CN114821725A - Miner face recognition system based on neural network - Google Patents

Miner face recognition system based on neural network Download PDF

Info

Publication number
CN114821725A
CN114821725A CN202210460972.3A CN202210460972A CN114821725A CN 114821725 A CN114821725 A CN 114821725A CN 202210460972 A CN202210460972 A CN 202210460972A CN 114821725 A CN114821725 A CN 114821725A
Authority
CN
China
Prior art keywords
face
miner
module
image
feature vectors
Prior art date
Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
Pending
Application number
CN202210460972.3A
Other languages
Chinese (zh)
Inventor
刘晓阳
王地
Current Assignee (The listed assignees may be inaccurate. Google has not performed a legal analysis and makes no representation or warranty as to the accuracy of the list.)
China University of Mining and Technology Beijing CUMTB
Original Assignee
China University of Mining and Technology Beijing CUMTB
Priority date (The priority date is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the date listed.)
Filing date
Publication date
Application filed by China University of Mining and Technology Beijing CUMTB filed Critical China University of Mining and Technology Beijing CUMTB
Priority to CN202210460972.3A priority Critical patent/CN114821725A/en
Publication of CN114821725A publication Critical patent/CN114821725A/en
Pending legal-status Critical Current

Links

Images

Classifications

    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F18/00—Pattern recognition
    • G06F18/20—Analysing
    • G06F18/21—Design or setup of recognition systems or techniques; Extraction of features in feature space; Blind source separation
    • G06F18/214—Generating training patterns; Bootstrap methods, e.g. bagging or boosting
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F18/00—Pattern recognition
    • G06F18/20—Analysing
    • G06F18/24—Classification techniques
    • G06F18/241—Classification techniques relating to the classification model, e.g. parametric or non-parametric approaches
    • G06F18/2411—Classification techniques relating to the classification model, e.g. parametric or non-parametric approaches based on the proximity to a decision surface, e.g. support vector machines
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
    • G06N3/00—Computing arrangements based on biological models
    • G06N3/02—Neural networks
    • G06N3/04—Architecture, e.g. interconnection topology

Landscapes

  • Engineering & Computer Science (AREA)
  • Theoretical Computer Science (AREA)
  • Data Mining & Analysis (AREA)
  • Physics & Mathematics (AREA)
  • Life Sciences & Earth Sciences (AREA)
  • Artificial Intelligence (AREA)
  • General Physics & Mathematics (AREA)
  • General Engineering & Computer Science (AREA)
  • Evolutionary Computation (AREA)
  • Bioinformatics & Computational Biology (AREA)
  • Evolutionary Biology (AREA)
  • Computer Vision & Pattern Recognition (AREA)
  • Bioinformatics & Cheminformatics (AREA)
  • Computational Linguistics (AREA)
  • Biomedical Technology (AREA)
  • Biophysics (AREA)
  • Health & Medical Sciences (AREA)
  • General Health & Medical Sciences (AREA)
  • Molecular Biology (AREA)
  • Computing Systems (AREA)
  • Mathematical Physics (AREA)
  • Software Systems (AREA)
  • Collating Specific Patterns (AREA)

Abstract

The invention provides a miner face recognition system based on a neural network, which comprises: the system comprises a miner information acquisition module, an attendance checking identification module and a miner data storage module. The system realizes the enhancement and the denoising of the miner face image through an image preprocessing algorithm of bilateral filtering and Retinex fusion, and can accurately detect the face position in the image through an SSD target detection network. After the face position is detected, subsequent face feature extraction work is carried out through a ResNet-18 feature extraction network improved based on a mixed attention mechanism, and the network can extract face feature vectors with high identifiability. And finally, performing cosine similarity calculation on the obtained feature vector and face information stored in the database to obtain attendance data. Compared with the traditional RFID attendance system, the system avoids the problem of one person with multiple cards, effectively improves the identification accuracy of miners, and is more convenient and efficient through a face identification algorithm based on a neural network.

