CN116189232A - Machine vision-based method and system for detecting abnormal behaviors of aged and elderly in nursing homes - Google Patents

Machine vision-based method and system for detecting abnormal behaviors of aged and elderly in nursing homes Download PDF

Info

Publication number
CN116189232A
CN116189232A CN202211573138.1A CN202211573138A CN116189232A CN 116189232 A CN116189232 A CN 116189232A CN 202211573138 A CN202211573138 A CN 202211573138A CN 116189232 A CN116189232 A CN 116189232A
Authority
CN
China
Prior art keywords
recognition model
target object
machine vision
behavior recognition
detecting abnormal
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
CN202211573138.1A
Other languages
Chinese (zh)
Inventor
杨明强
陈言
吴斌
杨奇
任静
赵浩博
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.)
Shandong University
Original Assignee
Shandong University
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 Shandong University filed Critical Shandong University
Priority to CN202211573138.1A priority Critical patent/CN116189232A/en
Publication of CN116189232A publication Critical patent/CN116189232A/en
Pending legal-status Critical Current

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
    • 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
    • G06V10/00—Arrangements for image or video recognition or understanding
    • G06V10/40—Extraction of image or video features
    • G06V10/44—Local feature extraction by analysis of parts of the pattern, e.g. by detecting edges, contours, loops, corners, strokes or intersections; Connectivity analysis, e.g. of connected components
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
    • G06V10/00—Arrangements for image or video recognition or understanding
    • G06V10/70—Arrangements for image or video recognition or understanding using pattern recognition or machine learning
    • G06V10/764—Arrangements for image or video recognition or understanding using pattern recognition or machine learning using classification, e.g. of video objects
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
    • G06V10/00—Arrangements for image or video recognition or understanding
    • G06V10/70—Arrangements for image or video recognition or understanding using pattern recognition or machine learning
    • G06V10/77—Processing image or video features in feature spaces; using data integration or data reduction, e.g. principal component analysis [PCA] or independent component analysis [ICA] or self-organising maps [SOM]; Blind source separation
    • G06V10/80—Fusion, i.e. combining data from various sources at the sensor level, preprocessing level, feature extraction level or classification level
    • G06V10/806—Fusion, i.e. combining data from various sources at the sensor level, preprocessing level, feature extraction level or classification level of extracted features
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
    • G06V10/00—Arrangements for image or video recognition or understanding
    • G06V10/70—Arrangements for image or video recognition or understanding using pattern recognition or machine learning
    • G06V10/82—Arrangements for image or video recognition or understanding using pattern recognition or machine learning using neural networks

Landscapes

  • Engineering & Computer Science (AREA)
  • Theoretical Computer Science (AREA)
  • Physics & Mathematics (AREA)
  • General Physics & Mathematics (AREA)
  • Evolutionary Computation (AREA)
  • Multimedia (AREA)
  • Computer Vision & Pattern Recognition (AREA)
  • General Health & Medical Sciences (AREA)
  • Software Systems (AREA)
  • Computing Systems (AREA)
  • Health & Medical Sciences (AREA)
  • Artificial Intelligence (AREA)
  • Databases & Information Systems (AREA)
  • Medical Informatics (AREA)
  • Biomedical Technology (AREA)
  • Life Sciences & Earth Sciences (AREA)
  • Human Computer Interaction (AREA)
  • Biophysics (AREA)
  • Computational Linguistics (AREA)
  • Data Mining & Analysis (AREA)
  • Molecular Biology (AREA)
  • General Engineering & Computer Science (AREA)
  • Mathematical Physics (AREA)
  • Alarm Systems (AREA)
  • Emergency Alarm Devices (AREA)

Abstract

The invention provides a machine vision-based method and a system for detecting abnormal behaviors of old people in a nursing home, comprising the following steps: acquiring video data shot by a camera; detecting abnormal behaviors of video data containing target objects based on different abnormal behavior recognition models respectively; the abnormal behavior recognition model comprises an active hand lifting help recognition model, a falling behavior recognition model and a long standing behavior recognition model. The detection of three abnormal behaviors solves the problem that whether the old in the nursing home contains abnormal behaviors or not is inaccurate due to the fact that the existing single detection behaviors.

