Base station data acquisition system and device for coal mine pithead multi-system fusion access

Abstract

The invention discloses a coal mine pithead multi-system fusion access base station data acquisition system and device, and relates to the detection field, and the system comprises a distributed acquisition module and a central analysis module; the distributed acquisition module is located at an entrance of a mine pit, senses an SIM card and an access target, generates registration information and a monitoring video, encrypts and transmits the registration information and the monitoring video to the central analysis module, and displays the counted number of people in real time; the central analysis module performs decryption to obtain the registration information and the monitoring video, analyzes the monitoring video by adopting a target detection model, determines a target type, determines a target behavior based on the registration information, and performs retrieval verification; selecting a frame screening algorithm based on a target type to extract a frame image, carrying out auxiliary verification on a target by adopting a face recognition model or an optical character model, generating a behavior record, generating an early warning prompt when verification fails, updating the statistical number of people based on a target behavior, and feeding back the statistical number of people to the distributed acquisition system. Precise detection of targets passing through a mine pit opening and real-time statistics of personnel in a mine are achieved, and the mine pit access management level is improved.

Description

Technical Field

[0001] The present invention relates to the field of detection, and particularly to a base station data acquisition system and device for multi-system fusion access at a coal mine pithead. Background Art

[0002] The safety production of coal mines has always been of top priority. Accurately grasping the entry and exit situation of personnel or vehicles is the key foundation for ensuring safety production. In the past, there were many problems to be solved in the traditional access management methods. It was impossible to identify and record the personnel or vehicles entering the pit in real time and accurately, there were situations of missing and inaccurate information, lack of statistical analysis function, unable to obtain the number of people in the mine pit in real time, and lack of means for identifying and alarming unauthorised personnel entering the well. Summary of the Invention

[0003] Aiming at the deficiencies of the existing technology, the present invention provides a base station data acquisition system and device for multi-system fusion access at a coal mine pithead to improve the real-time monitoring and management level of the entry and exit of personnel or vehicles.

[0004] The technical solution for achieving the object of the present invention is as follows:

[0005] A base station data acquisition system for multi-system fusion access at a coal mine pithead includes M distributed acquisition modules S1,…,S M and a central analysis module S0;

[0006] The M distributed acquisition modules respectively monitor M coal mine pitheads. The m-th distributed acquisition module S m responds to the access of the SIM card and generates registration information senses the entering and exiting personnel or vehicles and acquires monitoring videos encrypts the registration information and the monitoring videos generates an uplink ciphertext decrypts the downlink ciphertext acquires the statistical number of people and displays it, where m = 1,…,M and M is the total number of coal mine pitheads;

[0007] The central analysis module S0 decrypts the uplink ciphertext sent by the m-th distributed acquisition module S m obtains the registration information and the monitoring videos analyses the monitoring videos based on an object detection model to clarify the object type. Based on the registration information clarifies the object behavior and conducts retrieval verification. After the retrieval verification passes, a face image f or license plate information y is obtained. Different frame screening algorithms are selected based on the object type from the monitoring videos ​Extract frame images and use a face recognition model or an optical character model for auxiliary verification, and generate a behavior record r in combination with the system time t m (t) and decide whether to generate a warning prompt and update the statistical number of people And encrypt and generate the downlink ciphertext corresponding to M distributed acquisition modules respectively Encrypt and store the behavior record r m (t) and provide access verification for query, the target types include personnel and vehicles, and the target behaviors include entering the mine and leaving the mine

[0008] Furthermore, the M distributed acquisition module units are designed identically. The mth distributed acquisition module S m Includes a SIM response unit A video acquisition unit An uplink communication unit And a display unit

[0009] SIM response unit Based on the 1-cell 4G base station and the 2-cell 4G base station, respectively guide the network search and synchronization of the SIM cards entering their respective cell coverage ranges and process the random access requests, and generate the 1-cell registration information through registration authentication And the 2-cell registration information And take the union to construct the registration information

[0010] Video acquisition unit Based on the occlusion situation of the light beam array, sense the entry and exit of personnel or vehicles, start the camera recording and stop when the personnel or vehicle leaves, and generate a surveillance video

[0011] Uplink communication unit Use the randomly generated uplink session key To encrypt the registration information And the surveillance video To generate uplink encrypted data Use the public key PK0 of the central analysis module S0 to encrypt the uplink session key To generate an encrypted uplink session key Combine to generate the uplink ciphertext And send it to the central analysis module S0, and receive the downlink ciphertext And use its own private key SK m To decrypt the encrypted downlink session key To obtain the downlink session key Use the downlink session key To decrypt the encrypted downlink data To obtain the statistical number of people

[0012] Display unit Display the statistical number of people in real time and the system time t.

[0013] Furthermore, the 4G base station in Community 1 and the 4G base station in Community 2 guide the network search and synchronization of the SIM cards entering their respective community coverage areas and process random access requests, and generate the registration information of Community 1 and the registration information of Community 2 Taking the 4G base station in Community 1 as an example, it includes the following specific steps:

[0014] The 4G base station in Community 1 notifies the SIM cards within the range of Community 1 by sending frequency band signals and cell identification announcements, and guides the SIM cards to maintain frequency and time synchronization by sending synchronization signals and timing information;

[0015] The 4G base station in Community 1 receives the random access request sent by the SIM card and detects whether the signal strength meets the standard. If it meets the standard, the 4G base station in Community 1 feeds back access permission information;

[0016] The 4G base station in Community 1 obtains the registration request information sent by the SIM card and forwards it to the mobile management entity to authenticate the registration request information. Among them, the registration request information includes the international mobile subscriber identification number, and the mobile management entity is set by the operator;

[0017] The 4G base station in Community 1 receives the temporary identifier fed back by the mobile management entity, notifies the SIM card that the registration is successful, and packages the registration time of Community 1 with the international mobile subscriber identification number to generate the registration information of Community 1

[0018] Further, the central analysis module S0 includes a downlink communication unit Retrieval and verification unit Auxiliary verification unit Early warning and statistics unit Central storage unit and access display unit

[0019] Downlink communication unit Receive the uplink ciphertext and decrypt the encrypted uplink session key using its own private key SK0 to obtain the uplink session key and decrypt the uplink encrypted data to obtain the registration information and the monitoring video Based on the randomly generated downlink session key for the statistical number of people Encrypt to generate encrypted downstream data According to the public keys PK1, …, PK of M distributed acquisition modules m , …, PK M Encrypt the downstream session key respectively Generate M encrypted downstream session keys Successively pack with the encrypted downstream data To generate the downstream ciphertext And send it to the corresponding distributed acquisition module, m = 1, …, M, where M is the total number of mine portals;

[0020] Retrieval and verification unit Analyze the surveillance video frame by frame through the target detection model To identify the target type as a person or a vehicle, and according to the registration information The registration time of cell 1 And the registration time of cell 2 To clarify the target behavior as entering or leaving the mine based on the time sequence, convert the international mobile subscriber identification number in the registration information Into a user hash value and perform retrieval and verification, retrieve whether there is a face image f or license plate information y that matches the user hash value. If the retrieval and verification are passed, obtain the face image f or license plate information y. If the retrieval and verification fail, send a verification failure prompt to the warning and statistics unit

