CN111931634A - Deep learning-based campus protection method and system - Google Patents
Deep learning-based campus protection method and system Download PDFInfo
- Publication number
- CN111931634A CN111931634A CN202010784304.7A CN202010784304A CN111931634A CN 111931634 A CN111931634 A CN 111931634A CN 202010784304 A CN202010784304 A CN 202010784304A CN 111931634 A CN111931634 A CN 111931634A
- Authority
- CN
- China
- Prior art keywords
- feature
- neural network
- video
- picture
- network model
- 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
Links
Images
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V40/00—Recognition of biometric, human-related or animal-related patterns in image or video data
- G06V40/10—Human or animal bodies, e.g. vehicle occupants or pedestrians; Body parts, e.g. hands
- G06V40/16—Human faces, e.g. facial parts, sketches or expressions
- G06V40/161—Detection; Localisation; Normalisation
- G06V40/166—Detection; Localisation; Normalisation using acquisition arrangements
-
- 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
- G06N3/045—Combinations of networks
-
- 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
- G06N3/08—Learning methods
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V20/00—Scenes; Scene-specific elements
- G06V20/40—Scenes; Scene-specific elements in video content
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V40/00—Recognition of biometric, human-related or animal-related patterns in image or video data
- G06V40/10—Human or animal bodies, e.g. vehicle occupants or pedestrians; Body parts, e.g. hands
- G06V40/16—Human faces, e.g. facial parts, sketches or expressions
- G06V40/168—Feature extraction; Face representation
Landscapes
- Engineering & Computer Science (AREA)
- Physics & Mathematics (AREA)
- Theoretical Computer Science (AREA)
- Health & Medical Sciences (AREA)
- General Physics & Mathematics (AREA)
- General Health & Medical Sciences (AREA)
- Multimedia (AREA)
- Oral & Maxillofacial Surgery (AREA)
- Biomedical Technology (AREA)
- Mathematical Physics (AREA)
- Data Mining & Analysis (AREA)
- Molecular Biology (AREA)
- Computing Systems (AREA)
- General Engineering & Computer Science (AREA)
- Computational Linguistics (AREA)
- Evolutionary Computation (AREA)
- Software Systems (AREA)
- Biophysics (AREA)
- Artificial Intelligence (AREA)
- Human Computer Interaction (AREA)
- Life Sciences & Earth Sciences (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Image Analysis (AREA)
Abstract
The invention provides a campus protection method and system based on deep learning, wherein the method comprises the following steps: the method comprises the steps that a video camera is used for collecting and monitoring personnel conditions of all road sections or places in a campus in real time, and a plurality of images corresponding to videos are obtained; acquiring identity pictures corresponding to school students and school teaching employees stored in a school information system; establishing a neural network model for face recognition, and training the neural network model by using the identity picture; taking a plurality of images acquired by the video as the input of the trained neural network model, identifying the plurality of images through the neural network, and outputting an identification result; and judging whether the personnel of the non-school students and the teaching staff appear or not, and integrating the movement routes of the personnel of the non-school students and the teaching staff. The system comprises modules corresponding to the method steps.
Description
Technical Field
The invention provides a campus protection method and system based on deep learning, and belongs to the technical field of protection.
Background
The school is the main place of student's study and life, and its security is especially important, and current campus safety control is mainly video monitoring basically, but, and current video monitoring only stops just video acquisition broadcast and the mode of artifical monitoring video image carries out campus safety protection, but this kind of mode safety monitoring efficiency is lower, in case go wrong, needs the manual work to investigate through the video recording and obtains non-school personnel, wastes time and energy, and efficiency is extremely low.
Disclosure of Invention
The invention provides a campus protection method and system based on deep learning, which are used for solving the problems that in the existing campus video monitoring process, people outside a campus cannot be automatically identified, and the safety protection performance is poor, and adopt the following technical scheme:
a deep learning based campus protection method, the method comprising:
the method comprises the steps that a video camera is used for collecting and monitoring personnel conditions of all road sections or places in a campus in real time, and a plurality of images corresponding to videos are obtained;
acquiring identity pictures corresponding to school students and school teaching employees stored in a school information system;
establishing a neural network model for face recognition, and training the neural network model by using the identity picture;
taking a plurality of images acquired by the video as the input of the trained neural network model, identifying the plurality of images through the neural network, and outputting an identification result;
and judging whether the personnel of the non-school students and the teaching staff appear or not, and integrating the movement routes of the personnel of the non-school students and the teaching staff.
Further, training the neural network model by using the identity picture, including:
performing resolution reduction processing on the identity photo to obtain an initial processing picture with reduced resolution;
performing resolution increasing processing on the identity photo to obtain a secondary processing picture with increased resolution;
inputting the initial processing picture into the neural network model, extracting features through the neural network model to obtain each feature corresponding to the primary processing picture, and integrating each feature into an initial feature set;
inputting the secondary processing picture into the neural network model, extracting features through the neural network model to obtain each feature corresponding to the secondary processing picture, and integrating each feature into a secondary feature set;
comparing the feature set corresponding to the primary processing picture with the feature set corresponding to the secondary processing picture, and adjusting each feature in the feature set of the primary processing picture to be the same as the maximum approach of each feature in the feature set of the secondary processing picture to obtain an adjusted primary feature set;
restoring the primary processing picture according to each feature in the adjusted primary feature set to obtain an adjusted picture of the primary processing picture;
and distinguishing and identifying the adjusted primary processing picture and the secondary processing picture, and determining to finish the neural network training when the difference between the high-resolution picture and the secondary processing picture cannot be identified.