Description

Miner face recognition system based on neural network
Technical Field
The invention relates to the technical field of face recognition, in particular to a miner face recognition system based on a neural network.
Background
Under the development of modern science and technology, a face recognition technology is gradually applied to a mine attendance management system, and the identity of a card punching employee is determined by analyzing an image shot by a high-resolution camera, retrieving face information from the image, extracting facial features, and matching and comparing the facial features with the face information in a database. The technology can realize automatic identification of the personnel entering the well, is convenient for the managers to have comprehensive knowledge of the underground operation condition, effectively prevents the cheating of entering the well by multiple cards, provides timely preventive measures for coal mine managers, reduces the occurrence of accidents, and has important significance for improving the production safety.
However, in a coal mine environment, the face recognition algorithm faces many difficulties such as weak light, dust shielding, facial expression and the like, which leads to a reduction in recognition accuracy. Therefore, the invention provides a miner face recognition system based on a neural network aiming at the problems so as to promote the development and innovation of a coal mine safety face recognition technology.
Disclosure of Invention
The invention aims to provide a miner face recognition system based on a neural network, which is used for solving the problems of low illumination, more shielding, expression and posture and the like of the existing coal mine environment so as to improve the accuracy of an attendance system.
In order to achieve the above object, the present invention provides a miner face recognition system, comprising:
the miner information acquisition module is used for calling the camera to acquire a face image of the miner and acquiring the job number and name of the miner;
the attendance checking identification module is used for comparing and identifying the face information acquired by the camera with the face data in the database, outputting the personal information of the card punch and registering and recording the personal information;
and the miner data storage module is used for storing daily attendance records of miners and storing the face feature vectors extracted by the face feature extraction module and the job numbers and names of the miners acquired by the miner information acquisition module.
Further, in the above system, the miner information collecting module further includes:
and the face image enhancement module is used for enhancing the acquired face image and removing redundant noise by using an image preprocessing algorithm of bilateral filtering and Retinex fusion, and then sending the enhanced image to the attendance recognition module.
Further, in the above system, the attendance identification module further includes:
the human face position detection module is used for inputting the enhanced human face image into a pre-trained SSD target detection network so as to realize human face position detection in the image and cut the detected human face image;
the human face feature extraction module is used for inputting the cut human face image into a pre-trained ResNet-18 feature extraction network based on an attention mechanism so as to realize the extraction work of human face features;
and the face feature matching module is used for comparing and identifying the extracted feature vectors with the face feature vectors marked in the database, judging by calculating cosine similarity between the extracted feature vectors and the marked face feature vectors, if the extracted feature vectors are larger than a threshold value of 0.6, successfully checking in the well, storing the face feature vectors into the miner data storage module for storage, and otherwise, refusing to enter the well.
The method comprises the steps of enhancing and denoising the collected face image through an improved image preprocessing algorithm, detecting and cutting a face part through an SSD target detection network, extracting face features through an attention mechanism improved ResNet-18 feature extraction network, and finally realizing attendance recognition of miners through calculating cosine similarity between extracted feature vectors and database marked face feature vectors.
The invention can effectively enhance the quality of the face image through an improved image preprocessing algorithm, reduce the influence of interference factors such as low illumination, coal dust and the like under the well, strengthen the characteristic extraction capability of the ResNet-18 network by utilizing an attention mechanism, fully improve the accuracy of face recognition, avoid the defects of the traditional attendance mode and be a convenient and efficient attendance system.
Drawings
Fig. 1 is a block diagram of the structure of the miner's face recognition system according to the present invention.
Fig. 2 is a flow chart of the facial image enhancement module according to the present invention.
Fig. 3 is a flow chart of the face position detection module according to the present invention.
FIG. 4 is a flow chart of the face feature extraction module of the present invention.
FIG. 5 is a flow chart of the face feature matching module of the present invention.
Detailed Description
In order to better explain the present invention, the present invention will be described in further detail with reference to the accompanying drawings and detailed description.
As shown in fig. 1, the present invention provides a human face recognition system for miners based on a neural network, comprising:
the miner information acquisition module is used for calling the camera to acquire a face image of the miner and acquiring the job number and name of the miner;
the attendance checking identification module is used for comparing and identifying the face information acquired by the camera with the face data in the database, outputting the personal information of the card punch and registering and recording the personal information;
and the miner data storage module is used for storing daily attendance records of miners and storing the face feature vectors extracted by the face feature extraction module and the job numbers and names of the miners acquired by the miner information acquisition module.
The miner information acquisition module further comprises:
a face image enhancement module for enhancing the collected face image and removing redundant noise by using an image preprocessing algorithm of bilateral filtering and Retinex fusion, then sending the enhanced image to an attendance recognition module,
specifically, fig. 2 is a flow chart of a face image enhancement module provided by the embodiment of the present invention, as shown in fig. 2, including the technical solutions described in the following steps 1 to 4.
Step 1, taking a face image collected by a camera as an input image, performing logarithmic domain transformation, and then outputting.