[0021] Auxiliary verification unit Based on the target type being a person or a vehicle, select a face frame screening algorithm or a vehicle frame screening algorithm to extract the face frame image f In the surveillance video face Or license plate frame image f num , and select the auxiliary verification method as comparing the face frame image f face With the face image f or by comparing the license plate information y with the actual license plate information y generated by analyzing the license plate frame image f num By the optical character model num , and only send a verification failure prompt to the warning and statistics unit when the auxiliary verification fails

[0022] Warning and statistics unit Based on the target behavior of entering or leaving the mine, decide to increase or decrease the statistical number of people By 1, pack the registration information Surveillance video Target behavior and system time t to generate a behavior record r m (t), and encrypt it based on its own public key PK0 to generate the encrypted behavior record PK0(r m(t)) and add it to the behavior record table. If a verification failure prompt is received, send a warning prompt and mark the encrypted behavior record PK0(r in red m (t));

[0023] Central storage unit Store the user information table and the behavior record table. The user information table correspondingly records the user hash value, the user fingerprint hash value, the face image f, and the license plate information y;

[0024] Access display unit Obtain the input fingerprint and convert it into an input fingerprint hash value. Based on the matching result between the input fingerprint hash value and all user fingerprint hash values, decide whether to decrypt the encrypted behavior record in the behavior record table with its own private key SK0 and display it.

[0025] Furthermore, select a face frame screening algorithm to extract the face frame image f face And compare the face frame image f through a face recognition model face The auxiliary verification with the face image f includes the following specific steps:

[0026] Use a face detector based on HOG features to process the surveillance video Divide the frame image of each frame in the video into cells at equal intervals, calculate the direction and amplitude of the pixels in the cells, statistically generate a gradient histogram to describe the HOG features, and perform binary classification on the HOG features of the cells through a support vector machine to screen and generate a face region;

[0027] Convert the frame image into a frame grayscale image, transform the frame grayscale image through the Laplace operator and calculate the grayscale variance as the clarity score of the frame image;

[0028] Use a 68-key point detector to obtain 68 face key points of the frame image of each frame, and define the key point coverage rate as the ratio of the number of face feature points in the face region to 68;

[0029] Obtain the circumscribed rectangle of the face region, and define the face occupancy ratio as the area of the circumscribed rectangle divided by the area of the frame image;

[0030] Screen out the surveillance video Candidate frames in the video that meet the conditions that the clarity score is greater than or equal to the clarity threshold, the key point coverage rate is greater than or equal to the coverage rate threshold, and the face occupancy ratio meets the occupancy ratio range;

[0031] Normalize the clarity score of the candidate frame and add it to the key point coverage rate as the face comprehensive score, and select the frame image corresponding to the candidate frame with the highest face comprehensive score as the face frame image f face ;

[0032] Use the face recognition model to extract the face frame image f respectively faceAnd the feature vectors in the face image f are used to calculate the cosine similarity, which is compared with the similarity threshold for auxiliary verification;

[0033] If it is greater than or equal to the similarity threshold, the auxiliary verification passes. If it is less than the similarity threshold, the auxiliary verification fails, and a verification failure prompt is sent to the early warning statistics unit

[0034] Furthermore, a vehicle frame screening algorithm is selected to extract the surveillance video The license plate frame image f in num , and the auxiliary verification is carried out by comparing the license plate information y with the actual license plate information y generated by the optical character model num The auxiliary verification includes the following specific steps:

[0035] Input the frame image of each frame in the surveillance video into the SSD license plate detection model, extract the frame image features through convolution and pooling, and use the multi-scale feature map and anchor box mechanism to output the grid confidence;

[0036] The grid coverage area with the grid confidence greater than or equal to the confidence threshold is used as the candidate license plate area, and the Canny edge detection algorithm is used to perform Gaussian blur, non-maximum suppression, and double-threshold detection on the frame image in turn to correct and generate the license plate area;

[0037] Convert the frame image into a frame grayscale image, use the Sobel operator to calculate the total gradient amplitude of all pixels in the license plate area of the frame grayscale image, transform the frame grayscale image through the Laplace operator and calculate the gray variance as the clarity score of the frame image, normalize the total gradient amplitude and clarity score of each frame image and superimpose them to obtain the license plate comprehensive score, and select the frame image with the highest license plate comprehensive score as the license plate frame image f num ;

[0038] Use the optical character model to perform character recognition on the license plate frame image f num , and the optical character model uses a convolutional recurrent neural network to extract the features of the license plate frame image f num to generate a feature sequence. The recurrent layer processes the long-term dependencies in the feature sequence and models them as a hidden state sequence, and further inputs them into the fully connected layer for classification prediction, calculates the probability distribution of each character category to obtain the actual license plate information y num ;

[0039] Compare the actual license plate information y num with the license plate information y. If they are the same, the auxiliary verification passes. If they are different, the auxiliary verification fails, and a verification failure prompt is sent to the early warning statistics unit

[0040] Furthermore, as an improvement, a separable convolutional recurrent neural network is proposed. The multi-channel convolutional layer in the convolutional recurrent neural network is replaced by a depthwise separable convolutional layer. The depthwise separable convolutional layer extracts the features of the license plate frame image f num using a small number of channels of native convolution, and performs a linear operation on the basic feature sequence generated by the native convolution to generate a feature sequence with the number of channels equal to the output channels of the multi-channel convolutional layer. The linear operation includes element-wise addition and weighted summation to solve the problems of large parameter quantity of the multi-channel convolutional layer, low character recognition efficiency, and high device performance requirements.

[0041] Furthermore, the downlink communication unit and the M uplink communication units respectively need to randomly generate corresponding downlink session keys and uplink session keys The specific steps are as follows:

[0042] Randomly select a large prime number q greater than the public keys of all distributed acquisition modules and the public key of the central analysis module S0. The public keys of all modules are uniformly defined as PK. The large prime number q is a proper noun in the field of encryption, specifically referring to a relatively large prime number;

[0043] Perform prime factorization on q - 1 to find N prime factors q1,..., q N of q - 1. The prime factors q1,..., q N are all less than q - 1, and there are N corresponding integers a1,..., a N such that the product of the powers with N prime factors q1,..., q N as the base and N corresponding integers a1,..., a N as the exponents is equal to q - 1;

[0044] Generate a random integer g greater than 1 and less than q - 1;

[0045] For the N prime factors q1,..., q N of q - 1, perform generator verification in sequence to determine whether the remainder of the power with the random integer g as the base and the value of q - 1 divided by the nth prime factor q n as the exponent modulo the large prime number q is not 1, where n = 1,..., N;

[0046] If there is 1 in the generator verification result, regenerate a random integer g greater than 1 and less than q - 1. If the generator verification results are all not 1, the random integer g is the generator of the finite field G(q), and the finite field G(q) is constructed by the integers from 1 to q - 1 under the addition and multiplication operations modulo the large prime number q;

[0047] Generate a random number greater than 1 and less than PK - 1, and use the remainder of the power function of the generator g with respect to the random number modulo the large prime number q as the session key K. The session key K is uniformly represented as the downlink session key and the uplink session keys of M uplink communication units