Further, the step of using the plurality of images acquired by the video as the input of the trained neural network model, performing recognition processing on the plurality of images through the neural network, and outputting a recognition result includes:
inputting a plurality of video images acquired by the video into a neural network model, and adjusting the image resolution of the video images through the neural network model;
extracting feature points corresponding to facial features appearing in the video image and position information corresponding to the feature points, wherein the feature points refer to pixel points of the facial features; the position information refers to corresponding positions of the pixel points distributed in a face range;
integrating the position information into a feature vector, comparing the feature vector with feature vectors formed by feature points of facial features corresponding to the identity photo and the position information of the feature points, and determining the similarity between the feature vectors through a similarity model to realize the recognition of the facial image;
and when the similarity is lower than a preset similarity threshold, determining that the personnel corresponding to the face image are the personnel of non-school students and teaching workers, and storing a feature set corresponding to the face image of the personnel.
Further, the similarity model is:
wherein sim (A, B) represents the similarity between the feature vectors of the video images and the feature vectors corresponding to the identity photos; a represents a feature vector of the video image, and a ═ α1、α2……αn) (ii) a B represents a feature vector corresponding to the identity photograph, and B ═ β1、β2……βn) (ii) a And when sim (A, B) < 0.87, the similarity is lower than a preset similarity threshold value.
Further, the determining whether the presence of the people of the non-school students and the teaching staff exists and integrating the movement routes of the people of the non-school students and the teaching staff includes:
identifying whether the person exists in each video image or not by comparing feature sets corresponding to face images of the persons of non-school students and teaching staff, marking the video image in which the person appears, recording the time of the person appearing in each video scene, and obtaining a time line of the person;
integrating the video images according to the time line, identifying the campus places and positions corresponding to the video images, and obtaining position information of personnel of non-school students and teaching staff corresponding to the time line;
and integrating the position information and the time line to obtain the movement route and the time line of the personnel of the non-school students and the teaching staff.
A deep learning based campus protection system, the system comprising:
the image acquisition module is used for acquiring and monitoring the personnel conditions of each road section or place in the campus in real time by using the video camera and acquiring a plurality of images corresponding to the video;
the identity acquisition module is used for acquiring identity pictures corresponding to school students and school teaching staff stored in the school information system;
the model building and training module is used for building a neural network model for face recognition and training the neural network model by using the identity picture;
the neural network operation module is used for taking the plurality of images acquired by the video as the input of the trained neural network model, identifying the plurality of images through the neural network and outputting an identification result;
and the judging module is used for judging whether the personnel of the non-school students and the teaching staff appear or not and integrating the movement routes of the personnel of the non-school students and the teaching staff.
Further, the model building training module comprises:
the resolution reduction module is used for carrying out resolution reduction processing on the identity photo to obtain an initial processing picture with reduced resolution;
the resolution increasing module is used for increasing the resolution of the identity photo to obtain a secondary processing picture with increased resolution;
the first feature extraction module is used for inputting the initial processing picture into the neural network model, extracting features through the neural network model to obtain each feature corresponding to the primary processing picture, and integrating each feature into an initial feature set;
the second feature extraction module is used for inputting the secondary processed picture into the neural network model, extracting features through the neural network model to obtain each feature corresponding to the secondary processed picture, and integrating each feature into a secondary feature set;
the feature adjusting module is used for comparing the feature set corresponding to the primary processing picture with the feature set corresponding to the secondary processing picture, adjusting each feature in the feature set of the primary processing picture to be the same as the maximum approach of each feature in the feature set of the secondary processing picture, and obtaining an adjusted primary feature set;
the image processing module is used for restoring the primary processing picture according to each feature in the adjusted primary feature set and acquiring the adjusted picture of the primary processing picture;
and the identification module is used for distinguishing and identifying the adjusted primary processing picture and the secondary processing picture, and when the difference between the high-resolution picture and the secondary processing picture cannot be identified, the neural network training is determined to be finished.
Further, the neural network operation module includes:
the image input module is used for inputting a plurality of video images acquired by the video into a neural digging network model and adjusting the image resolution of the video images through the neural network model;
the extraction module is used for extracting feature points corresponding to facial features appearing in the video image and position information corresponding to the feature points, wherein the feature points refer to pixel points of the facial features; the position information refers to corresponding positions of the pixel points distributed in a face range;
the comparison module is used for integrating the position information into a feature vector, comparing the feature vector with feature vectors formed by feature points of facial features corresponding to the identity photos and the position information of the feature points, and determining the similarity between the feature vectors through a similarity model to realize the recognition of the facial image;
and the storage module is used for determining that the personnel corresponding to the face image are the personnel of non-school students and teaching workers when the similarity is lower than a preset similarity threshold, and storing the feature set corresponding to the face image of the personnel.
Further, the similarity model is:
wherein sim (A, B) represents the feature vector of the video image and the feature vector corresponding to the identity photoSimilarity between quantities; a represents a feature vector of the video image, and a ═ α1、α2……αn) (ii) a B represents a feature vector corresponding to the identity photograph, and B ═ β1、β2……βn) (ii) a And when sim (A, B) < 0.87, the similarity is lower than a preset similarity threshold value.
Further, the judging module comprises:
the time module is used for identifying whether the person exists in each video image or not through comparison of feature sets corresponding to the face images of the persons of the non-school students and the teaching staff, marking the video image in which the person appears, recording the time of the person appearing in each video scene, and obtaining a time line of the person;
the position information acquisition module is used for integrating the video images according to the time line, identifying the campus places and positions corresponding to the video images and acquiring the position information of the non-school students and the staff of the teaching staff corresponding to the time line;
and the integration module is used for integrating the position information and the time line to obtain the movement route and the time line of the personnel of the non-school students and the teaching staff.
