[0056] If the pallets are separated, the overlapping area meets the threshold area;

[0057] Judgment: If the overlapping area is less than the threshold, it is judged as separation.

[0058] The data security monitoring system based on the large model includes a body sensing device, a slot contact sensing device, a positioning plate damage detection device, a card board movement detection device, a movement stop judgment device, a card board separation detection device, a voice broadcast device, and a main control device, wherein the main control device is respectively connected to the body sensing device, the slot contact sensing device, the positioning plate damage detection device, the card board movement detection device, the movement stop judgment device, the card board separation detection device, and the voice broadcast device;

[0059] The body sensing device is used to sense whether the monitor body is located in the protective shell; the slot contact sensing device is used to sense whether the slot of the positioning slot is in contact with the positioning plate; the positioning plate damage detection device is used to detect whether the positioning plate is damaged; the card movement detection device is used to detect whether the card moves; the movement stop judgment device is used to detect whether the movement process of the card is completed; the card separation detection device is used to detect whether the card is completely separated from the monitor body; the voice broadcast device is used to voice broadcast warning information to maintenance personnel.

[0060] Furthermore, the main control device controls the initial state of the main body sensing device to be turned on; controls the initial states of the notch contact sensing device, the positioning plate damage detection device, the card movement detection device, the movement stop judgment device, the card separation detection device, and the voice broadcast device to be turned off;

[0061] When the body sensing device senses that the monitor body is located within the protective shell, the main control device controls the slot contact sensing device to turn on;

[0062] When the notch contact sensing device senses that the notch of the positioning slot contacts the positioning plate, the main control device controls the positioning plate damage detection device to turn on;

[0063] When the positioning plate damage detection device detects that the positioning plate is damaged, the main control device controls the card plate movement detection device;

[0064] When the card board movement detection device detects that the card board has moved, the main control device controls the movement stop judgment device to turn on;

[0065] When the movement stop judgment device detects that the movement process of the card board is completed, the main control device controls the card board separation detection device to start;

[0066] When the card board separation detection device detects that the card board is not completely separated from the monitor body, the main control device controls the voice broadcast device to broadcast a warning message that the monitor body cannot be taken out temporarily and manual troubleshooting is required.

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

[0068] One of the beneficial effects of this solution is that an automated monitoring solution is designed for inspection and maintenance of the monitoring device, so that when the monitor body is removed, the monitoring device will not be damaged due to improper removal. BRIEF DESCRIPTION OF THE DRAWINGS

[0069] Figure 1 This is a schematic diagram of the method steps of an embodiment of the implementation of this solution. DETAILED DESCRIPTION

[0070] Example: Data security monitoring method based on large model, such as Figure 1 As shown, the following steps are included:

[0071] S1, detecting whether the monitoring body is located in the protective shell, if so, proceeding to step S2;

[0072] S2, detecting whether the notch of the positioning groove is in contact with the positioning plate, if so, proceeding to step S4, otherwise proceeding to step S3;

[0073] S3. Check whether the positioning plate is damaged;

[0074] S4, check whether the card board moves, if so, proceed to step S5;

[0075] S5, detecting and judging whether the moving process of the card board is completed, if so, proceeding to step S6;

[0076] S6. Check whether the card board is completely separated from the monitor body. If so, the monitor body can be taken out. Otherwise, the monitor body cannot be taken out temporarily and manual troubleshooting is required.

[0077] Furthermore, in step S3, the specific position where the notch of the positioning groove does not contact the positioning plate is first determined, and then whether the positioning plate at the specific position is damaged is detected. If not damaged, whether the positioning plate is worn is detected.

[0078] Furthermore, the large model is combined with the sound to determine whether the movement process of the pallet is completed, as follows:

[0079] Data collection: Collect sound samples when the card is moving and when it stops. Ensure that a preset number of samples are collected in a real environment, including movement sounds at different speeds and positions, as well as silence or background noise when the card is stopped.

