CN111860308A

CN111860308A - Intelligent student sleep safety management method based on video behavior recognition

Info

Publication number: CN111860308A
Application number: CN202010695618.XA
Authority: CN
Inventors: 周凯凯; 高善恒
Original assignee: Suzhou Enterprise Intelligence Information Technology Co ltd
Current assignee: Suzhou Enterprise Intelligence Information Technology Co ltd
Priority date: 2020-07-17
Filing date: 2020-07-17
Publication date: 2020-10-30

Abstract

The invention discloses a student dormitory safety intelligent management method based on video behavior recognition. The method relates to the fields of informatization, video analysis and image recognition, in particular to a method for detecting object behavior specifications based on a student sleeping scene environment. The behavior specification intelligent judgment and detection method only aims at scenes of dormitory of students and abnormal early warning and checking processing, under the scene, the students need to leave the bedroom at fixed time and return to the bedroom at fixed time, and the students are in abnormal conditions no matter whether coming late, returning early or returning late, and need to be concerned and processed by school managers. Meanwhile, due to the particularity and the privacy of the bedroom, people entering the bedroom at irregular access time, especially strangers, need dormitory managers and school logistics managers to pay attention to the dormitory managers and the school logistics managers. The method comprises the steps of arranging people in a scene according to school sleeping time regulations in a school student sleeping scene, combining a face recognition technology and a video behavior analysis technology, carrying out intelligent training depending on correct and wrong behavior specifications in the scene, intelligently judging the sleeping state of the student, enabling a stranger to enter a bedroom state, and outputting abnormal behaviors in combination with illegal pictures.

Description

Intelligent student sleep safety management method based on video behavior recognition

Technical Field

The invention relates to the technical field of computers, in particular to an algorithm application based on image recognition.

Background

Students are the future of society, and all things related to students are the central importance of the whole society. For accommodation students, the safety problem is always a great concern of schools. Currently, no good means is provided for supervising the sleeping of students, and only manpower is relied on. While manpower is limited at all, for many thousands of students' dormitories, a hoster team consisting of only a few people can be called the salary cup. And if the check is relied on, the check can only be used as a temporary means and can not be normalized.

At present, compared with schools, the layout of informatization and even intellectualization is gradually completed in other fields, and the block of going back to bed management is blank all the time. The operation is simple, the labor input is low, students can not enter or exit the dormitory, and the difficulty is high. Moreover, due to the special condition of the student bedroom, the tide rule of the students entering and exiting the dormitory is obvious, and the difficulty of manpower management is greatly increased.

At present, the manpower management of dormitory management is mainly relied on in the dormitory management process, and staff is arranged to patrol at a specific time stage so as to ensure that all students are confirmed to be in dormitory as far as possible. The whole process has the following disadvantages:

1. For dormitories that have closed doors, inconvenient access checks

2. General dormitory managers cannot judge whether students on beds are themselves or not

3. The consumption of manpower resources is huge, and the long-term complete coverage inspection cannot be realized

Disclosure of Invention

Aiming at the defects, the invention provides an intelligent, fair, efficient and extensible safe and intelligent management method for sleeping in school bedrooms.

The technical scheme for solving the technical problems of the invention is as follows: a method for bedroom safety management based on video behavior recognition. The method comprises the steps of face information acquisition of school students and teaching staff, face detection and capture algorithm based on edge calculation of an intelligent camera with a calculation platform, background face detection and capture algorithm based on video image recognition and face recognition algorithm. Based on the algorithms, a set of student sleep safety level model is constructed by combining information such as the installation attribute of the camera and the installation position of the camera, and the model is used for judging the abnormal conditions of each student, outputting the abnormal conditions, generating an abnormal processing flow through an application program, processing the abnormal conditions by related personnel, and finally ensuring the sleep safety of the students.