And 2, filtering low-frequency components, namely illumination components, in the face image through bilateral filtering, and outputting the rest reflection components, so that the aim of separating the illumination components from the reflection components is fulfilled.
And 3, performing weighted dot multiplication on the reflection component and the color recovery factor, and outputting the result.
And 4, performing exponential domain transformation on the dot product result to obtain a final output image, and sending the enhanced face image to an attendance checking identification module.
The attendance identification module further comprises:
and the face position detection module is used for inputting the enhanced face image into the SSD target detection network which is pre-trained on the LFW large-scale face data set, and a loss function used by the training network is SoftMax, and the function can enable the network to obtain good classification capability. Then, feature images of different scales are obtained through a feature pyramid in a VGG-16 basic network, then default prior frames of different sizes are generated on the feature images of different scales, prior frames which do not belong to the face class are filtered out through a class confidence threshold, redundant prior frames are filtered out through an NMS algorithm, and finally the remaining prior frames are the detected face of the target miner, and are cut out and sent to a face feature extraction module.
Specifically, fig. 3 is a flowchart of a face position detection module according to an embodiment of the present invention, and as shown in fig. 3, the flowchart includes the following technical solutions described in steps 5 to 9.
And 5, sending the input image into a pre-trained SSD target detection network, and obtaining feature maps with different scales through a feature pyramid in a VGG-16 basic network.
And 6, generating default prior frames with different sizes on the feature maps with different scales.
And 7, filtering out prior frames which do not belong to the face class, namely removing background frames through a class confidence threshold (the threshold is 0.5).
And 8, filtering redundant prior boxes by an NMS algorithm, wherein the steps comprise the following step q 1-step q 3.
Step q1, screening out the bounding box M with the maximum category score from the set B of all bounding boxes;
step q2, moving M from the set B to a final result set D;
step q3, in B, the remaining bounding box with the IoU value of M greater than the threshold F (the present threshold takes 0.6) is removed.
Step q4, repeat q1 through q3 until all bounding boxes in B are removed.
And 9, cutting the detected face of the target miner, and sending the cut face to a face feature extraction module.
The face feature extraction module is used for inputting the cut face images into a ResNet-18 feature extraction network which is pre-trained on a large miner face data set and is based on an attention mechanism, the network comprises three layers of original convolution blocks and two layers of improved convolution blocks inserted into a mixed attention module (CBAM), the CBAM enables the network to pay more attention to important feature information in the face images, so that the feature images extracted by the network contain richer semantic contents, a loss function used by the training network is ArcFace, the function can further enhance the clustering capability of the network, the purposes of increasing the inter-class distance and reducing the intra-class distance are achieved, and the effect of improving the accuracy of a face classification model is achieved.
Specifically, fig. 4 is a flow chart of a face feature extraction module provided in the embodiment of the present invention, as shown in fig. 4, including the technical solutions described in the following steps 10 to 15.
Step 10, the input image is sent to a pre-trained ResNet-18 feature extraction network inserted into a CBAM module, firstly, a feature map with the size of 112 x 64 is obtained through a first layer convolution block conv1, and the feature map is sent to a next layer convolution block.
And step 11, inputting the feature map output by the previous layer, obtaining the feature map with the size of 56 × 56 × 64 through the second layer convolution block conv2, and sending the feature map to the next layer convolution block.
And step 12, inputting the feature map output by the previous layer, obtaining the feature map with the size of 28 multiplied by 128 through the third layer convolution block conv3, and sending the feature map to the next layer convolution block.
And step 13, inputting the feature map output by the previous layer, passing through a fourth layer convolution block conv4_ CBAM, adding a CBAM module into the layer convolution block to obtain a feature map with the size of 14 multiplied by 256, and sending the feature map into the next layer convolution block.
And step 14, inputting the feature map output by the previous layer, passing through a fifth layer convolution block conv5_ CBAM, adding a CBAM module into the layer convolution block to obtain a feature map with the size of 7 multiplied by 512, and sending the feature map into the FC layer.
And step 15, inputting the feature map output by the previous layer, obtaining a face feature vector with the size of 1 multiplied by 1000 through maximum pooling and a full connection layer, and sending the face feature vector to a face feature matching module.
And the face feature matching module is used for comparing and identifying the extracted feature vectors with the face feature vectors marked in the database, judging by calculating cosine similarity between the extracted feature vectors and the marked face feature vectors, if the extracted feature vectors are larger than a threshold value of 0.6, successfully checking in the well, storing the face feature vectors into the miner data storage module for storage, and otherwise, refusing to enter the well.
Specifically, fig. 5 is a flow chart of a face feature matching module provided in the embodiment of the present invention, as shown in fig. 5, including the technical solutions described in the following steps 16 to 18.
And step 16, calculating cosine similarity between the input feature vector and the marked feature vector of the database.
In step 17, the calculation result is subjected to threshold judgment (the threshold value is 0.6).
Step 18, if the human face characteristic vector is larger than the threshold value, the same person is considered to be successfully checked in the attendance card, the person is allowed to enter the well, and then the human face characteristic vector extracted this time is input into a database in a miner data storage module for storage; if the attendance is smaller than the threshold value, the attendance checking of different personnel fails, and the personnel is refused to enter the well.
The above embodiments and specific parameters in the embodiments are only used for explicitly describing the implementation procedures of the present invention, and not for limiting the patent scope of the present invention, and all equivalent structures or procedures that are used in the content of the description and drawings of the present invention or are directly or indirectly applied to other related technical fields shall be included in the protection scope of the appended claims of the present invention.