[0048] A base station data acquisition device for multi - system integrated access at a coal mine pithead, including a 4G high - power integrated base station and a central device;

[0049] The 4G high - power integrated base station is equipped with a distributed acquisition module, establishes a network communication link with the central device to achieve encrypted data interaction, responds to the access of the SIM card and generates registration information, senses the incoming and outgoing personnel or vehicles and acquires surveillance videos, decrypts to obtain the statistical number of people and displays it in real - time;

[0050] The central device includes a host computer and a fingerprint verifier. The host computer has a built - in central analysis module S0, realizes encrypted data interaction with the 4G high - power integrated base station through a network communication link, decrypts to obtain the registration information and surveillance videos and conducts retrieval verification and auxiliary verification, generates behavior records and encrypts and stores them, and decides whether to generate a warning prompt and mark the encrypted behavior records in red based on the results of the retrieval verification and auxiliary verification, updates the statistical number of people, obtains the input fingerprint for access verification to provide a viewable behavior record form, and the fingerprint verifier is used to obtain the input fingerprint.

[0051] Furthermore, the 4G high - power integrated base station includes a 1 - cell 4G base station, a 2 - cell 4G base station, an entrance - exit display screen, an active infrared reflection sensor, a camera, a tower mast, a counterweight base, a power supply, and an intelligent gateway;

[0052] The 1 - cell 4G base station and the 2 - cell 4G base station are installed at the top of the tower mast, with different antenna orientations to cover the 1 - cell and the 2 - cell respectively, and sense the access of SIM cards within the range of the 1 - cell or the 2 - cell to generate 1 - cell registration information or 2 - cell registration information;

[0053] The entrance - exit display screen is installed 4 meters above the tower mast, and is used to display the statistical number of people and the system time;

[0054] The active infrared reflection sensor is installed 2 meters above the tower mast, emits a light beam array to the ground to sense the entry and exit of personnel or vehicles, and controls the start and stop of the camera to record surveillance videos through a hardware interface;

[0055] The camera is installed 2 meters above the tower mast and is controlled by the active infrared reflection sensor to record surveillance videos;

[0056] The tower mast is fixed to the counterweight base and is used to install the 1 - cell 4G base station, the 2 - cell 4G base station, the entrance - exit display screen, the active infrared reflection sensor, and the camera;

[0057] The counterweight base is located directly below the tower mast for stability. The internal storage bin of the counterweight base is used to install the power supply and the intelligent gateway;

[0058] The power supply supplies power to the 4G base station in Area 1, the 4G base station in Area 2, the entrance and exit display screens, the active infrared reflection sensors, the cameras, and the intelligent gateway;

[0059] The intelligent gateway establishes a network communication link with the central device to achieve encrypted data interaction.

[0060] Compared with the prior art, in the present invention, the distributed acquisition module deployed at the mine pit entrance senses the access of the SIM card and the entry and exit of the target to generate registration information and surveillance videos, which are encrypted and transmitted to the central analysis module, and the decrypted obtained statistical number of people is displayed in real time. The central analysis module decrypts to obtain the registration information and surveillance videos, uses the target detection model to analyze the target types in the surveillance videos as personnel or vehicles, determines the target behavior as entering the mine or exiting the mine based on the time sequence in the registration information, and preliminarily judges whether the SIM card belongs to an authorized person by retrieval and verification, identifies the situation of unauthorized personnel entering, further selects different frame screening algorithms based on the target type to extract frame images, and uses the face recognition model or the optical character model to assist in verifying the face frame images or license plate frame images to determine whether the SIM card matches the entering or exiting personnel or vehicles, identifies the situation of others stealing, generates behavior records and generates a warning prompt when the verification fails, updates the statistical number of people based on the target behavior and feeds it back to the distributed acquisition system, realizes the accurate detection of the passing targets at the mine pit entrance and the real-time statistics of the personnel in the mine, and improves the management level of the mine access. Description of the Drawings

[0061] Figure 1 It is a schematic diagram of the base station data acquisition system for the multi-system fusion access at the coal mine pit entrance in the present invention;

[0062] Figure 2 It is a flow chart of generating cell registration information by the 4G base station in the cell in the present invention;

[0063] Figure 3 It is a flow chart of assisting verification through the face frame screening algorithm and the face recognition model in the present invention;

[0064] Figure 4 It is a diagram of the base station data acquisition device for the multi-system fusion access at the coal mine pit entrance in the present invention;

[0065] Figure 5 It is a structure diagram of the 4G high-power integrated base station in the present invention.

[0066] Reference numerals: 1. 4G high-power integrated base station; 11. 4G base station in Area 1; 12. 4G base station in Area 2; 13. Entrance and exit display screen; 14. Active infrared reflection sensor; 15. Camera; 16. Tower mast; 17. Counterweight base; 18. Power supply; 19. Intelligent gateway; 2. Central device; 21. Host computer; 22. Fingerprint verifier. Detailed implementation mode

[0067] The present invention will be further described in detail below with reference to the accompanying drawings and embodiments.

[0068] Embodiment 1

[0069] As Figure 1 shown, the present invention discloses a base station data acquisition system for multi-system fusion access at a coal mine pithead, including M distributed acquisition modules S1, …, S M and a central analysis module S0;

[0070] The M distributed acquisition modules S1, …, S m , …, S M respectively monitor M different coal mine pitheads. The mth distributed acquisition module S m responds to the access of the SIM card and generates registration information senses the incoming and outgoing personnel or vehicles based on infrared technology and acquires surveillance videos combines symmetric encryption and asymmetric encryption for the registration information and the surveillance videos to perform joint encryption to generate an uplink ciphertext and sends it to the central analysis module S0, and performs joint decryption on the downlink ciphertext to obtain the counted number of people and display it, where m = 1, …, M, and M is the total number of coal mine pitheads;

[0071] The central analysis module S0 combines symmetric encryption and asymmetric encryption to perform joint decryption on the uplink ciphertext m sent by the mth distributed acquisition module S to obtain the registration information and the surveillance videos analyzes the surveillance videos based on the target detection model to clarify the target type, clarifies the target behavior based on the registration information and performs retrieval verification. After the retrieval verification passes, the face image f or license plate information y is obtained. Different frame screening algorithms are selected based on the target type to extract the face frame image f in the surveillance videos face or the license plate frame image f num , and uses the face recognition model to process the face frame image f facePerform auxiliary verification with the face image f or analyze the license plate frame image f using an optical character model num Obtain the actual license plate information y num And perform auxiliary verification with the license plate information y, generate a behavior record r in combination with the system time t m (t) And decide whether to generate a warning prompt based on the results of retrieval verification or auxiliary verification, and update the statistical number of people Use joint encryption to generate the downlink ciphertexts of M distributed acquisition modules S1,…,S m ,…,S M respectively And send them correspondingly, encrypt and store the behavior record r based on its own public key PK0 m (t) In the behavior record table and provide access verification to ensure that authorized users can view the behavior record r m (t), where the target types include personnel and vehicles, and the target behaviors include entering the mine and leaving the mine