The face collection mode provided by the invention is based on the combination of original data of a school and collection tools, and students have complete life attributes in the school, such as class, bedroom, bed and the like, and have complete in-school relationships, such as classmates, beddings and the like. And the data are important to the safety management of the student's sleep. The human face acquisition is carried out by the provided data acquisition tool, and meanwhile, the human face data is combined with other data of students to construct a simple data center station which covers structured attribute data and unstructured head portrait data. And the connection point of the data is the face ID of the student, so that the correlation of the student can be ensured based on the face.

The invention aims to acquire the face in a specific sleeping scene, and the installation position and the installation direction of a camera have great influence on the acquisition result in the scene. As the dormitory doors of a school are not too many, the relative number of acquisition points is controllable, and the school intranet is relatively smooth, the edge calculation and the center calculation do not have too many difference in quality.

The present invention also focuses on accuracy, efficiency, and cost in studying the process of data acquisition. According to the requirements of schools, if schools want to reduce false alarms, the probability of missing human faces is reduced as much as possible, double measures can be selected, namely edge calculation and center calculation are used in a combined mode, if schools pay more attention to efficiency, the human face recognition and grabbing acquisition part in calculation can be scattered to a calculation platform of a camera, and therefore processing efficiency can be guaranteed. And if the school bedroom has a high-definition camera and the angle position is appropriate, the video can be directly accessed to the GPU server cluster, and only the back-end calculation is carried out. The invention is flexible and efficient in the face acquisition part, can meet various scenes and requirements, and has flexible and changeable coping means no matter what way the face acquisition part is used.

After the face detection and the grabbing are finished, the face recognition stage is started. The invention can realize preprocessing of the images of the base library and extract the core characteristics in order to improve the efficiency of face recognition. After the camera or the GPU server cluster extracts face information, the collected faces are transmitted into the model, the model carries out face comparison based on a face comparison algorithm, face id of the introduced faces is returned after the face comparison is successful, and the id is used as a basis for subsequent algorithm processing.

In the process of studying the sleeping safety of students, the method mainly comprises the steps of extracting sleeping labels, predicting sleeping abnormity and reporting sleeping abnormity. The whole judgment process depends on a 'sleeping safety intelligent algorithm', the algorithm is based on data support of a sleeping data center, student life data, student relation data, student sleeping history data, acquisition equipment basic information, acquisition equipment history information, and other supplementary data such as week data, weather data, class schedule data and the like are combined, the probability of student sleeping abnormity is predicted according to the condition of each student, the student face comparison data is compared according to the predicted value, the student abnormity grade is evaluated, a customized safety management strategy is formulated for each student, and sleeping safety of the student is guaranteed. Based on the algorithm, the students are guaranteed to uniformly obey rules, namely basic regulations of schools are not violated, simultaneously, the entering and exiting rules of each student are intelligently calculated, safety strategies are formulated for each student and provided for teacher administrators of schools to refer to, the intelligent degree of sleeping safety is improved to the maximum extent, the management cost is reduced, and the management efficiency is improved.

The process system, namely the ' sleeping safety management system ' in the method can automatically generate early warning information and a verification process after finding that a student is abnormal in sleeping, the verification process can send the early warning process to the hands of related managers and comprises student's office officers, life teachers, dormitory managers or students, when the related managers receive abnormal reminding, verification is needed, for example, the related managers go to a specific bed in a specific dormitory to verify information, the information verification needs to comprise face photos, position information and the like, and the fact that the student is required to be in a specific place at a specific time is guaranteed. Every abnormality must have a complete process to ensure the safety of each student.

Drawings

The invention is further described below with reference to the accompanying drawings.

Fig. 1 is a schematic block diagram of an intelligent student sleep safety management method based on video behavior recognition.

Detailed Description

The invention is further described with reference to the following figures and specific examples.