The invention has the beneficial effects that:
the campus protection method and system based on deep learning can effectively improve the identification accuracy rate and the identification corresponding time of the personnel outside the campus, so that the identification accuracy rate of the personnel outside the campus is up to 92%, and the identification efficiency and the monitoring strength of the personnel outside the campus are effectively improved. According to the invention, the personnel outside the school can be automatically identified through the video monitoring image, and the operation track and the time line are locked, so that the consumption of human resources is greatly reduced. The method has the advantages that the resolution ratio processing is carried out on the images acquired by the video through the neural network model, so that the acquired video images with lower quality can be adjusted to have higher image quality, the accuracy of face recognition is improved, and the phenomenon of inaccurate face recognition caused by poor video image quality due to the problems of weather, light, angles and the like is effectively prevented. Meanwhile, the feature vectors of the video images are compared with the feature vectors corresponding to the identity information pictures of the school system, so that the effective identification of the personnel outside the school is realized, the identification accuracy is ensured, the response time of face identification is effectively shortened, and the face identification speed is increased. And the whole course of movement of the personnel outside the school can be monitored by acquiring the running track and the time line of the personnel outside the school through information integration, the strength and the efficiency of campus protection are improved to a great extent, and the traditional campus monitoring and protection manpower is effectively reduced.
Drawings
FIG. 1 is a flow chart of the method of the present invention;
fig. 2 is a schematic diagram of the system of the present invention.
Detailed Description
The preferred embodiments of the present invention will be described in conjunction with the accompanying drawings, and it will be understood that they are described herein for the purpose of illustration and explanation and not limitation.
The invention provides a campus protection method and system based on deep learning, which are used for solving the problem that people outside a campus cannot be automatically identified in the existing campus video monitoring process.
The embodiment of the invention provides a campus protection method based on deep learning, and as shown in fig. 1, the method comprises the following steps:
s1, acquiring and monitoring the personnel conditions of each road section or place in the campus in real time by using a video camera, and acquiring a plurality of images corresponding to the video;
s2, acquiring identity pictures corresponding to school students and school teaching employees stored in the school information system;
s3, establishing a neural network model for face recognition, and training the neural network model by using the identity picture;
s4, taking the plurality of images acquired by the video as the input of the trained neural network model, identifying the plurality of images through the neural network, and outputting an identification result;
and S5, judging whether the personnel of the non-school students and the teaching staff appear or not, and integrating the movement routes of the personnel of the non-school students and the teaching staff.
The working principle of the technical scheme is as follows: firstly, acquiring and monitoring personnel conditions of each road section or place in a campus in real time by using a video camera, and acquiring a plurality of images corresponding to videos; then, acquiring identity pictures corresponding to school students and school teaching employees stored in the school information system; then, establishing a neural network model for face recognition, and training the neural network model by using the identity picture; then, taking a plurality of images acquired by the video as the input of the trained neural network model, carrying out recognition processing on the plurality of images through the neural network, and outputting a recognition result; and finally, judging whether the personnel of the non-school students and the teaching staff appear or not, and integrating the movement routes of the personnel of the non-school students and the teaching staff.
The effect of the above technical scheme is as follows: the method can effectively improve the identification accuracy and the identification corresponding time of the personnel outside the school, so that the identification accuracy of the personnel outside the school is up to 92 percent, and the identification efficiency and the monitoring strength of the personnel outside the school are effectively improved. According to the invention, the personnel outside the school can be automatically identified through the video monitoring image, and the operation track and the time line are locked, so that the consumption of human resources is greatly reduced. The method has the advantages that the resolution ratio processing is carried out on the images acquired by the video through the neural network model, so that the acquired video images with lower quality can be adjusted to have higher image quality, the accuracy of face recognition is improved, and the phenomenon of inaccurate face recognition caused by poor video image quality due to the problems of weather, light, angles and the like is effectively prevented. Meanwhile, the feature vectors of the video images are compared with the feature vectors corresponding to the identity information pictures of the school system, so that the effective identification of the personnel outside the school is realized, the identification accuracy is ensured, the response time of face identification is effectively shortened, and the face identification speed is increased. And the whole course of movement of the personnel outside the school can be monitored by acquiring the running track and the time line of the personnel outside the school through information integration, the strength and the efficiency of campus protection are improved to a great extent, and the traditional campus monitoring and protection manpower is effectively reduced.
In an embodiment of the present invention, training the neural network model using the identity picture includes:
s301, performing resolution reduction processing on the identity photo to obtain an initial processing picture with reduced resolution;
s302, performing resolution increasing processing on the identity photo to obtain a secondary processing picture with increased resolution;
s303, inputting the initial processing picture into the neural network model, extracting features through the neural network model to obtain each feature corresponding to the primary processing picture, and integrating each feature into an initial feature set;
s304, inputting the secondary processing picture into the neural network model, extracting features through the neural network model to obtain each feature corresponding to the secondary processing picture, and integrating each feature into a secondary feature set;
s305, comparing the feature set corresponding to the primary processed picture with the feature set corresponding to the secondary processed picture, and adjusting each feature in the feature set of the primary processed picture to be the same as the maximum approach of each feature in the feature set of the secondary processed picture to obtain an adjusted primary feature set;
s306, restoring the primary processing picture according to each feature in the adjusted primary feature set, and obtaining an adjusted picture of the primary processing picture;
s307, distinguishing and identifying the adjusted primary processing picture and the secondary processing picture, and determining to finish the neural network training when the difference between the high-resolution picture and the secondary processing picture cannot be identified.
The effect of the above technical scheme is as follows: by training the neural network module in the mode, the neural network model has a strong image quality adjusting function, and images with low quality caused by environment, optical fibers, weather and angles can achieve image recognition with high accuracy. The accuracy of pattern recognition is effectively improved, and meanwhile, the trained neural network model still has a high corresponding speed of recognition on video pictures with poor quality, so that the efficiency and the speed of face recognition are effectively improved. The problem that the face recognition efficiency and accuracy are reduced due to the fact that the corresponding time process is recognized due to the fact that the quality of the video image is low is solved.