[0072] Furthermore, the M distributed acquisition modules S1,…,S m ,…,S M Have the same unit design. The mth distributed acquisition module S m Includes a SIM response unit A video acquisition unit An uplink communication unit And a display unit

[0073] The SIM response unit Is implemented based on the 1-cell 4G base station and the 2-cell 4G base station arranged at the mth mine pit entrance. The 1-cell 4G base station and the 2-cell 4G base station have different cell coverage ranges, respectively guide the network search and synchronization of the SIM cards within their respective cell coverage ranges, process the random access requests of the SIM cards, and generate corresponding 1-cell registration information And 2-cell registration information When the SIM card in the terminal device carried by a person or vehicle enters or exits the mine pit entrance, it will pass through the cell coverage ranges of the 1-cell 4G base station and the 2-cell 4G base station in forward and reverse order respectively, and generate 1-cell registration information And 2-cell registration information Further take the union to construct the registration information The registration information Includes the international mobile subscriber identification number, the 1-cell registration time And the 2-cell registration time Wherein, the cell is the basic unit of the 4G network coverage range;

[0074] The video acquisition unit Implemented based on an active infrared reflection sensor and a camera. When a person or vehicle enters or exits, it blocks the emission beam array of the active infrared reflection sensor. The light intensity and angle of the receiving beam array of the active infrared reflection sensor change compared to the default light intensity and default angle without occlusion. The camera is activated for recording until the light intensity and angle of the receiving beam return to the default light intensity and default angle, and a surveillance video is generated.

[0075] Uplink communication unit Randomly generate an uplink session key For the registration information And the surveillance video Perform symmetric encryption to generate uplink encrypted data Use the public key PK0 of the central analysis module S0 to perform asymmetric encryption on the uplink session key Generate an encrypted uplink session key Pack with the uplink encrypted data Generate an uplink ciphertext And send it to the central analysis module S0 to receive the downlink ciphertext And use its own private key SK m Perform asymmetric decryption on the encrypted downlink session key To obtain the downlink session key Use the downlink session key Perform symmetric decryption on the encrypted downlink data To obtain the counted number of people

[0076] Display unit Realtime display the counted number of people on the entrance and exit display screen And the system time t.

[0077] Such as Figure 2 Shown, furthermore, the 4G base station of Cell 1 or the 4G base station of Cell 2 guides the network search and synchronization of the SIM card, processes the random access request, and generates the registration information of Cell 1 Or the registration information of Cell 2 Including the following specific steps:

[0078] The 4G base station of Cell 1 or the 4G base station of Cell 2 continuously sends frequency band signals, cell identifiers, synchronization signals, and timing information within the 4G frequency band to the scope where Cell 1 or Cell 2 is located;

[0079] Since the terminal device with a SIM card installed has a built-in 4G network search function, the 4G base station in Cell 1 or the 4G base station in Cell 2 notifies the SIM card within the range of Cell 1 or Cell 2 that accesses the 4G base station through the frequency band signal and cell identifier, and guides the SIM card to maintain frequency synchronization and time synchronization through the synchronization signal and timing information to establish a random access channel;

[0080] The 4G base station in Cell 1 or the 4G base station in Cell 2 receives the random access request of the SIM card within the range of Cell 1 or Cell 2 on the random access channel, and detects whether the signal strength of the random access request meets the standard. If it meets the standard, the 4G base station in Cell 1 or the 4G base station in Cell 2 sends access permission information to the SIM card through the random access response channel. Among them, the random access request is obtained by the SIM card selecting from a predefined set of random access preambles;

[0081] The 4G base station in Cell 1 or the 4G base station in Cell 2 obtains the registration request information sent by the SIM card, and the 4G base station in Cell 1 or the 4G base station in Cell 2 forwards it to the mobile management entity in the 4G network. Among them, the registration request information includes the international mobile subscriber identification number, and the mobile management entity is set by the operator and cooperates with the home subscriber server also set by the operator to implement the authentication of the registration request information;

[0082] The 4G base station in Cell 1 or the 4G base station in Cell 2 receives the temporary identifier fed back by the mobile management entity, notifies the SIM card that the registration is successful, and packages the registration time in Cell 1 or the registration time in Cell 2 with the international mobile subscriber identification number to generate the registration information in Cell 1 or the registration information in Cell 2

[0083] Furthermore, the central analysis module S0 is implemented based on the host computer and the fingerprint verifier, and includes a downlink communication unit retrieval and verification unit auxiliary verification unit early warning and statistics unit central storage unit and access display unit

[0084] Downlink communication unit Receives the uplink ciphertext Uses its own private key SK0 to perform asymmetric decryption on the encrypted uplink session key to obtain the uplink session key and further performs symmetric decryption on the uplink encrypted data to obtain the registration information and the monitoring video Randomly generates a downlink session key and for the statistical number of people Perform symmetric encryption to generate encrypted downstream data According to the public keys PK1, …, PK of M distributed acquisition modules S1, …, S m , …, S M Perform asymmetric encryption on the downstream session key m , …, PK M respectively to generate M encrypted downstream session keys corresponding to the distributed acquisition modules S1, …, S m M , …, S M Corresponding encrypted downstream session keys Successively pack with the encrypted downstream data to generate downstream ciphertext and send it to the corresponding distributed acquisition module, where m = 1, …, M and M is the total number of mine portals;

[0085] Retrieval and verification unit Read the surveillance video frame by frame through the target detection model to identify that the target type in the surveillance video is a person or a vehicle, extract the registration information of the registration time in Community 1 and the registration time in Community 2 and compare the chronological order. If the registration time in Community 1 is earlier than the registration time in Community 2 clarify that the target behavior is entering the mine. If the registration time in Community 1 is later than the registration time in Community 2 clarify that the target behavior is leaving the mine, extract the international mobile subscriber identification number in the registration information and convert it into a user hash value through a hash function. Based on the target type, retrieve the face image f or license plate information y that matches the user hash value in the central storage unit for retrieval and verification. If there is a face image f or license plate information y that matches the user hash value, it indicates that the registration information of the SIM card belongs to an authorized user and the retrieval and verification are passed. If there is no face image f or license plate information y that matches the user hash value, it indicates that the registration information of the SIM card belongs to an unauthorized user and the retrieval and verification fails. Send a verification failure prompt to the warning and statistics unit Among them, the target detection model includes RCNN, Fast-RCNN, Faster-RCNN, YOLO, SSD, RetinaNet, Detection Transformer, and SwinTransformer;

[0086] Auxiliary verification unit Identify the target type. If the target type is a person, select the face frame screening algorithm to extract the face frame image f with the best face clarity and face angle from the surveillance video and calculate the similarity between the face frame image f face and the face image f using the face recognition model for auxiliary verification. If the target type is a vehicle, select the vehicle frame screening algorithm to extract the license plate frame image f with the highest license plate clarity from the surveillance video face and analyze the license plate frame image f using the optical character model to obtain the actual license plate information y num and compare it with the license plate information y for auxiliary verification. If and only if the similarity between the face frame image f num and the face image f is greater than or equal to the similarity threshold or the actual license plate information y num matches the license plate information y, the auxiliary verification passes. Otherwise, it indicates that the person or vehicle cannot match the authorized user to whom the SIM card's registration information face belongs, and the auxiliary verification fails. A verification failure prompt is sent to the early warning statistics unit num The early warning statistics unit updates the statistical number of people based on the target behavior