Claims (5)

1. A face recognition system for miners based on a neural network, comprising:
the miner information acquisition module is used for calling the camera to acquire a face image of the miner and acquiring the job number and name of the miner;
the attendance checking identification module is used for comparing and identifying the face information acquired by the camera with the face data in the database, outputting the personal information of the card punch and registering and recording the personal information;
and the miner data storage module is used for storing daily attendance records of miners and storing the face feature vectors extracted by the face feature extraction module and the job numbers and names of the miners acquired by the miner information acquisition module.
2. The miner face recognition system of claim 1, further comprising:
and the face image enhancement module is used for enhancing the acquired face image and removing redundant noise by using an image preprocessing algorithm of bilateral filtering and Retinex fusion, and then sending the enhanced image to the attendance recognition module.
3. The miner face recognition system of claim 1, further comprising:
and the face position detection module is used for inputting the enhanced face image into a pre-trained SSD target detection network so as to realize face position detection in the image and cut the detected face image.
4. The miner face recognition system of claim 1, further comprising:
and the face feature extraction module is used for inputting the cut face image into a pre-trained ResNet-18 feature extraction network based on an attention mechanism so as to realize the extraction work of the face features.
5. The miner face recognition system of claim 1, further comprising:
and the face feature matching module is used for comparing and identifying the extracted feature vectors with the face feature vectors marked in the database, judging by calculating cosine similarity between the extracted feature vectors and the marked face feature vectors, if the extracted feature vectors are larger than a threshold value of 0.6, successfully checking in the well, storing the face feature vectors into the miner data storage module for storage, and otherwise, refusing to enter the well.
CN202210460972.3A 2022-04-28 2022-04-28 Miner face recognition system based on neural network Pending CN114821725A (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN202210460972.3A CN114821725A (en) 2022-04-28 2022-04-28 Miner face recognition system based on neural network