In an embodiment of the present invention, the taking a plurality of images acquired by the video as the input of the trained neural network model, performing recognition processing on the plurality of images through the neural network, and outputting a recognition result includes:
s401, inputting a plurality of video images acquired by the video into a neural network model, and adjusting the image resolution of the video images through the neural network model;
s402, extracting feature points corresponding to facial features appearing in the video image and position information corresponding to the feature points, wherein the feature points refer to pixel points of the facial features; the position information refers to corresponding positions of the pixel points distributed in a face range;
s403, integrating the position information into a feature vector, comparing the feature vector with feature vectors formed by feature points of facial features corresponding to the identity photo and the position information of the facial features, and determining the similarity between the feature vectors through a similarity model to realize the recognition of the facial image;
s404, when the similarity is lower than a preset similarity threshold, determining that the personnel corresponding to the face image are the personnel of non-school students and teaching workers, and storing a feature set corresponding to the face image of the personnel.
The working principle of the technical scheme is as follows: firstly, inputting a plurality of video images acquired by a video into a neural network model, and adjusting the image resolution of the video images through the neural network model; then, extracting feature points corresponding to facial features appearing in the video image and position information corresponding to the feature points, wherein the feature points refer to pixel points of the facial features; the position information refers to corresponding positions of the pixel points distributed in a face range; then, integrating the position information into a feature vector, comparing the feature vector with feature vectors formed by feature points of facial features corresponding to the identity photo and the position information of the facial features, and determining the similarity between the feature vectors through a similarity model to realize the recognition of the facial image; and then, when the similarity is lower than a preset similarity threshold, determining that the personnel corresponding to the face image are the personnel of non-school students and teaching workers, and storing a feature set corresponding to the face image of the personnel.
The effect of the above technical scheme is as follows: the feature vectors of the video images are compared with the feature vectors corresponding to the identity information pictures of the school system, so that the effective identification of the personnel outside the school is realized, the identification accuracy is ensured, the response time of face identification is effectively shortened, and the face identification speed is increased.
In an embodiment of the present invention, the similarity model is:
wherein sim (A, B) represents the similarity between the feature vectors of the video images and the feature vectors corresponding to the identity photos; a represents a feature vector of the video image, and a ═ α1、α2……αn) (ii) a B represents a feature vector corresponding to the identity photograph, and B ═ β1、β2……βn) (ii) a And when sim (A, B) < 0.87, the similarity is lower than a preset similarity threshold value.
The effect of the above technical scheme is as follows: the similarity obtained by the similarity model can effectively improve the accuracy of similarity judgment, and meanwhile, the similarity threshold value is set to be 0.87, so that the definition standard of face recognition under the conditions of campus environment, people coming and going and high personnel density can be better met. And the misjudgment and the missed judgment of the personnel outside the school are effectively avoided.
In an embodiment of the present invention, the determining whether the presence of the people of the non-school students and the teaching staff exists and integrating the movement routes of the people of the non-school students and the teaching staff includes:
s501, identifying whether the person exists in each video image or not through comparison of feature sets corresponding to face images of the persons of non-school students and teaching staff, marking the video image in which the person appears, recording the time of the person appearing in each video scene, and obtaining a timeline of the person;
s502, integrating the video images according to the time line, identifying campus places and positions corresponding to the video images, and obtaining position information of non-school students and staff of teaching staff corresponding to the time line;
and S503, integrating the position information and the time line to obtain the movement route and the time line of the personnel of the non-school students and the teaching staff.
The working principle of the technical scheme is as follows: firstly, identifying whether a person exists in each video image or not by comparing feature sets corresponding to face images of persons of non-school students and teaching staff, marking the video image in which the person appears, recording the time of the person appearing in each video scene, and obtaining a time line of the person; then, integrating the video images according to the time line, identifying the campus places and positions corresponding to the video images, and obtaining position information of personnel of non-school students and teaching staff corresponding to the time line; and finally, integrating the position information and the time line to obtain the movement route and the time line of the personnel of the non-school students and the teaching staff.
The effect of the above technical scheme is as follows: the whole-course motion monitoring of the integral type that obtains the operation track and the time line of school personnel outside through information integration can realize going on to school personnel outside, has improved the dynamics and the efficiency of campus protection to a very big extent, effectively reduces traditional campus control and protection manpower.
One embodiment of the present invention provides a deep learning-based campus protection system, as shown in fig. 2, the system includes:
the image acquisition module is used for acquiring and monitoring the personnel conditions of each road section or place in the campus in real time by using the video camera and acquiring a plurality of images corresponding to the video;
the identity acquisition module is used for acquiring identity pictures corresponding to school students and school teaching staff stored in the school information system;
the model building and training module is used for building a neural network model for face recognition and training the neural network model by using the identity picture;
the neural network operation module is used for taking the plurality of images acquired by the video as the input of the trained neural network model, identifying the plurality of images through the neural network and outputting an identification result;
and the judging module is used for judging whether the personnel of the non-school students and the teaching staff appear or not and integrating the movement routes of the personnel of the non-school students and the teaching staff.
The working principle of the technical scheme is as follows: the method comprises the steps that a video camera is utilized through an image acquisition module to acquire and monitor personnel conditions of all road sections or places in a campus in real time, and a plurality of images corresponding to videos are acquired; an identity acquisition module is adopted to acquire identity pictures corresponding to school students and school teaching staff stored in a school information system; establishing a neural network model for face recognition by using a model establishment training module, and training the neural network model by using the identity picture; adopting a neural network operation module to perform identification processing on the plurality of images and outputting an identification result; whether the personnel of the non-school students and the teaching staff appear is judged through the judging module, and the movement routes of the personnel of the non-school students and the teaching staff are integrated.