[0080] Data preprocessing: converting sound signals into a format suitable for model input; extracting sound features to capture the spectral characteristics of the sound;

[0081] Dataset construction: Pair the processed sound features with labels and divide them into training, validation, and test sets. Labels include: moving and stopping.

[0082] Model selection: RNN, CNN, CRNN or Transformer;

[0083] Model training: Train the model on the training set and adjust the hyperparameters on the validation set;

[0084] Model evaluation: Evaluate model performance on the test set to ensure that the model can accurately determine whether the card has stopped;

[0085] Deployment: Deploy the trained model to the actual environment to monitor the sound in real time and determine the status of the card board.

[0086] The specific implementation is as follows:

[0087] 1. Data Collection:

[0088] - Use a microphone to record the sounds of the board moving and stopping.

[0089] - Moving state: Record the entire process of the card from starting to stopping, including the collision sound during movement.

[0090] - Stop state: records the sound after the card is stopped, which may include background noise.

[0091] - Note: It is necessary to collect sounds under different working conditions to enhance the generalization ability of the model.

[0092] 2. Data preprocessing:

[0093] - Read audio files (such as .wav format), usually using a 16kHz sampling rate.

[0094] - Split the audio into small time segments (e.g. 0.5 seconds each) and mark whether the segment is moving or stopped. Note that the stopped state sound segment should be taken from the time period after the card is stopped.

[0095] - Extract MFCC features: Multiple MFCC coefficients (e.g., 40) can be extracted for each segment and formed into a two-dimensional matrix (time frame × MFCC coefficients). Other features such as mel-spectrograms can also be used.

[0096] - Normalization: Normalize the features.

[0097] 3. Dataset construction:

[0098] - The feature matrix and label (0 for stop, 1 for movement) corresponding to each sound clip are combined into a sample.

[0099] - Divide the training set, validation set and test set into proportions (e.g. 70% training, 15% validation, 15% test).

[0100] 4. Model construction:

[0101] - Using CNN: Treat MFCC features as images (timeframe as height, MFCC coefficients as width, number of channels as 1). Use several convolutional and pooling layers, followed by fully connected layers.

[0102] - Using RNN: The features of each time frame are treated as a time step of the sequence, input into the RNN (such as LSTM layer), and then connected to the fully connected layer.

[0103] - CRNN can also be used: first use CNN to extract high-level features, and then input the feature sequence into RNN.

[0104] 5. Model training:

[0105] - Loss function: For binary classification problems, binary cross entropy loss is used.

[0106] - Optimizer: Adam.

[0107] - Monitor the accuracy and loss on the validation set during training to avoid overfitting (you can use early stopping and dropout).

[0108] 6. Model Evaluation:

[0109] - Use the test set to evaluate and calculate indicators such as accuracy, precision, recall, and F1 value.

[0110] - Ensure that the model performs well on unseen data.

[0111] 7. Deployment:

[0112] - In a real-world environment, the sound is collected in real time through a microphone, and features are extracted using the same preprocessing method. The features are then input into the model for real-time classification.

[0113] - A threshold can be set. For example, if multiple consecutive clips are classified as stopped, it is determined that the card has stopped to avoid occasional misjudgments.

[0114] challenge:

[0115] - Environmental noise: Various noises may exist in the actual environment, so it is necessary to add noise enhancement to the training data or use a noise reduction algorithm.

[0116] - Real-time: The model must be able to process data quickly, so it should not be too complex.

[0117] Example code skeleton (using TensorFlow and Keras):

[0118] Here is a simple example using a CNN model:

[0119] Step 1: Import the library

[0120] """

[0121] import numpy as np

[0122] import tensorflow as tf

[0123] from tensorflow.keras import layers, models

[0124] import librosa

[0125] import os

[0126] # Step 2: Data loading and preprocessing

[0127] def load_audio_files(folder, label, max_files=None):

[0128] """

[0129] Load the audio files in the specified folder, extract MFCC features, and label them.