[0087] If the target behavior is entering the mine, the statistical number of people is incremented by 1. If the target behavior is leaving the mine, the statistical number of people is decremented by 1. The registration information is packaged along with the surveillance video the target behavior and the system time t to generate a behavior record r which is asymmetrically encrypted based on its own public key PK0 to generate an encrypted behavior record PK0(r m (t)) and added to the behavior record table. When receiving a verification failure prompt from the retrieval verification unit m or the auxiliary verification unit(t)) and added to the behavior record table. When receiving a verification failure prompt from the retrieval verification unit or the auxiliary verification unit a warning prompt is sent to the staff and the encrypted behavior record PK0(r m (t)) is marked in red;

[0088] The central storage unit stores the user hash value, the user fingerprint hash value, the face image f, and the license plate information y in the form of a user information table. At the same time, the behavior record table is stored. The data in the central storage unit can be called at any time. Since the user hash value and the user fingerprint hash value cannot be reversely restored to the International Mobile Subscriber Identity and the user fingerprint, and all the stored in the behavior record table are encrypted behavior records, even if an outsider obtains them, no valid information can be known. Among them, the user information table is pre-entered by the authorized user in the system;

[0089] Access display unit Through access verification, ensure that authorized users decrypt and view the behavior record table, obtain the input fingerprints of authorized users, convert them into input fingerprint hash values through a hash function, and compare them with all user fingerprint hash values stored in the central storage unit for traversal matching. If there is a matching user fingerprint hash value, the access verification passes, and each encrypted behavior record in the behavior record table is asymmetrically decrypted and displayed using its own private key SK0. If there is no matching user fingerprint hash value, the access verification fails, and the decryption of the behavior record table is rejected.

[0090] Such as Figure 3 shown, furthermore, extract the face frame image f from the surveillance video through the face frame screening algorithm face , and use the face recognition model for auxiliary verification, including the following specific steps:

[0091] For each frame image in the surveillance video , use a face detector based on HOG features, divide the frame image into cells at equal intervals, calculate the direction and amplitude of the pixels in each cell, and statistically generate a gradient histogram to describe the HOG features of the frame image. Use a support vector machine to perform binary classification on the HOG features of the cells in the high-dimensional space, and screen out the cells belonging to the face and frame the face area;

[0092] Convert each frame image into a frame grayscale image, enhance the high-frequency components, amplify the gray-scale changes of the edges and details in the frame grayscale image through the Laplace operator transformation, generate a transformed grayscale image and calculate the gray-scale variance. The gray-scale variance reflects the degree of gray-scale dispersion. The higher the clarity of the image, that is, the more high-frequency information the image contains, the more dispersed the gray-scale distribution, and the larger the gray-scale variance. Therefore, the gray-scale variance can be used as the clarity score of the frame image;

[0093] Use 68 key point detectors in Dlib to obtain 68 face key points in each frame image of the surveillance video. Define the key point coverage rate as the ratio of the number of face feature points in the face area to 68. The higher the key point coverage rate, the more face key points are included in the frame image, that is, the more complete the face is photographed. Among them, the 68 key point detectors are existing pre-trained models and are widely used standards in the face detection industry;

[0094] Obtain the circumscribed rectangle of the face area, and define the face occupancy ratio as the area of the circumscribed rectangle divided by the area of the frame image. If the face occupancy ratio is too small, it indicates that the face is too small relative to the frame image. If the face occupancy ratio is too close to 1, it indicates that the face may exceed the boundary of the frame image;

[0095] Screen out the surveillance video The candidate frames in

[0096] satisfy that the clarity score is greater than or equal to the clarity threshold, the key point coverage rate is greater than or equal to the coverage rate threshold, and the face occupancy ratio satisfies the occupancy ratio range; the clarity scores of all candidate frames are normalized to the range of 0 to 1 and added to the key point coverage rate to obtain the comprehensive face score, and the frame image with the highest comprehensive face score among the candidate frames is selected as the face frame image f face ;

[0097] The first feature vector and the second feature vector in the face frame image f face and the face image f are respectively extracted by using a face recognition model, the cosine similarity between the first feature vector and the second feature vector is calculated and compared with the similarity threshold for auxiliary verification. Among them, the face recognition model includes VGG-Face, Facenet, SphereFace, MobileFace, and ArcFace, and the first feature vector and the second feature vector are respectively used to describe the face feature data in the face frame image f face and the face image f;

[0098] If the cosine similarity is greater than or equal to the similarity threshold, it indicates that the face frame image f face corresponds to the same person as the face image f, that is, the registration information of the person and the SIM card belongs to the authorized user and the auxiliary verification is passed;

[0099] If the cosine similarity is less than the similarity threshold, it indicates that the face frame image f face corresponds to a different person from the face image f, that is, the registration information of the person and the SIM card does not belong to the authorized user, the auxiliary verification fails, and a verification failure prompt is sent to the warning statistics unit

[0100] Furthermore, the license plate frame image f is extracted from the surveillance video through a vehicle frame screening algorithm num , and the actual license plate information y num is obtained by using an optical character model and the auxiliary verification includes the following specific steps:

[0101] The frame image of each frame in the surveillance video is input into the SSD license plate detection model. The SSD license plate detection model extracts the frame image features through convolution and pooling, and uses the multi-scale feature map and the anchor box mechanism to output the grid confidence. The higher the grid confidence, the higher the possibility that the grid is a license plate;

[0102] Regard the area covered by grids with grid confidence greater than or equal to the confidence threshold as the candidate license plate area. Introduce the Canny edge detection algorithm, and correct the edges of the candidate license plate area by performing Gaussian blur processing, non-maximum suppression, and double-threshold detection on the frame image to generate the license plate area;

[0103] Convert the frame image of each frame into a frame grayscale image. Use the Sobel operator to calculate the gradient magnitude of each pixel in the license plate area of the frame grayscale image and sum it as the total gradient magnitude of the license plate area. The larger the total gradient magnitude, the clearer the edges of the license plate area. Transform the frame grayscale image through the Laplace operator and calculate the gray variance as the clarity score of the frame image. Normalize the total gradient magnitude and clarity score of the frame image of each frame to the range of 0 to 1 respectively and superimpose them to obtain the license plate comprehensive score. Select the frame image with the highest license plate comprehensive score as the license plate frame image f num ;

[0104] Use the optical character model to perform character recognition on the license plate frame image f num The optical character model uses a convolutional recurrent neural network. The convolutional recurrent neural network is used to perform end-to-end recognition of text sequences with variable lengths, eliminating the need for individual segmentation of characters, and transforming the text recognition task into a sequence learning problem based on temporal dependence. The convolutional recurrent neural network extracts features from the license plate frame image f num to generate a feature sequence. The recurrent layer processes the long-term dependencies in the feature sequence, sequentially models the feature sequence as a hidden state sequence to capture the context information between characters, and inputs the hidden state sequence into the fully connected layer for classification prediction, calculates the probability distribution of each character category to obtain the actual license plate information y num ;