Description

Machine vision-based method and system for detecting abnormal behaviors of aged and elderly in nursing homes
Technical Field
The invention belongs to the field of image processing, and particularly relates to a method and a system for detecting abnormal behaviors of old people in a nursing home based on machine vision.
Background
The statements in this section merely provide background information related to the present disclosure and may not necessarily constitute prior art.
With the high-speed improvement of the modern economic level of China, the life quality of people realizes breakthrough crossing, the aspects of scientific and technological production, medical treatment and health and the like are improved to a great extent, the average life of the old population of China is obviously prolonged, and the ageing degree of the population structure is gradually increased. However, many children are busy and do not have time to care for the elderly, and institutional care is also a popular way of care. The deterioration of the physical functions of the elderly and the like is affected by various chronic diseases, so that the safety and nursing problems of the elderly are more important. The research design of the method for detecting the abnormal behaviors of the aged and the college players has important practical significance, and has great social significance for guaranteeing the safety of daily behavior activities of the aged, reducing the possibility of accidental risks, and constructing a safe and intelligent aged service environment.
Currently, the number of service practitioners and the technical expertise in some existing care institutions are not yet meeting the social requirements. When the abnormal behavior of the senior citizen is detected by the traditional modes of observing and monitoring video data by human eyes, on-site watching by on-site caregivers and the like, the problems of large working intensity, untimely detection, insufficient objectivity in judgment, easy occurrence of missed detection, false detection and the like of the caregivers exist, and the working efficiency is low. In addition, there are also some abnormal behavior detection methods based on the wearable device and the environmental sensor, the methods are not accurate enough to detect, and the problems of high equipment cost, short battery endurance time, limited detection range, easy environmental interference and the like exist, and the problems of inconvenience in wearing equipment, poor experience and the like exist for the elderly with inconvenient actions, and some elderly also forget to wear or are unwilling to wear the equipment and the like, so that the detection effect is often influenced. With the rapid development of computer vision and artificial intelligence, an abnormal behavior detection technology based on machine learning gradually enters the field of vision of people, and the method has the advantages of avoiding various inconveniences of wearing equipment, improving the detection accuracy and solving the problem more efficiently.
In the prior art, the detection of abnormal behaviors of a senior citizen based on machine vision still has some technical problems, on one hand, the lack of corresponding related data sets aiming at specific scenes such as the senior citizen leads to insufficient data, so that the training and testing effects are poor, and the detection accuracy is insufficient and the comprehensiveness is not high; secondly, the traditional monitoring equipment cannot perform monitoring analysis and detection alarm simultaneously, and the existing behavior detection systems are single, and only one or a class of behaviors can be identified; moreover, the detection effect of small targets in some pictures is poor, and the detection effect of abnormal behaviors of the aged and the elderly is still to be improved. Therefore, how to detect the behavior state of the elderly in real time and efficiently and feed back the behavior state to the relevant caretaker in time is an urgent problem to be solved at present.
Disclosure of Invention
In order to overcome the defects of the prior art, the invention provides a machine vision-based method and a machine vision-based system for detecting abnormal behaviors of old people in a nursing home, which are used for detecting and identifying three abnormal behaviors of lifting hands of the old people in a public area of the nursing home, such as active help seeking, falling and standing for a long time based on different detection models, and solves the problem that whether the old people in the nursing home contain abnormal behaviors or not due to the existing single detection behavior.
To achieve the above object, one or more embodiments of the present invention provide the following technical solutions: a machine vision-based method for detecting abnormal behaviors of aged people in a nursing home comprises the following steps:
acquiring video data shot by a camera;
detecting abnormal behaviors of video data containing target objects based on different abnormal behavior recognition models respectively; the abnormal behavior recognition model comprises an active hand lifting help recognition model, a falling behavior recognition model and a long standing behavior recognition model.
A second aspect of the present invention provides a machine vision-based system for detecting abnormal behavior of aged people in a nursing home, comprising:
the acquisition module is used for: acquiring video data shot by a camera;