The effect of the above technical scheme is as follows: the method can effectively improve the identification accuracy and the identification corresponding time of the personnel outside the school, so that the identification accuracy of the personnel outside the school is up to 92 percent, and the identification efficiency and the monitoring strength of the personnel outside the school are effectively improved. According to the invention, the personnel outside the school can be automatically identified through the video monitoring image, and the operation track and the time line are locked, so that the consumption of human resources is greatly reduced. The method has the advantages that the resolution ratio processing is carried out on the images acquired by the video through the neural network model, so that the acquired video images with lower quality can be adjusted to have higher image quality, the accuracy of face recognition is improved, and the phenomenon of inaccurate face recognition caused by poor video image quality due to the problems of weather, light, angles and the like is effectively prevented. Meanwhile, the feature vectors of the video images are compared with the feature vectors corresponding to the identity information pictures of the school system, so that the effective identification of the personnel outside the school is realized, the identification accuracy is ensured, the response time of face identification is effectively shortened, and the face identification speed is increased. And the whole course of movement of the personnel outside the school can be monitored by acquiring the running track and the time line of the personnel outside the school through information integration, the strength and the efficiency of campus protection are improved to a great extent, and the traditional campus monitoring and protection manpower is effectively reduced.
In an embodiment of the present invention, the model building training module includes:
the resolution reduction module is used for carrying out resolution reduction processing on the identity photo to obtain an initial processing picture with reduced resolution;
the resolution increasing module is used for increasing the resolution of the identity photo to obtain a secondary processing picture with increased resolution;
the first feature extraction module is used for inputting the initial processing picture into the neural network model, extracting features through the neural network model to obtain each feature corresponding to the primary processing picture, and integrating each feature into an initial feature set;
the second feature extraction module is used for inputting the secondary processed picture into the neural network model, extracting features through the neural network model to obtain each feature corresponding to the secondary processed picture, and integrating each feature into a secondary feature set;
the feature adjusting module is used for comparing the feature set corresponding to the primary processing picture with the feature set corresponding to the secondary processing picture, adjusting each feature in the feature set of the primary processing picture to be the same as the maximum approach of each feature in the feature set of the secondary processing picture, and obtaining an adjusted primary feature set;
the image processing module is used for restoring the primary processing picture according to each feature in the adjusted primary feature set and acquiring the adjusted picture of the primary processing picture;
and the identification module is used for distinguishing and identifying the adjusted primary processing picture and the secondary processing picture, and when the difference between the high-resolution picture and the secondary processing picture cannot be identified, the neural network training is determined to be finished.
The working principle of the technical scheme is as follows: performing resolution reduction processing on the identity photo by using a resolution reduction module to obtain an initial processing picture with reduced resolution; performing resolution increasing processing on the identity photo through a resolution increasing module to obtain a secondary processing picture with increased resolution; inputting the initial processing picture into the neural network model by adopting a first feature extraction module, extracting features through the neural network model to obtain each feature corresponding to the primary processing picture, and integrating each feature into an initial feature set; inputting the secondary processing picture into the neural network model through a second feature extraction module, extracting features through the neural network model to obtain each feature corresponding to the secondary processing picture, and integrating each feature into a secondary feature set; comparing the feature set corresponding to the primary processing picture with the feature set corresponding to the secondary processing picture by adopting a feature adjusting module, and adjusting each feature in the feature set of the primary processing picture to be the same as the maximum approach of each feature in the feature set of the secondary processing picture to obtain an adjusted primary feature set; restoring the primary processing picture according to each feature in the adjusted primary feature set by using an image processing module to obtain an adjusted picture of the primary processing picture; and distinguishing and identifying the adjusted primary processing picture and the secondary processing picture through an identification module, and determining to finish the neural network training when the difference between the high-resolution picture and the secondary processing picture cannot be identified.
The effect of the above technical scheme is as follows: by training the neural network module in the mode, the neural network model has a strong image quality adjusting function, and images with low quality caused by environment, optical fibers, weather and angles can achieve image recognition with high accuracy. The accuracy of pattern recognition is effectively improved, and meanwhile, the trained neural network model still has a high corresponding speed of recognition on video pictures with poor quality, so that the efficiency and the speed of face recognition are effectively improved. The problem that the face recognition efficiency and accuracy are reduced due to the fact that the corresponding time process is recognized due to the fact that the quality of the video image is low is solved.
In one embodiment of the present invention, the neural network operation module includes:
the image input module is used for inputting a plurality of video images acquired by the video into a neural digging network model and adjusting the image resolution of the video images through the neural network model;
the extraction module is used for extracting feature points corresponding to facial features appearing in the video image and position information corresponding to the feature points, wherein the feature points refer to pixel points of the facial features; the position information refers to corresponding positions of the pixel points distributed in a face range;
the comparison module is used for integrating the position information into a feature vector, comparing the feature vector with feature vectors formed by feature points of facial features corresponding to the identity photos and the position information of the feature points, and determining the similarity between the feature vectors through a similarity model to realize the recognition of the facial image;
and the storage module is used for determining that the personnel corresponding to the face image are the personnel of non-school students and teaching workers when the similarity is lower than a preset similarity threshold, and storing the feature set corresponding to the face image of the personnel.
The working principle of the technical scheme is as follows: inputting a plurality of video images acquired by the video into a neural digging network model through an image input module, and adjusting the image resolution of the video images through the neural network model; then, extracting feature points corresponding to facial features appearing in the video image and position information corresponding to the feature points by using an extraction module, wherein the feature points refer to pixel points of the facial features; the position information refers to corresponding positions of the pixel points distributed in a face range; then, integrating the position information into a feature vector through a comparison module, comparing the feature vector with feature vectors formed by feature points of facial features corresponding to the identity photos and the position information of the feature points, and determining the similarity among the feature vectors through a similarity model to realize the recognition of the facial image; and finally, when the similarity is lower than a preset similarity threshold, determining that the personnel corresponding to the face image are the personnel of non-school students and teaching workers by using a storage module, and storing a feature set corresponding to the face image of the personnel.