[0105] Compare the actual license plate information y num with the license plate information y. If the actual license plate information y num is the same as the license plate information y, it indicates that the vehicle matches the authorized user to whom the registration information of the SIM card belongs, and the auxiliary verification passes. If the actual license plate information y num is different from the license plate information y, it indicates that the vehicle does not match the authorized user to whom the registration information of the SIM card belongs, and the auxiliary verification fails. Send a verification failure prompt to the early warning statistics unit

[0106] Furthermore, the convolutional recurrent neural network adopted by the optical character model relies on the network architecture of VGG-16, and extracts complex features of the license plate frame image f num through multi-channel convolutional layers. The input license plate frame image f numIt is often an RGB three-channel color image. Therefore, the number of input channels L of the multi-channel convolutional layer in is 3, the convolutional kernel size is 3×3, and the number of output channels L of the multi-channel convolutional layer out is a power of 2 and generally greater than or equal to 128. At this time, the computational complexity of the conventional multi-channel convolutional layer is 3×L out ×3×3, which leads to a large number of network parameters and the expansion of the volume of the optical character model, reduces the character recognition efficiency, and increases the performance operation requirements of the device. As an improvement, a separable convolutional recurrent neural network is proposed, which uses a depthwise separable convolutional layer to replace the conventional multi-channel convolutional layer. First, the original convolution is used to extract the features of the license plate frame image f num . The original convolution refers to the convolution operation with a small number of channels. The size of the original convolutional kernel is still 3×3, and the number of output channels L of the original convolution prim is a power of 2 and much smaller than the number of output channels L of the multi-channel convolutional layer out , generally set to 32 or 64. For large-scale image recognition or high-precision semantic segmentation tasks, it can be set to 128 or 256. At this time, the computational complexity of the original convolution is 3×L prim ×3×3. A series of linear operations are performed on the basic feature sequence generated by the original convolution to generate a feature sequence with more channels to reach the number of output channels L of the multi-channel convolutional layer out . The linear operations include element-wise addition and weighted summation. The computational complexity of the linear operation is L prim -1. Therefore, the total computational complexity of the depthwise separable convolutional layer is the computational complexity of the original convolution plus the computational complexity of the linear operation. In this embodiment, the number of output channels L of the multi-channel convolutional layer out is set to 256, and the number of output channels L of the original convolution prim is 64. The computational complexity of the separable convolutional recurrent neural network is approximately reduced by 75% compared with that of the convolutional recurrent neural network, improving the character recognition efficiency of the optical character model.

[0107] Furthermore, the downlink communication unit and M uplink communication units respectively need to randomly generate the corresponding downlink session key and uplink session key , which includes the following specific steps:

[0108] Randomly select a large prime number q that satisfies being greater than the public keys PK0, PK1,..., PK of the central analysis module S0 and M distributed acquisition modules S1,..., S m ,..., S M ,..., PK m ,..., PK M . Define the public key PK to uniformly represent the public keys PK0, PK1,..., PK m,…,PK M , where the large prime number q is a proprietary term in the field of encryption, specifically referring to a relatively large prime number;

[0109] Factorize q - 1 into prime factors, that is, find N prime numbers q1,…,q less than q - 1 N and N corresponding integers a1,…,a N , such that the product of the powers with N prime numbers q1,…,q N as the bases and N corresponding integers a1,…,a N as the exponents is equal to q - 1. The specific formula is At this time, the N prime numbers q1,…,q less than q - 1 N are called the N prime factors of q - 1;

[0110] Generate a random integer g that satisfies being greater than 1 and less than q - 1;

[0111] For the N prime factors q1,…,q of q - 1 N Perform the generator verification in sequence. Determine whether the remainder of the power with the random integer g as the base and the value of q - 1 divided by the nth prime factor q n as the exponent modulo the large prime number q is not 1. The specific formula is mod represents the modulo operation;

[0112] If there is a case where the generator verification result of the random integer g for the N prime factors q1,…,q N is equal to 1, then regenerate a random integer g that satisfies being greater than 1 and less than q - 1;

[0113] If the generator verification results of the random integer g for the N prime factors q1,…,q N are all not 1, then the random integer g is the generator of the finite field G(q), the large prime number q is the order of the finite field G(q), and the finite field G(q) is constructed by the addition and multiplication operations of integers from 1 to q - 1 modulo the large prime number q;

[0114] Generate a random number rand(1, PK - 1) that is greater than 1 and less than PK - 1. Take the remainder of the power function of the generator g with respect to the random number rand(1, PK - 1) modulo the large prime number q as the session key K. The session key K uniformly represents the downlink session key of the downlink communication unit and the uplink session keys of M uplink communication units

[0115] Example 2

[0116] Such as Figure 4As shown in the figure, the present invention also discloses a base station data acquisition device for multi-system fusion access at a coal mine pithead, including a 4G high-power integrated base station 1 and a central device 2;

[0117] The 4G high-power integrated base station 1 is deployed at the mine pithead, equipped with a distributed acquisition module, establishing a network communication link with the central device 2 to achieve encrypted data interaction, responding to the access of the SIM card and generating registration information, sensing the incoming and outgoing personnel or vehicles based on infrared technology and collecting surveillance videos, jointly decrypting to obtain the statistical number of people and displaying it in real time;

[0118] The central device 2 includes a host computer 21 and a fingerprint verifier 22. The host computer 21 has a built-in central analysis module S0, achieving encrypted data interaction with the 4G high-power integrated base station 1 through a network communication link, jointly decrypting to obtain registration information and surveillance videos and conducting retrieval verification and auxiliary verification, generating behavior records and encrypting and storing them, deciding whether to generate a warning prompt and marking the encrypted behavior records in red based on the verification results, updating the statistical number of people, and obtaining the input fingerprint for access verification to ensure that authorized users can view the behavior records; the fingerprint verifier 22 is used to obtain the input fingerprint to assist the host computer 21 in access verification.

[0119] As Figure 5 shown in the figure, further, the 4G high-power integrated base station 1 includes a 1-cell 4G base station 11, a 2-cell 4G base station 12, an entrance and exit display screen 13, an active infrared reflection sensor 14, a camera 15, a tower mast 16, a counterweight base 17, a power supply 18, and an intelligent gateway 19;

[0120] The 1-cell 4G base station 11 and the 2-cell 4G base station 12 have the same base station antenna configuration, installed at the top of the tower mast 16, with different antenna orientations, continuously sending frequency band signals, cell identifiers, synchronization signals, and timing information to the respective ranges of the 1-cell and 2-cell to guide the network search and synchronization of SIM cards within the range, processing random access requests and generating 1-cell registration information or 2-cell registration information through registration authentication;

[0121] The entrance and exit display screen 13 is installed at 4 meters of the tower mast 16, used to display the statistical number of people and the system time;

[0122] The active infrared reflection sensor 14 is installed at 2 meters of the tower mast 16, emitting a beam array to the ground to sense the entry and exit of personnel or vehicles, connected to the trigger pin of the camera 15 through a hardware interface. When it senses the occlusion and the leaving of the occlusion of the beam array by personnel or vehicles, it generates different electrical signals to control the start and stop of the camera 15 to record surveillance videos;