[0130] Each audio file is split into multiple segments (e.g. one segment every 0.5 seconds)

[0131] """

[0132] data = []

[0133] labels = []

[0134] files = os.listdir(folder)[:max_files]

[0135] for file in files:

[0136] file_path = os.path.join(folder, file)

[0137] # Load the audio file

[0138] y, sr = librosa.load(file_path, sr=16000) # Set the sampling rate to 16kHz

[0139] # Split the audio into 0.5 second segments

[0140] segment_length = int(0.5 * sr) # Number of sample points in 0.5 seconds

[0141] n_segments = len(y) / / segment_length

[0142] for i in range(n_segments):

[0143] segment = y[i*segment_length:(i+1)*segment_length]

[0144] # Extract MFCC features, assuming 40 coefficients per frame

[0145] mfcc = librosa.feature.mfcc(y=segment, sr=sr, n_mfcc=40)

[0146] # Transpose so that the time step is in the first dimension (for convenience of CNN processing, the time step is used as height and the feature is used as width)

[0147] mfcc = mfcc.T # Now the shape is (timesteps, 40)

[0148] # Normalization

[0149] mfcc = (mfcc - np.mean(mfcc)) / np.std(mfcc)

[0150] data.append(mfcc)

[0151] labels.append(label)

[0152] return data, labels

[0153] # Assume there are two folders: 'moving_sounds' and 'stopped_sounds'

[0154] moving_data, moving_labels = load_audio_files('moving_sounds', 1)

[0155] stopped_data, stopped_labels = load_audio_files('stopped_sounds', 0)

[0156] # Merge data

[0157] X = moving_data + stopped_data

[0158] y = moving_labels + stopped_labels

[0159] # Convert to numpy array and fill it to the same length (because the number of time steps in different segments may be different, but usually the number of time steps extracted by MFCC is fixed, if it is not fixed, it needs to be filled)

[0160] # First find the maximum number of time steps

[0161] max_len = max([x.shape[0] for x in X])

[0162] # Fill with 0 to max_len

[0163] X_padded = np.array([np.pad(x, ((0, max_len - x.shape[0]), (0,0)),mode='constant') for x in X])

[0164] # At this time, the shape of X_padded is (number of samples, time step max_len, 40)

[0165] # Increase the channel dimension, because CNN requires channels

[0166] X_padded = X_padded[..., np.newaxis] # shape becomes (number of samples, time step max_len, 40, 1)

[0167] y = np.array(y)

[0168] # Divide the dataset

[0169] from sklearn.model_selection import train_test_split

[0170] X_train, X_test, y_train, y_test = train_test_split(X_padded, y,test_size=0.2, random_state=42)

[0171] X_train, X_val, y_train, y_val = train_test_split(X_train, y_train,test_size=0.2, random_state=42)

[0172] # Step 3: Build CNN model

[0173] model = models.Sequential()

[0174] model.add(layers.Conv2D(32, (3, 3), activation='relu', input_shape=(max_len, 40, 1)))

[0175] model.add(layers.MaxPooling2D((2, 2)))

[0176] model.add(layers.Conv2D(64, (3, 3), activation='relu'))

[0177] model.add(layers.MaxPooling2D((2, 2)))

[0178] model.add(layers.Conv2D(64, (3, 3), activation='relu'))

[0179] model.add(layers.Flatten())

[0180] model.add(layers.Dense(64, activation='relu'))

[0181] model.add(layers.Dense(1, activation='sigmoid'))

[0182] model.compile(optimizer='adam',

[0183] loss='binary_crossentropy',

[0184] metrics=['accuracy'])

[0185] # Train the model

[0186] history = model.fit(X_train, y_train, epochs=10,

[0187] validation_data=(X_val, y_val))

[0188] # Evaluate the model

[0189] test_loss, test_acc = model.evaluate(X_test, y_test)

[0190] print(f'Test accuracy: {test_acc}')

[0191] # Save the model

[0192] model.save('cardboard_sound_classifier.h5')

[0193] # During deployment, collect 0.5 seconds of audio in real time, extract MFCC features, fill them to max_len, increase the channel dimension, and then input them into the model for prediction.