The effect of the above technical scheme is as follows: the feature vectors of the video images are compared with the feature vectors corresponding to the identity information pictures of the school system, so that the effective identification of the personnel outside the school is realized, the identification accuracy is ensured, the response time of face identification is effectively shortened, and the face identification speed is increased.
In an embodiment of the present invention, the determining module includes:
the time module is used for identifying whether the person exists in each video image or not through comparison of feature sets corresponding to the face images of the persons of the non-school students and the teaching staff, marking the video image in which the person appears, recording the time of the person appearing in each video scene, and obtaining a time line of the person;
the position information acquisition module is used for integrating the video images according to the time line, identifying the campus places and positions corresponding to the video images and acquiring the position information of the non-school students and the staff of the teaching staff corresponding to the time line;
and the integration module is used for integrating the position information and the time line to obtain the movement route and the time line of the personnel of the non-school students and the teaching staff.
The working principle of the technical scheme is as follows: identifying whether the person exists in each video image by utilizing a time module through comparison of feature sets corresponding to face images of persons of non-school students and teaching staff, marking the video image in which the person appears, recording the time of the person appearing in each video scene, and obtaining a time line of the person; integrating the video images according to the time line through a position information acquisition module, identifying the campus places and positions corresponding to the video images, and acquiring position information of non-school students and staff of teaching staff corresponding to the time line; and then, integrating the position information and the time line by using an integration module to obtain the movement route and the time line of the personnel of the non-school students and the teaching staff.
The effect of the above technical scheme is as follows: the whole-course motion monitoring of the integral type that obtains the operation track and the time line of school personnel outside through information integration can realize going on to school personnel outside, has improved the dynamics and the efficiency of campus protection to a very big extent, effectively reduces traditional campus control and protection manpower.
It will be apparent to those skilled in the art that various changes and modifications may be made in the present invention without departing from the spirit and scope of the invention. Thus, if such modifications and variations of the present invention fall within the scope of the claims of the present invention and their equivalents, the present invention is also intended to include such modifications and variations.
Claims (10)
1. A campus protection method based on deep learning is characterized in that the method comprises the following steps:
the method comprises the steps that a video camera is used for collecting and monitoring personnel conditions of all road sections or places in a campus in real time, and a plurality of images corresponding to videos are obtained;
acquiring identity pictures corresponding to school students and school teaching employees stored in a school information system;
establishing a neural network model for face recognition, and training the neural network model by using the identity picture;
taking a plurality of images acquired by the video as the input of the trained neural network model, identifying the plurality of images through the neural network, and outputting an identification result;
and judging whether the personnel of the non-school students and the teaching staff appear or not, and integrating the movement routes of the personnel of the non-school students and the teaching staff.
2. The method of claim 1, wherein training the neural network model using the identity picture comprises:
performing resolution reduction processing on the identity photo to obtain an initial processing picture with reduced resolution;
performing resolution increasing processing on the identity photo to obtain a secondary processing picture with increased resolution;
inputting the initial processing picture into the neural network model, extracting features through the neural network model to obtain each feature corresponding to the primary processing picture, and integrating each feature into an initial feature set;
inputting the secondary processing picture into the neural network model, extracting features through the neural network model to obtain each feature corresponding to the secondary processing picture, and integrating each feature into a secondary feature set;
comparing the feature set corresponding to the primary processing picture with the feature set corresponding to the secondary processing picture, and adjusting each feature in the feature set of the primary processing picture to be the same as the maximum approach of each feature in the feature set of the secondary processing picture to obtain an adjusted primary feature set;
restoring the primary processing picture according to each feature in the adjusted primary feature set to obtain an adjusted picture of the primary processing picture;
and distinguishing and identifying the adjusted primary processing picture and the secondary processing picture, and determining to finish the neural network training when the difference between the high-resolution picture and the secondary processing picture cannot be identified.
3. The method of claim 1, wherein the taking a plurality of images of the video capture as input of the trained neural network model, performing recognition processing on the plurality of images through the neural network, and outputting a recognition result comprises:
inputting a plurality of video images acquired by the video into a neural network model, and adjusting the image resolution of the video images through the neural network model;
extracting feature points corresponding to five sense organs of a face part appearing in the video image and position information corresponding to the feature points;
integrating the position information into a feature vector, comparing the feature vector with feature vectors formed by feature points of facial features corresponding to the identity photo and the position information of the feature points, and determining the similarity between the feature vectors through a similarity model to realize the recognition of the facial image;
and when the similarity is lower than a preset similarity threshold, determining that the personnel corresponding to the face image are the personnel of non-school students and teaching workers, and storing a feature set corresponding to the face image of the personnel.
4. The method of claim 3, wherein the similarity model is:
wherein sim (A, B) represents the similarity between the feature vectors of the video images and the feature vectors corresponding to the identity photos; a represents a feature vector of the video image, and a ═ α1、α2……αn) (ii) a B represents a feature vector corresponding to the identity photograph, and B ═ β1、β2……βn) (ii) a And when sim (A, B) < 0.87, the similarity is lower than a preset similarity threshold value.