and a detection module: after processing the video data containing the target object, detecting abnormal behaviors in different abnormal behavior recognition models; the abnormal behavior recognition model comprises an active hand lifting help recognition model, a falling behavior recognition model and a long standing behavior recognition model.
A third aspect of the invention provides a computer readable storage medium storing computer instructions which, when executed by a processor, perform the steps of the method described above.
A fourth aspect of the invention provides an electronic device comprising a memory and a processor and computer instructions stored on the memory and running on the processor, which when executed by the processor, perform the steps of the method described above.
The one or more of the above technical solutions have the following beneficial effects:
1. according to the invention, three abnormal behaviors of lifting hands of the old in the public area of the nursing home, such as active help seeking, falling and standing for a long time, are detected and identified based on different detection models, so that the problem that whether the old in the nursing home contains abnormal behaviors or not due to the existing single detection behavior is solved.
2. According to the invention, abnormal behaviors of the old are detected based on the YOLOv5 algorithm, a CBAM attention mechanism module is added to the algorithm, the accuracy of small target detection is improved, and the overall detection accuracy is higher.
3. According to the invention, by means of a method combining target tracking and template matching, the long-standing motionless behavior is detected, the influence of factors such as illumination and the like is small, the recognition accuracy is higher than that of the conventional recognition equipment, and the method has strong timeliness and robustness.
Additional aspects of the invention will be set forth in part in the description which follows and, in part, will be obvious from the description, or may be learned by practice of the invention.
Drawings
The accompanying drawings, which are included to provide a further understanding of the invention and are incorporated in and constitute a part of this specification, illustrate embodiments of the invention and together with the description serve to explain the invention.
FIG. 1 is a schematic diagram of a network of an active help-seeking identification model in accordance with a first embodiment of the present invention;
fig. 2 is a schematic diagram of a fall detection flow in accordance with the first embodiment of the present invention;
FIG. 3 is a flowchart of a long standing still detection in accordance with a first embodiment of the present invention;
FIG. 4 is a flowchart of the DeepSort algorithm in accordance with the first embodiment of the present invention;
FIG. 5 is a diagram of an abnormal behavior detection system according to a second embodiment of the present invention.
Detailed Description
It should be noted that the following detailed description is exemplary and is intended to provide further explanation of the invention. Unless defined otherwise, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this invention belongs.
It is noted that the terminology used herein is for the purpose of describing particular embodiments only and is not intended to be limiting of exemplary embodiments according to the present invention.
Embodiments of the invention and features of the embodiments may be combined with each other without conflict.
Example 1
The embodiment discloses a machine vision-based method for detecting abnormal behaviors of old people in a nursing home, which comprises the following steps:
acquiring video data shot by a camera;
detecting abnormal behaviors of video data containing target objects based on different abnormal behavior recognition models respectively; the abnormal behavior recognition model comprises an active hand lifting help recognition model, a falling behavior recognition model and a long standing behavior recognition model.
In this embodiment, in terms of data set preparation and production, firstly, pedestrians in monitoring video data of a pension institution are subjected to frame-separating interception operation and stored as pictures, and in daily life, as few lifting hands are required for active help, in order to obtain data of single person and multiple persons lifting hands, in the embodiment, partial pictures are collected in the pictures disclosed on the internet and pictures shot in a scene of a simulated pension, and 1384 pictures are formed together with the intercepted video frames, including pictures of single person and multiple persons lifting hands and not lifting hands. In this embodiment, the VOC data set format is used for training, so that the data set needs to be calibrated before training, the LabelImg software is used for pulling out the characters in the training set pictures from the rectangular frame and labeling, the hand is marked as "handup" instead of "nonhandup", so that the image marking is completed and the xml format file is generated, finally, the processed data set is placed in the folder, all the pictures in the data set are randomly divided into two parts, and the two parts are combined according to (training set+verification set): test set = 9:1, training set: validation set = 9:1, data set divided, no picture using data enhancement pre-processing. When training, the network model reads the xml file content and continuously learns, and finally the characteristics are extracted to realize prediction.