[0194] Note: In practical applications, it may be necessary to optimize the model, such as adjusting the network structure, hyperparameters, and dealing with data imbalance.

[0195] Alternatively, you can also try using the RNN model, for example:

[0196] model = models.Sequential()

[0197] model.add(layers.LSTM(64, input_shape=(max_len, 40)))

[0198] model.add(layers.Dense(1, activation='sigmoid'))

[0199] Or using a 1D CNN (convolution in the time dimension):

[0200] model = models.Sequential()

[0201] model.add(layers.Conv1D(32, 3, activation='relu', input_shape=(max_len, 40)))

[0202] model.add(layers.MaxPooling1D(2))

[0203] model.add(layers.Conv1D(64, 3, activation='relu'))

[0204] model.add(layers.GlobalMaxPooling1D())

[0205] model.add(layers.Dense(1, activation='sigmoid'))

[0206] Select the best model based on actual performance.

[0207] The above is an example of a framework, which needs to be adjusted according to specific circumstances in actual applications.

[0208] Furthermore, combined with the large model, image recognition is used to detect whether the card board and the monitor body are completely separated, as follows:

[0209] Image acquisition:

[0210] Ensure a stable lighting environment to avoid reflections and shadows;

[0211] Use a fixed-position image acquisition unit to ensure consistent shooting angle and distance each time;

[0212] Image preprocessing:

[0213] Denoising: Use Gaussian filtering, median filtering, etc. to remove image noise;

[0214] Grayscale: Convert a color image to a grayscale image;

[0215] Contrast enhancement: Use methods such as histogram equalization to enhance the contrast of the edges of the card, the main structure, and potential gaps;

[0216] ROI selection: focus on the key area of ​​the card insertion interface;

[0217] Feature extraction and separation judgment:

[0218] Use one or more of the following methods in combination:

[0219] a. Edge detection and gap analysis:

[0220] Use edge detection algorithm to extract significant edges of ROI area;

[0221] Objective: Detect whether there is a continuous gap edge between the edge of the card board and the edge of the monitor body, with a width exceeding the threshold;

[0222] method:

[0223] In the reference image, the edge of the cardboard and the edge of the body are closely fitted or partially overlapped;

[0224] In the image to be tested, if the card board is separated, a new, relatively parallel gap edge line will be generated between its edge and the edge of the body;

[0225] The algorithm detects the existence, continuity, and average width of this gap line;

[0226] If the gap is continuous and the width is greater than the preset threshold, it is determined to be separated;

[0227] b. Feature point matching and displacement detection:

[0228] Extract stable feature points in the ROI area of ​​the reference image and the corresponding ROI area of ​​the image to be tested;

[0229] Perform feature point matching;

[0230] analyze:

[0231] If the pallet is not separated, the position deviation between the matching point pairs is within the set range;

[0232] If the pallet is completely separated, the feature points of the pallet area will undergo a preset displacement of integrity and consistency relative to the feature points of the body area;

[0233] Calculation: Calculate the average displacement vector or affine transformation matrix of the matching point pair; if the displacement exceeds the threshold or the transformation matrix shows obvious translation / rotation, it is judged as separation;

[0234] c. Contour analysis / Blob analysis:

[0235] Perform binary segmentation on the pre-processed image to separate the card board and the body area;

[0236] Perform morphological operations to optimize the segmentation results, remove noise, and connect broken edges;

[0237] Extract the outline of the pallet;

[0238] analyze:

[0239] Calculate the overlapping area or Hausdorff distance between the cardboard contour and the reference contour of the body;

[0240] If the card is inserted correctly, the overlapping area should meet a certain value;

[0241] If the pallets are separated, the overlapping area meets the threshold area;

[0242] Judgment: If the overlapping area is less than the threshold, it is judged as separation.