5. The method of claim 1, wherein said determining whether the presence of people other than school students and professors is present and integrating movement routes of the people other than school students and professors comprises:
identifying whether the person exists in each video image or not by comparing feature sets corresponding to face images of the persons of non-school students and teaching staff, marking the video image in which the person appears, recording the time of the person appearing in each video scene, and obtaining a time line of the person;
integrating the video images according to the time line, identifying the campus places and positions corresponding to the video images, and obtaining position information of personnel of non-school students and teaching staff corresponding to the time line;
and integrating the position information and the time line to obtain the movement route and the time line of the personnel of the non-school students and the teaching staff.
6. A deep learning based campus protection system, the system comprising:
the image acquisition module is used for acquiring and monitoring the personnel conditions of each road section or place in the campus in real time by using the video camera and acquiring a plurality of images corresponding to the video;
the identity acquisition module is used for acquiring identity pictures corresponding to school students and school teaching staff stored in the school information system;
the model building and training module is used for building a neural network model for face recognition and training the neural network model by using the identity picture;
the neural network operation module is used for taking the plurality of images acquired by the video as the input of the trained neural network model, identifying the plurality of images through the neural network and outputting an identification result;
and the judging module is used for judging whether the personnel of the non-school students and the teaching staff appear or not and integrating the movement routes of the personnel of the non-school students and the teaching staff.
7. The system of claim 6, wherein the model building training module comprises:
the resolution reduction module is used for carrying out resolution reduction processing on the identity photo to obtain an initial processing picture with reduced resolution;
the resolution increasing module is used for increasing the resolution of the identity photo to obtain a secondary processing picture with increased resolution;
the first feature extraction module is used for inputting the initial processing picture into the neural network model, extracting features through the neural network model to obtain each feature corresponding to the primary processing picture, and integrating each feature into an initial feature set;
the second feature extraction module is used for inputting the secondary processed picture into the neural network model, extracting features through the neural network model to obtain each feature corresponding to the secondary processed picture, and integrating each feature into a secondary feature set;
the feature adjusting module is used for comparing the feature set corresponding to the primary processing picture with the feature set corresponding to the secondary processing picture, adjusting each feature in the feature set of the primary processing picture to be the same as the maximum approach of each feature in the feature set of the secondary processing picture, and obtaining an adjusted primary feature set;
the image processing module is used for restoring the primary processing picture according to each feature in the adjusted primary feature set and acquiring the adjusted picture of the primary processing picture;
and the identification module is used for distinguishing and identifying the adjusted primary processing picture and the secondary processing picture, and when the difference between the high-resolution picture and the secondary processing picture cannot be identified, the neural network training is determined to be finished.
8. The system of claim 6, wherein the neural network operation module comprises:
the image input module is used for inputting a plurality of video images acquired by the video into a neural digging network model and adjusting the image resolution of the video images through the neural network model;
the extraction module is used for extracting feature points corresponding to five sense organs of a face part appearing in the video image and position information corresponding to the feature points;
the comparison module is used for integrating the position information into a feature vector, comparing the feature vector with feature vectors formed by feature points of facial features corresponding to the identity photos and the position information of the feature points, and determining the similarity between the feature vectors through a similarity model to realize the recognition of the facial image;
and the storage module is used for determining that the personnel corresponding to the face image are the personnel of non-school students and teaching workers when the similarity is lower than a preset similarity threshold, and storing the feature set corresponding to the face image of the personnel.
9. The system of claim 8, wherein the similarity model is:
wherein sim (A, B) represents the similarity between the feature vectors of the video images and the feature vectors corresponding to the identity photos; a represents a feature vector of the video image, and a ═ α1、α2……αn) (ii) a B represents a feature vector corresponding to the identity photograph, and B ═ β1、β2……βn) (ii) a And when sim (A, B) < 0.87, the similarity is lower than a preset similarity threshold value.
10. The system of claim 6, wherein the determining module comprises:
the time module is used for identifying whether the person exists in each video image or not through comparison of feature sets corresponding to the face images of the persons of the non-school students and the teaching staff, marking the video image in which the person appears, recording the time of the person appearing in each video scene, and obtaining a time line of the person;
the position information acquisition module is used for integrating the video images according to the time line, identifying the campus places and positions corresponding to the video images and acquiring the position information of the non-school students and the staff of the teaching staff corresponding to the time line;
and the integration module is used for integrating the position information and the time line to obtain the movement route and the time line of the personnel of the non-school students and the teaching staff.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202010784304.7A CN111931634A (en) | 2020-08-06 | 2020-08-06 | Deep learning-based campus protection method and system |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202010784304.7A CN111931634A (en) | 2020-08-06 | 2020-08-06 | Deep learning-based campus protection method and system |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN111931634A true CN111931634A (en) | 2020-11-13 |