[0123] The camera 15 is installed at 2 meters of the tower mast 16, controlled by the active infrared reflection sensor 14 to record surveillance videos;

[0124] The tower mast 16 is fixed to the counterweight base 17, and is used for installing and fixing the 4G base station 11 of Community 1, the 4G base station 12 of Community 2, the entrance and exit display screen 13, the active infrared reflection sensor 14, and the camera 15;

[0125] The counterweight base 17 is located directly below the tower mast 16, and is used for improving the stability of the 4G high-power integrated base station 1 and preventing the 4G high-power integrated base station 1 from tipping over. The internal storage bin of the counterweight base 17 is used for installing the power supply 18 and the intelligent gateway 19;

[0126] The power supply 18 is used for supplying power to the 4G base station 11 of Community 1, the 4G base station 12 of Community 2, the entrance and exit display screen 13, the active infrared reflection sensor 14, the camera 15, and the intelligent gateway 19;

[0127] The intelligent gateway 19 establishes a network communication link with the central device 2 to achieve encrypted data interaction.

[0128] The present invention discloses a base station data acquisition system and device for multi-system fusion access at the coal mine pithead, including a distributed acquisition module and a central analysis module; the distributed acquisition module generates registration information by sensing the access of the SIM card through the 4G high-power integrated base station deployed at the mine entrance, and generates a monitoring video by sensing the incoming and outgoing targets based on the cooperation of the active infrared reflection sensor and the camera, encrypts and transmits it to the central analysis module, and decrypts and displays the obtained statistical number of people in real time; the central analysis module decrypts and obtains the registration information and the monitoring video, analyzes the monitoring video using the target detection model to clarify the target type, clarifies the target behavior based on the registration information and conducts retrieval and verification, selects a frame screening algorithm based on the target type to extract the frame image, and uses the face recognition model or the optical character model to assist in verifying the target, generates a behavior record and generates a warning prompt when the verification fails, updates the statistical number of people based on the target behavior and feeds it back to the distributed acquisition system, realizes the accurate detection of the passing targets at the mine pithead and the real-time statistics of the people in the mine, and improves the management level of the mine access.

[0129] The above is only the preferred embodiment of the present invention, and the protection scope of the present invention is not limited to the above embodiments. All technical solutions falling within the idea of the present invention belong to the protection scope of the present invention. It should be noted that for those of ordinary skill in the art in this technical field, without departing from the principle of the present invention, several improvements and refinements should also be regarded as the protection scope of the present invention.

Claims

1. The base station data acquisition system with multi-system integrated access at the coal mine entrance includes a distributed acquisition module and a central analysis module; The distributed collection module senses the connected SIM card and the people or vehicles entering and leaving to generate registration information and monitoring video, encrypts and generates uplink ciphertext, decrypts the downlink ciphertext, obtains the statistical number of people and displays it; The central analysis module decrypts the uplink ciphertext to obtain registration information and surveillance video, analyzes the surveillance video based on the target detection model to clarify the target type, clarifies the target behavior based on the registration information and performs retrieval verification to obtain face images or license plate information, selects different frame screening algorithms based on the target type to extract frame images in the surveillance video and uses a face recognition model or an optical character model to assist in verifying the face or license plate, generates behavior records and decides whether to generate an early warning prompt based on the verification result, updates the number of people counted and encrypts them to generate downlink ciphertext, encrypts and stores the behavior records and provides access verification for query. The target types include personnel and vehicles, and the target behaviors include entering and exiting the mine.

2. The base station data acquisition system for coal mine pit entrance multi-system fusion access as claimed in claim 1, characterized in that: The central analysis module includes a retrieval verification unit, an auxiliary verification unit and an early warning statistics unit; The retrieval verification unit analyzes the surveillance video frame by frame through the target detection model to identify the target type, clarifies the target behavior based on the time sequence of the registration time of cell 1 and the registration time of cell 2 in the registration information, converts the international mobile user identity code in the registration information into a user hash value and performs retrieval verification. If the retrieval verification passes, a face image or license plate information matching the user hash value is obtained, and if the retrieval verification fails, a verification failure prompt is sent; The auxiliary verification unit selects a face frame screening algorithm or a vehicle frame screening algorithm based on the target type as a person or a vehicle to extract a face frame image or a license plate frame image in the surveillance video, and selects an auxiliary verification method as comparing the face frame image with the face image through a face recognition model or analyzing the license plate frame image by comparing the license plate information with an optical character model to generate the actual license plate information, and sends a verification failure prompt when the auxiliary verification fails; The early warning statistics unit adds 1 or subtracts 1 to the number of people counted based on the target behavior being the decision to enter or exit the mine, packages the registration information, monitoring video, target behavior and system time to generate a behavior record, generates an encrypted behavior record based on the public key encryption of the central analysis module and stores it in the behavior record table, and sends a warning prompt when receiving a verification failure prompt and marks the encrypted behavior record in red.

3. The base station data acquisition system for coal mine pit entrance multi-system fusion access as claimed in claim 2, characterized in that: Selecting the face frame screening algorithm to extract the face frame image, and comparing the face frame image with the face image through the face recognition model for auxiliary verification includes the following specific steps: A face detector based on HOG features is used to divide each frame image in the surveillance video into cells. The direction and amplitude statistics of the pixels in the cells are calculated to generate a gradient histogram to describe the HOG features. The cells are classified into two categories based on the HOG features through a support vector machine to screen and generate the face area. The frame image is converted into a frame grayscale image, the frame grayscale image is transformed by the Laplace operator and the grayscale variance is calculated as the clarity score, 68 key point detectors are used to obtain 68 facial key points of the frame image, the ratio of the number of facial feature points in the face area to 68 is calculated as the key point coverage rate, and the ratio of the circumscribed rectangle area of ​​the face area to the frame image area is calculated as the face ratio; Filter out all candidate frames that meet the definition score and key point coverage ratio that are greater than or equal to the corresponding threshold and whose face ratio meets the ratio range, normalize the definition score of the candidate frame and add it to the key point coverage ratio as the comprehensive face score, and select the frame image corresponding to the candidate frame with the highest comprehensive face score as the face frame image; The face recognition model is used to extract the feature vectors in the face frame image and the face image and calculate the cosine similarity, which is compared with the similarity threshold for auxiliary verification. If it is greater than or equal to the similarity threshold, the auxiliary verification passes. If it is less than the similarity threshold, the auxiliary verification fails and a verification failure prompt is sent.