In this embodiment, an active help-seeking recognition model is used to detect the abnormal behavior of active help-seeking, the active help-seeking recognition model is a target detection algorithm based on YOLOv5, and YOLOv5 is a high-performance and general target detection model.
The YOLOv5 network can be divided into three parts, namely a backbond, an FPN and a yolhead, wherein the backbond refers to a trunk feature extraction network, and the YOLOv5 uses a CSPDarknet residual network; the FPN feature pyramid is a reinforced feature extraction network of YOLOv 5; yolhead is a classifier and regressor used to obtain the prediction results.
Obtaining a video frame image, inputting the video frame image into a backstone for feature extraction, and outputting three effective feature layers with different sizes; the FPN performs feature fusion on three effective feature layers obtained from the trunk part, not only performs feature fusion by upsampling, but also further feature fusion by downsampling, and outputs the three reinforced effective feature layers. And judging the feature points by the yolhead, and judging whether the object is contained or not, namely, predicting whether the feature points correspond to the object or not.
In this embodiment, a convolutional attention mechanism module (CBAM) is combined with the YOLOv5 network, the core of the attention mechanism is to focus on the places where the network needs to focus on, the attention mechanism is a way to realize self-adaptive attention of the network, and the attention mechanism is a plug-and-play module, which can be theoretically placed behind any feature layer, but the pre-training weight of the network cannot be used due to the placement on the backbone. Therefore, in this embodiment, the convolution attention mechanism module is applied to the enhanced feature extraction network, i.e., the FPN feature pyramid, the three effective feature layers extracted by the Backbone network, i.e., the Backbone, are input to the convolution attention mechanism module, and then the output of the convolution attention mechanism module is input to the strong feature extraction network, i.e., the FPN feature pyramid, and at the same time, the up-sampled result and the down-sampled result in the FPN feature pyramid are also input to the convolution attention mechanism module for feature enhancement.
CBAM is a lightweight convolution attention module, comprising two sub-modules of a Channel Attention Mechanism (CAM) and a spatial attention mechanism (CBM), which enhances useful features in a feature map by fusing attention features in the channel and spatial dimensions and suppresses useless features, thereby improving the network feature extraction capability, and achieving better effects than SENet focusing on the attention mechanism of the channel only.
The operation process of the CBAM is totally divided into two parts, a feature layer extracted from a Backbone network, namely a Backbone is used as input, global maximum pooling and global average pooling are carried out on the input according to channels, two one-dimensional vectors after pooling are sent into a shared full-connection layer for processing and then added, then normalized channel attention weight is obtained through a Sigmoid activation function, and the weight is multiplied with the input feature layer to obtain a feature layer after channel attention adjustment; and secondly, carrying out global maximum pooling and global average pooling on the feature layer after channel attention adjustment according to space, splicing two-dimensional vectors generated by pooling, carrying out convolution adjustment on the channel number which is 1 once, and multiplying the space attention weight subjected to normalization processing of the Sigmoid activation function by the input feature layer to obtain the finally generated feature layer.
And (3) training the active hand-lifting distress identification model, running a train. Py file to train a homemade data set based on the training set and the verification set, and performing parallel operation by using a GPU to obtain a YOLOV5-CBAM detection model, and evaluating the YOLOV5-CBAM detection model based on the test set to obtain a detection model with qualified evaluation. The network is set up for a total of 200 iterations. The iteration number of the freezing stage is 50 times, and the iteration number of the thawing stage is 150 times.
The training curve of the network is shown in fig. 1, and the training set loss is 0.08 and the verification set loss is 0.019 when the network is epochs=194. And (3) putting the generated optimal weight file into a prediction file for prediction, and inputting a path of a picture or video to obtain a prediction result. The weight obtained by the 194 th iteration is put into a prediction code, single-person detection and multi-person detection are respectively carried out, and a rectangular frame where the human body is located, the category to which the action belongs and the probability thereof are predicted and generated.
The active hand lifting help-seeking recognition model adopted by the embodiment can accurately detect two types of limb behaviors of lifting hands and not lifting hands, the predicted probability is about 0.8, multi-person target detection can be carried out, the functions of accurately detecting and judging human body actions in images are realized, positioning is accurate and the detection effect is good under the condition of no shielding, when people overlap or shielding, the detection probability is slightly reduced, and the algorithm performance is good as a whole.