[0243] The data security monitoring system based on the large model includes a body sensing device, a slot contact sensing device, a positioning plate damage detection device, a card board movement detection device, a movement stop judgment device, a card board separation detection device, a voice broadcast device, and a main control device, wherein the main control device is respectively connected to the body sensing device, the slot contact sensing device, the positioning plate damage detection device, the card board movement detection device, the movement stop judgment device, the card board separation detection device, and the voice broadcast device;

[0244] The body sensing device is used to sense whether the monitor body is located in the protective shell; the slot contact sensing device is used to sense whether the slot of the positioning slot is in contact with the positioning plate; the positioning plate damage detection device is used to detect whether the positioning plate is damaged; the card movement detection device is used to detect whether the card moves; the movement stop judgment device is used to detect whether the movement process of the card is completed; the card separation detection device is used to detect whether the card is completely separated from the monitor body; the voice broadcast device is used to voice broadcast warning information to maintenance personnel.

[0245] Furthermore, the main control device controls the initial state of the main body sensing device to be turned on; controls the initial states of the notch contact sensing device, the positioning plate damage detection device, the card movement detection device, the movement stop judgment device, the card separation detection device, and the voice broadcast device to be turned off;

[0246] When the body sensing device senses that the monitor body is located within the protective shell, the main control device controls the slot contact sensing device to turn on;

[0247] When the notch contact sensing device senses that the notch of the positioning slot contacts the positioning plate, the main control device controls the positioning plate damage detection device to turn on;

[0248] When the positioning plate damage detection device detects that the positioning plate is damaged, the main control device controls the card plate movement detection device;

[0249] When the card board movement detection device detects that the card board has moved, the main control device controls the movement stop judgment device to turn on;

[0250] When the movement stop judgment device detects that the movement process of the card board is completed, the main control device controls the card board separation detection device to start;

[0251] When the card board separation detection device detects that the card board is not completely separated from the monitor body, the main control device controls the voice broadcast device to broadcast a warning message that the monitor body cannot be taken out temporarily and manual troubleshooting is required.

[0252] The logical sequence is: check if the body is in the shell → check if the notch contacts → check if the positioning plate is damaged → detect if the card moves → determine if the movement stops → detect if the card is separated.

[0253] The system includes the following devices, each of which has a specific function and is controlled by the main control device:

[0254] Main body sensing device: detects whether the monitor body is inside the protective shell. The initial state is open.

[0255] Notch contact sensing device: Detects whether the notch of the positioning slot is in contact with the positioning plate. The initial state is closed and it only opens when the body sensing device detects that the body is inside the shell.

[0256] Positioning plate damage detection device: Detects whether the positioning plate is damaged (such as deformed or broken). The initial state is closed and it only opens when the notch contact sensing device detects contact.

[0257] Pallet movement detection device: Detects whether the pallet has moved. Initially closed, it is only opened when the positioning plate damage detection device detects damage.

[0258] Movement stop judgment device: determines whether the movement process of the pallet is completed. The initial state is closed and it is only turned on when the pallet movement detection device detects movement.

[0259] Card separation detection device: detects whether the card is completely separated from the monitor body. The initial state is closed and it is only turned on when the movement stop judgment device determines that the movement has ended.

[0260] Voice broadcast device: broadcasts warning information to maintenance personnel. The initial state is off, and the broadcast is triggered only when the card separation detection device detects that the separation is not complete.

[0261] Master device: As the core controller, it manages the initial state and state switching logic of all devices.