4. The base station data acquisition system for coal mine pit entrance multi-system fusion access as claimed in claim 2, characterized in that: Selecting the vehicle frame screening algorithm to extract the license plate frame image, and performing auxiliary verification by comparing the license plate information with the actual license plate information generated by the optical character model includes the following specific steps: Input each frame image in the surveillance video into the SSD license plate detection model, extract frame image features through convolution and pooling, and output grid confidence using multi-scale feature maps and anchor frame mechanism; The grid coverage area with a grid confidence greater than or equal to the confidence threshold is taken as a candidate license plate area, and the Canny edge detection algorithm is used to refine the edge of the candidate license plate area of ​​the frame image to generate a license plate area; Convert the frame image into a frame grayscale image, use the Sobel operator to calculate the total gradient amplitude of all pixels in the license plate area of ​​the frame grayscale image, transform the frame grayscale image through the Laplace operator and calculate the grayscale variance as the clarity score of the frame image, normalize the total gradient amplitude and clarity score of the frame image and superimpose them to obtain the comprehensive score of the license plate, and select the frame image with the highest comprehensive score of the license plate as the license plate frame image; The optical character model is used to recognize characters on the license plate frame image. The optical character model uses a convolutional recurrent neural network. The multi-channel convolution layer extracts the license plate frame image to generate a feature sequence. The recurrent layer processes the dependency of the feature sequence and models it as a hidden state sequence. The fully connected layer classifies and predicts the hidden state sequence and calculates the probability distribution of each character category to obtain the actual license plate information. Compare the actual license plate information with the license plate information. If they are the same, the auxiliary verification passes. If they are different, the auxiliary verification fails and a verification failure prompt is sent.

5. The base station data acquisition system for coal mine pit entrance multi-system fusion access as claimed in claim 4, characterized in that: The optical character model can be optimized using a separable convolutional recurrent neural network, which replaces the multi-channel convolutional layer of the convolutional recurrent neural network with a depth-separable convolutional layer. The depth-separable convolutional layer uses native convolution to extract features of the license plate frame image to generate a basic feature sequence, and generates a feature sequence equal to the number of output channels of the multi-channel convolutional layer through a linear operation combination.

6. The base station data acquisition system for coal mine pit entrance multi-system fusion access as claimed in claim 1, characterized in that: The central analysis module also includes a downlink communication unit, a central storage unit and an access display unit; The downlink communication unit receives the uplink ciphertext and uses the private key of the central analysis module to decrypt the encrypted uplink session key, obtains the uplink session key and decrypts the uplink encrypted data, obtains the registration information and the surveillance video, encrypts the number of people based on the randomly generated downlink session key to generate encrypted downlink data, encrypts the downlink session key based on the public key of the distributed acquisition module to generate an encrypted downlink session key, and packages it with the encrypted downlink data to generate a downlink ciphertext; The central storage unit stores a user information table and a behavior record table, wherein the user information table records user hash values, user fingerprint hash values, face images and license plate information; The access display unit obtains the input fingerprint and converts it into an input fingerprint hash value, and decides whether to decrypt and display the encrypted behavior record in the behavior record table through the private key of the central analysis module based on the matching result between the input fingerprint hash value and the user fingerprint hash value.

7. The base station data acquisition system for coal mine pit entrance multi-system fusion access as claimed in claim 6, characterized in that: The random generation of the downlink session key comprises the following specific steps: Randomly select a large prime number q that is larger than the public key of the distributed acquisition module and the public key of the central analysis module; Factorize q-1 and find N prime factors of q-1. All N prime factors are less than q-1, and there are N corresponding integers, satisfying that the product of the powers with N prime factors as base and N corresponding integers as exponents is equal to q-1. Generate a random integer greater than 1 and less than q-1; Perform generator verification for the N prime factors of q-1 in turn, and determine whether the remainder of taking the modulus of the large prime number q with the random integer as the base and the value of q-1 divided by the value of the nth prime factor as the exponent is not 1, n = 1, ..., N; If there is 1 in the generator verification result, a new random integer is generated. If the generator verification results are not 1, the random integer is defined as the generator; Generate a random number that is greater than 1 and less than the public key of the central analysis module, and use the remainder of the power function of the generator with respect to the random number to the large prime number q as the downlink session key.

8. The base station data acquisition system for coal mine pit entrance multi-system fusion access as claimed in claim 1, characterized in that: The distributed acquisition module includes a SIM response unit, a video acquisition unit, an uplink communication unit and a display unit; The SIM response unit generates 1-cell registration information and 2-cell registration information for SIM cards entering the coverage of their respective cells based on 1-cell 4G base station and 2-cell 4G base station through guidance and registration authentication, and takes the union to construct registration information; The video acquisition unit senses the entry and exit of a person or vehicle based on the occlusion of the beam array, starts camera recording and stops when the person or vehicle leaves, generating a surveillance video; The uplink communication unit uses the randomly generated uplink session key to encrypt the registration information and the surveillance video to generate uplink encrypted data, uses the public key of the central analysis module to encrypt the uplink session key to generate an encrypted uplink session key, combines to generate uplink ciphertext, receives the downlink ciphertext and decrypts the encrypted downlink session key through the private key of the distributed acquisition module to obtain the downlink session key, and uses the downlink session key to decrypt the encrypted downlink data to obtain the statistical number of people; The display unit shows the number of people counted and system time in real time.

9. The base station data acquisition system for coal mine pit entrance multi-system fusion access as claimed in claim 8, characterized in that: The 1-cell 4G base station generates 1-cell registration information for the SIM card entering the cell coverage through guidance and registration authentication, including the following specific steps: Send frequency band signals and cell identification to notify SIM cards entering the cell range, send synchronization signals and timing information to guide SIM cards to maintain frequency and time synchronization; Receive the random access request sent by the SIM card and detect whether the signal strength meets the standard. If it meets the standard, feedback the access permission information; Obtain the registration request information sent by the SIM card and forward it to the mobile management entity to authenticate the registration request information, wherein the registration request information includes the international mobile subscriber identity code, and the mobile management entity is set by the operator; Receive the temporary identifier fed back by the mobile management entity, notify the SIM card of successful registration, and package the 1-cell registration time and the international mobile user identity code to generate 1-cell registration information.

10. Base station data collection device for multi-system integrated access at the coal mine entrance, including 4G high-power integrated base station and central equipment; The 4G high-power integrated base station includes a 1-cell 4G base station, a 2-cell 4G base station, an entrance and exit display screen, an active infrared reflection sensor, a camera and an intelligent gateway; The 4G base station for cell 1 and the 4G base station for cell 2 have different antenna orientations to cover cell 1 and cell 2 respectively, and sense SIM card access within the range of cell 1 or cell 2 to generate registration information for cell 1 or cell 2; The entrance and exit display screen is used to display the number of people counted and the system time; The active infrared reflection sensor emits an array of light beams to the ground to sense the entry and exit of people or vehicles, and controls the start and stop of the camera recording surveillance video through a hardware interface connection; The camera is controlled by an active infrared reflection sensor to record surveillance video; The intelligent gateway establishes a network communication link with the central device to perform data encryption interaction; The central device includes a host computer and a fingerprint verifier; The host computer and the 4G high-power integrated base station realize data encryption interaction through a network communication link, decrypt and obtain registration information and monitoring video, perform retrieval verification and auxiliary verification, generate and encrypt behavior records, store them, and decide whether to generate early warning prompts and mark the encrypted behavior records in red based on the results of retrieval verification and auxiliary verification, update the number of people counted, obtain input fingerprints for access verification to provide a behavior record table; The fingerprint verifier is used to obtain an input fingerprint.

Images