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN202210460972.3A CN114821725A (en) 2022-04-28 2022-04-28 Miner face recognition system based on neural network

Publications (1)

Publication Number Publication Date
CN114821725A true CN114821725A (en) 2022-07-29

Family

ID=82508561

Family Applications (1)

Application Number Title Priority Date Filing Date
CN202210460972.3A Pending CN114821725A (en) 2022-04-28 2022-04-28 Miner face recognition system based on neural network

Country Status (1)

Country Link
CN (1) CN114821725A (en)

Cited By (2)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN116680622A (en) * 2023-08-03 2023-09-01 西安核音智言科技有限公司 Residual LSTM network-based coal mine well logging personnel prediction method
CN117218783A (en) * 2023-09-12 2023-12-12 广东云百科技有限公司 Internet of things safety management system and method

Cited By (3)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN116680622A (en) * 2023-08-03 2023-09-01 西安核音智言科技有限公司 Residual LSTM network-based coal mine well logging personnel prediction method
CN116680622B (en) * 2023-08-03 2023-10-20 西安核音智言科技有限公司 Residual LSTM network-based coal mine well logging personnel prediction method
CN117218783A (en) * 2023-09-12 2023-12-12 广东云百科技有限公司 Internet of things safety management system and method

Similar Documents

Publication Publication Date Title
Asghar et al. Copy-move and splicing image forgery detection and localization techniques: a review
Pourshahabi et al. Offline handwritten signature identification and verification using contourlet transform
CN108664996A (en) A kind of ancient writing recognition methods and system based on deep learning
CN114821725A (en) Miner face recognition system based on neural network
CN105678612A (en) Mobile terminal original certificate electronic intelligent filling system and method
CN109446345A (en) Nuclear power file verification processing method and system
CN108323209A (en) Information processing method, system, cloud processing device and computer program product
Kalunga et al. Development of fingerprint biometrics verification and vetting management system
CN110222660B (en) Signature authentication method and system based on dynamic and static feature fusion
SP et al. Automatic Number Plate Recognition System for Entry and Exit Management
CN117558011B (en) Image text tampering detection method based on self-consistency matrix and multi-scale loss
CN104463091A (en) Face image recognition method based on LGBP feature subvectors of image
CN114140025A (en) Multi-modal data-oriented vehicle insurance fraud behavior prediction system, method and device
Pinthong et al. Globally Harmonized System Label detection using Color Segmentation
Pinthong et al. The License Plate Recognition system for tracking stolen vehicles
Piekarczyk et al. Hierarchical Graph-Grammar Model for Secure and Efficient Handwritten Signatures Classification.
CN115795079A (en) Engineering cost analysis data acquisition and processing method and system
CN114596439A (en) Image-based intelligent damage assessment and check method and device for car insurance and electronic equipment
CN114663899A (en) Financial bill processing method, device, equipment and medium
CN105718972A (en) Intelligent acquiring method of information
Soma et al. An automated fraud detection of hall ticket in an offline examination system using ANN classifier
Girinath et al. Automatic Number Plate Detection using Deep Learning
Rani et al. Recognition of High-Security Registration Plate for Indian Vehicles
Jenitta et al. License Plate Recognition Using Raspberry Pi
Yuan et al. Study on forensic image retrieval

Legal Events

Date Code Title Description
PB01 Publication
PB01 Publication
SE01 Entry into force of request for substantive examination
SE01 Entry into force of request for substantive examination