[0262] The specific logic is as follows:

[0263] start

[0264] │

[0265] ├─ The main body sensor is turned on (initial state)

[0266] │ ├─ If the monitor body is detected in the protective shell → open the slot contact sensing device

[0267] │ └─ If not detected (remain closed or alarm)

[0268] │

[0269] ├─ Notch contact sensing device is turned on

[0270] │ ├─ If the notch is detected to be in contact with the positioning plate → turn on the positioning plate damage detection device

[0271] │ └─ If no contact is detected (do not trigger subsequent steps)

[0272] │

[0273] ├─ Positioning plate damage detection device is turned on

[0274] │ ├─ If the positioning plate is detected to be damaged → turn on the card movement detection device

[0275] │ └─ If not damaged (skip subsequent or allow direct separation)

[0276] │

[0277] ├─ Pallet movement detection device is turned on

[0278] │ ├─ If the card movement is detected → start the movement stop judgment device

[0279] │ └─ If no movement (waiting or alarm)

[0280] │

[0281] ├─ Movement stop judgment device is turned on

[0282] │ ├─ If it is judged that the movement is completed → turn on the card separation detection device

[0283] │ └─ If the movement is not completed (continue detection)

[0284] │

[0285] ├─ Pallet separation detection device is turned on

[0286] │ ├─ If it is detected that the card board is not completely separated from the monitor body → the main control turns on the voice broadcast device

[0287] │ └─ Voice broadcast: "The monitor body cannot be removed temporarily. Manual troubleshooting is required."

[0288] │ └─ If completely separated (allowing safe removal of the monitor body)

[0289] │

[0290] Finish.

Claims

1. A data security monitoring method based on a large model, characterized in that: The following steps are involved: S1, detecting whether the monitoring body is located in the protective shell, if so, proceeding to step S2; S2, detecting whether the notch of the positioning groove is in contact with the positioning plate, if so, proceeding to step S4, otherwise proceeding to step S3; S3. Check whether the positioning plate is damaged; S4, check whether the card board moves, if so, proceed to step S5; S5, detecting and judging whether the moving process of the card board is completed, if so, proceeding to step S6; S6. Check whether the card board is completely separated from the monitor body. If so, the monitor body can be removed. Otherwise, the monitor body cannot be removed temporarily and manual troubleshooting is required. Among them, in step S5, the large model is combined with the sound to determine whether the movement process of the card board is completed, as follows: Data collection: Collect sound samples when the card is moving and when it stops. Ensure that a preset number of samples are collected in a real environment, including movement sounds at different speeds and positions, as well as silence or background noise when the card is stopped. Data preprocessing: converting sound signals into a format suitable for model input; extracting sound features to capture the spectral characteristics of the sound; Dataset construction: Pair the processed sound features with labels and divide them into training, validation, and test sets. Labels include: moving and stopping. Model selection: RNN, CNN, CRNN or Transformer; Model training: Train the model on the training set and adjust the hyperparameters on the validation set; Model evaluation: Evaluate model performance on the test set to ensure that the model can accurately determine whether the card has stopped; Deployment: Deploy the trained model to the actual environment to monitor the sound and determine the status of the card in real time; In step S6, image recognition is used to detect whether the card board and the monitor body are completely separated in combination with the large model, as follows: Image acquisition: Ensure a stable lighting environment to avoid reflections and shadows; Use a fixed-position image acquisition unit to ensure consistent shooting angle and distance each time; Image preprocessing: Denoising: Use Gaussian filtering and median filtering to remove image noise; Grayscale: Convert a color image to a grayscale image; Contrast enhancement: Use histogram equalization to enhance the contrast of the edges, structure, and potential gaps of the cardboard; ROI selection: focus on the key area of ​​the card insertion interface; Feature extraction and separation judgment: Use one or more of the following methods in combination: a. Edge detection and gap analysis: Use edge detection algorithm to extract significant edges of ROI area; Objective: Detect whether there is a continuous gap edge between the edge of the card board and the edge of the monitor body, with a width exceeding the threshold; method: In the reference image, the edge of the cardboard and the edge of the body are closely fitted or partially overlapped; In the image to be tested, if the card board is separated, a new, relatively parallel gap edge line will be generated between its edge and the edge of the body; The algorithm detects the existence, continuity, and average width of this gap line; If the gap is continuous and the width is greater than the preset threshold, it is determined to be separated; b. Feature point matching and displacement detection: Extract stable feature points in the ROI area of ​​the reference image and the corresponding ROI area of ​​the image to be tested; Perform feature point matching; analyze: If the pallet is not separated, the position deviation between the matching point pairs is within the set range; If the pallet is completely separated, the feature points of the pallet area will undergo a preset displacement of integrity and consistency relative to the feature points of the body area; Calculation: Calculate the average displacement vector or affine transformation matrix of the matching point pair; if the displacement exceeds the threshold or the transformation matrix shows obvious translation / rotation, it is judged as separation; c. Contour analysis / Blob analysis: Perform binary segmentation on the pre-processed image to separate the card board and the body area; Perform morphological operations to optimize the segmentation results, remove noise, and connect broken edges; Extract the outline of the pallet; analyze: Calculate the overlapping area or Hausdorff distance between the cardboard contour and the reference contour of the body; If the card is inserted correctly, the overlapping area should meet a certain value; If the pallets are separated, the overlapping area meets the threshold area; Judgment: If the overlapping area is less than the threshold, it is judged as separation.