The method is an intelligent sleeping safety management method based on video behavior recognition, and based on the practical situation of a school bedroom and the practical situation of a school, a proper camera is installed in a proper point position direction, the camera can be an intelligent camera carrying intelligent computing equipment or a high-definition common camera, and the camera needs to cover the whole sleeping channel. After the camera is erected, edge calculation or center calculation is selected according to the requirements of a school to perform face detection and face acquisition. The collected data are compared with one another, the comparison conclusion data are transmitted to enter a 'sleeping safety intelligent algorithm', students corresponding to the head image are judged according to the sleeping safety based on the algorithm, the judgment conclusion is output to a sleeping safety management process by the algorithm, all related persons needing to know the abnormality of the students in the school regulations can obtain corresponding messages in the process, the related persons need to track and process the messages carefully until the conclusion is finally reached, and through the complete process, the artificial intelligence full strength is provided, the communication of related responsible persons is provided, and the sleeping safety of the students is finally ensured to the maximum extent.

Example 1:

the technician selects a bedroom of a specific school as an example, and in the example, the installation position and selection of the camera are determined according to the actual situation of the bedroom. According to technical staff's field observation, the bedroom contains 2 access & exits, selects high definition digtal camera to monitor and head portrait snatchs, and two cameras of business turn over need be installed to each export, have had data guarantee promptly after the installation is accomplished.

Technicians access the 4 paths of camera data to the GPU server cluster, deploy a face detection acquisition algorithm and a face recognition algorithm into the GPU server cluster, and guarantee the reliability of the image recognition algorithm. The GPU server firstly processes the accessed video, after the face is detected, the face is captured, collected and judged in quality, and the face with higher quality is selected by the same person and output to a face comparison algorithm.

Technicians cooperate with schools to collect student information, in the collection process, gaps between system data layers are opened based on school student management systems, face data and other data are correlated, and a data middle platform with student face id as a coupling point is constructed.

The face id of the face recognition and the face id of the student data are fused, the sleeping safety of each student is judged based on a sleeping safety intelligent algorithm, and the judgment result is output to a school management large screen and a small program in hands of related personnel.

After abnormal information is found by school related managers from the small program or the large screen, the school related managers need to go to the field or confirm the current state of the students in other modes according to actual conditions, complete or track the processes at hand based on the current state, and ensure the sleeping safety of the students.

With the above embodiments, all the bedroom managers can easily implement the invention. Any technical engineer may implement the process in our way.

Claims

1. An intelligent student safety management method based on a school student dormitory scene. Contain with bedroom, key crossing and two-way installation high definition intelligence camera in school gate, the camera contains conventional camera video stream and records the collection function and the face snatchs the function. And transmitting the video stream and the face capturing data of the corresponding equipment back to the GPU server, performing face capturing work on the video stream in the server, and performing cross verification with the face capturing of the front-end camera to ensure the high quality and integrity of the face capturing. In the GPU cluster, the system compares the pictures with picture base libraries of students, teachers and workers, and inputs comparison results into a sleeping safety management system. In the safety management system for sleeping, the system judges the safety condition of sleeping of each student based on the safety rules formulated by the school, and outputs the abnormity in combination with the safety strategies formulated by the school.

2. The method for acquiring the head portrait information of all students according to the claim 1, correlating the head portrait with other information of the students, and importing the head portrait into a GPU server cluster to be used as a basis for comparison of face data acquired by a subsequent camera. The acquisition end adopts the intelligent camera carrying the computing power, the intelligent camera can write in a proper algorithm by combining with the requirements of a sleeping scene, the edge computing of the equipment is utilized to reduce the computing power pressure of a central GPU cluster, meanwhile, according to the management requirements of a school, in order to further reduce the omission ratio of the face, the video stream can be accessed into a GPU server cluster, and the face and background pictures are acquired by the back end computing for the follow-up face comparison service.

S1, face collection can collect the front face and side face photos of the same student for modeling, so that the recognition accuracy at specific corners is guaranteed.

And S2, both edge calculation and center calculation can be used for detecting the face, judging the face quality and grabbing the face.

And S3, judging attributes of the students based on the equipment information and the face information and the position information.

3. The system of claim 1, wherein the face comparison result is output to a sleep management system, and the system performs a safety logic judgment in combination with related information.

4. The method of claim 1, wherein the calculation conclusion of the homed safety management system is output to various applications.