Family
ID=73307754
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202010784304.7A Pending CN111931634A (en) | 2020-08-06 | 2020-08-06 | Deep learning-based campus protection method and system |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN111931634A (en) |
Cited By (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN113569709A (en) * | 2021-07-23 | 2021-10-29 | 西安电子科技大学 | Campus security image recognition early warning method based on convolutional neural network |
| CN114973153A (en) * | 2022-07-27 | 2022-08-30 | 广州宏途数字科技有限公司 | Smart campus security detection method, device, equipment and storage medium |
| CN115063839A (en) * | 2022-07-01 | 2022-09-16 | 姚懿宸 | Non-standard unsafe behavior detection system in campus |
Citations (9)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN108564049A (en) * | 2018-04-22 | 2018-09-21 | 北京工业大学 | A kind of fast face detection recognition method based on deep learning |
| CN109523656A (en) * | 2018-09-27 | 2019-03-26 | 福建省南安市大大电子有限公司 | A kind of application method applied in the intelligent things management system of school |
| CN110175583A (en) * | 2019-05-30 | 2019-08-27 | 重庆跃途科技有限公司 | It is a kind of in the campus universe security monitoring analysis method based on video AI |
| CN110363181A (en) * | 2019-07-24 | 2019-10-22 | 安徽锦星信息技术有限公司 | Campus security management system based on AI recognition of face |
| CN110502999A (en) * | 2019-07-25 | 2019-11-26 | 北京华三通信技术有限公司 | Campus Management System and method |
| US20190370618A1 (en) * | 2016-06-06 | 2019-12-05 | Mutualink, Inc. | System and method for intelligent pattern recognition |
| CN110910549A (en) * | 2019-11-15 | 2020-03-24 | 江苏高泰软件技术有限公司 | Campus personnel safety management system based on deep learning and face recognition features |
| CN110941993A (en) * | 2019-10-30 | 2020-03-31 | 东北大学 | Dynamic personnel classification and storage method based on face recognition |
| CN111127317A (en) * | 2019-12-02 | 2020-05-08 | 深圳供电局有限公司 | Image super-resolution reconstruction method and device, storage medium and computer equipment |
-
2020
- 2020-08-06 CN CN202010784304.7A patent/CN111931634A/en active Pending
Patent Citations (9)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20190370618A1 (en) * | 2016-06-06 | 2019-12-05 | Mutualink, Inc. | System and method for intelligent pattern recognition |
| CN108564049A (en) * | 2018-04-22 | 2018-09-21 | 北京工业大学 | A kind of fast face detection recognition method based on deep learning |
| CN109523656A (en) * | 2018-09-27 | 2019-03-26 | 福建省南安市大大电子有限公司 | A kind of application method applied in the intelligent things management system of school |
| CN110175583A (en) * | 2019-05-30 | 2019-08-27 | 重庆跃途科技有限公司 | It is a kind of in the campus universe security monitoring analysis method based on video AI |
| CN110363181A (en) * | 2019-07-24 | 2019-10-22 | 安徽锦星信息技术有限公司 | Campus security management system based on AI recognition of face |
| CN110502999A (en) * | 2019-07-25 | 2019-11-26 | 北京华三通信技术有限公司 | Campus Management System and method |
| CN110941993A (en) * | 2019-10-30 | 2020-03-31 | 东北大学 | Dynamic personnel classification and storage method based on face recognition |
| CN110910549A (en) * | 2019-11-15 | 2020-03-24 | 江苏高泰软件技术有限公司 | Campus personnel safety management system based on deep learning and face recognition features |
| CN111127317A (en) * | 2019-12-02 | 2020-05-08 | 深圳供电局有限公司 | Image super-resolution reconstruction method and device, storage medium and computer equipment |
Non-Patent Citations (1)
| Title |
|---|
| 张怡等: "基于深度学习算法的改进型人脸识别***实现——以智慧校园安防***为例", 《信息***工程》 * |
Cited By (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN113569709A (en) * | 2021-07-23 | 2021-10-29 | 西安电子科技大学 | Campus security image recognition early warning method based on convolutional neural network |
| CN115063839A (en) * | 2022-07-01 | 2022-09-16 | 姚懿宸 | Non-standard unsafe behavior detection system in campus |
| CN114973153A (en) * | 2022-07-27 | 2022-08-30 | 广州宏途数字科技有限公司 | Smart campus security detection method, device, equipment and storage medium |
| CN114973153B (en) * | 2022-07-27 | 2022-11-04 | 广州宏途数字科技有限公司 | Smart campus security detection method, device, equipment and storage medium |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN111931634A (en) | Deep learning-based campus protection method and system | |
| CN109819208B (en) | Intensive population security monitoring management method based on artificial intelligence dynamic monitoring | |
| WO2021142902A1 (en) | Danet-based unmanned aerial vehicle coastline floating garbage inspection system | |
| CN103824070B (en) | A kind of rapid pedestrian detection method based on computer vision | |
| CN109190475B (en) | Face recognition network and pedestrian re-recognition network collaborative training method | |
| CN107481343A (en) | A kind of check class attendance based on face recognition technology is registered system and its method of work | |
| CN109376637A (en) | Passenger number statistical system based on video monitoring image processing | |
| CN110837784A (en) | Examination room peeping cheating detection system based on human head characteristics | |
| CN108921038A (en) | A kind of classroom based on deep learning face recognition technology is quickly called the roll method of registering | |
| CN106339657B (en) | Crop straw burning monitoring method based on monitor video, device | |
| CN110543811B (en) | Deep learning-based non-cooperative examination personnel management method and system | |
| CN109635758A (en) | Wisdom building site detection method is dressed based on the high altitude operation personnel safety band of video | |
| CN107230267A (en) | Intelligence In Baogang Kindergarten based on face recognition algorithms is registered method | |
| CN106541968A (en) | The subway carriage real-time prompt system of view-based access control model analysis and recognition methodss | |
| CN111914761A (en) | Thermal infrared face recognition method and system | |
| CN104463232A (en) | Density crowd counting method based on HOG characteristic and color histogram characteristic | |
| Huang et al. | Attendance system based on dynamic face recognition | |
| CN107516076A (en) | Portrait identification method and device | |
| CN111639580A (en) | Gait recognition method combining feature separation model and visual angle conversion model | |
| CN113705510A (en) | Target identification tracking method, device, equipment and storage medium | |
| CN115116137A (en) | Pedestrian detection method based on lightweight YOLO v5 network model and space-time memory mechanism | |
| TWI733616B (en) | Reconition system of human body posture, reconition method of human body posture, and non-transitory computer readable storage medium | |
| CN108307158A (en) | People's air defense method for automatically regulating, apparatus and system | |
| CN110351268B (en) | Digital law enforcement system for smart city | |
| CN108197579B (en) | Method for detecting number of people in protection cabin |
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 | ||
| RJ01 | Rejection of invention patent application after publication | ||
| RJ01 | Rejection of invention patent application after publication |
Application publication date: 20201113 |