2. The data security monitoring method based on a large model according to claim 1 is characterized in that: In step S3, the specific position where the notch of the positioning groove does not contact the positioning plate is first determined, and then the positioning plate at the specific position is detected to see if it is damaged. If not, the positioning plate is detected to see if it is worn.

3. The data security monitoring system based on the large model is characterized by: The data security monitoring method based on the large model as claimed in claim 1 is used to perform data security monitoring, comprising a body sensing device, a notch contact sensing device, a positioning plate damage detection device, a card movement detection device, a movement stop judgment device, a card separation detection device, a voice broadcast device, and a main control device, wherein the main control device is connected to the body sensing device, the notch contact sensing device, the positioning plate damage detection device, the card movement detection device, the movement stop judgment device, the card separation detection device, and the voice broadcast device respectively; The body sensing device is used to sense whether the monitor body is located in the protective shell; the slot contact sensing device is used to sense whether the slot of the positioning slot is in contact with the positioning plate; the positioning plate damage detection device is used to detect whether the positioning plate is damaged; the card movement detection device is used to detect whether the card moves; the movement stop judgment device is used to detect whether the movement process of the card is completed; the card separation detection device is used to detect whether the card is completely separated from the monitor body; the voice broadcast device is used to voice broadcast warning information to maintenance personnel.

4. The data security monitoring system based on a large model according to claim 3 is characterized in that: The main control device controls the initial state of the main body sensing device to be turned on; controls the initial states of the notch contact sensing device, the positioning plate damage detection device, the card movement detection device, the movement stop judgment device, the card separation detection device, and the voice broadcast device to be turned off; When the body sensing device senses that the monitor body is located within the protective shell, the main control device controls the slot contact sensing device to turn on; When the notch contact sensing device senses that the notch of the positioning slot contacts the positioning plate, the main control device controls the positioning plate damage detection device to turn on; When the positioning plate damage detection device detects that the positioning plate is damaged, the main control device controls the card plate movement detection device; When the card board movement detection device detects that the card board has moved, the main control device controls the movement stop judgment device to turn on; When the movement stop judgment device detects that the movement process of the card board is completed, the main control device controls the card board separation detection device to start; When the card board separation detection device detects that the card board is not completely separated from the monitor body, the main control device controls the voice broadcast device to broadcast a warning message that the monitor body cannot be taken out temporarily and manual troubleshooting is required.

Images