As shown in fig. 2, in this embodiment, a fall behavior recognition model is used for detecting an abnormal fall behavior, and the fall behavior recognition model uses the YOLOV5-CBAM detection model to detect a target object, and the shape of an external rectangular frame of a human body is obviously different from that of an external rectangular frame of the human body in a falling state, so that the aspect ratio of the external rectangular frame of the human body detection can be used for judging to determine whether the detection target falls. Setting a first preset threshold for the aspect ratio of the detected person, wherein the first preset threshold in the embodiment is 1.4, if the first preset threshold is larger than the first preset threshold, the person is judged to fall, and if the first preset threshold is smaller than the first preset threshold, the person is judged to fall.
But it is considered that sometimes the aspect ratio of the human detection external frame is less than 1.4 may be a normal action such as sleeping or the like. The falling process is usually very violent, and the time from standing to falling is much shorter than the time of lying down during sleeping, so the embodiment adopts the judgment of the time difference of two status frames in the video to judge whether the old falls, and specifically:
the label is labeled "hall down" when the detected person aspect ratio is in the (1.4,1.9) interval, and "hall down" when the detected person aspect ratio is below 1.4. Then, a timer is added when each frame is processed, and the embodiment takes the falling time of 1.2s as a second preset threshold value, and judges that the user falls when the time interval between the labels of "no handup" and "fall down" is smaller than 1.2 s.
By judging the time difference of the two video states by the threshold value, normal life actions such as sleeping and the like can be eliminated, and the detection of falling is more accurate.
In this embodiment, as shown in fig. 3, the abnormal behavior of long standing still is detected by using a long standing still behavior recognition model, and based on the detection result of the YOLOV5-CBAM detection model, the existing target is tracked by combining with a deep target tracking model, a new target is created, and a target which is not matched for a long time is deleted.
The deep Sort algorithm carries out motion estimation by utilizing a Kalman filter and carries out front and back frame association by adopting a Hungary algorithm, depth association measurement is added on the basis to distinguish different pedestrians, and appearance information is added to solve the problem of target tracking when long-time shielding exists. The whole algorithm structure of the deep sort algorithm is shown in fig. 4, the result of the detection based on the YOLOV5-CBAM detection model is used as the input of the deep sort algorithm, the prior state prediction is carried out on the next frame target by utilizing a kalman prediction formula according to the current frame target state, then cascade matching and IOU matching are carried out, if the matching with the existing target is successful, the tracking operation is continued, if the unmatched detection target exists, the track is initialized by utilizing the kalman filtering and a new tracking object is created for the unmatched detection target, and if the unmatched tracking object exists for a long time, the tracking object is deleted.
After obtaining the detection frame of the target object according to the YOLOV5-CBAM and deep pedestrian tracking system, selecting a point with a larger mean square error area on the human body contour as a human body characteristic point. In order to ensure that the feature points are selected to the outline of the human body as far as possible, the length and the width of the target detection frame are respectively three quarters, a rectangular frame close to the center of the human body is selected, then the rectangular frame is equally divided into four parts, the mean square error of each part is calculated, and the center point of a small block with the maximum mean square error is selected as the feature point, so that four feature points of the pedestrian can be obtained.
Because the camera angle is fixed, the problems of rotation, scaling, visual angle change and the like are not involved, so that the template matching algorithm can be applied to match the positions of the characteristic points of the current frame and the same characteristic points in the subsequent frames. Selecting the size of a template according to the size of a detection frame, selecting a square taking a human body characteristic point as a center and taking one sixth of the width of the detection frame as a side length as a template image, then taking the square as a template, and matching the human body characteristic point of a subsequent video frame with the template to obtain the position of the human body characteristic point in a new video frame.
In this embodiment, after a pedestrian is detected to stay at a certain place for a period of time, the characteristic points are taken, and then the four characteristic point positions (x 1 ,y 1 )、(x 2 ,y 2 )、(x 3 ,y 3 )、(x 4 ,y 4 ) The four points are then combined into an eight-dimensional vector (x 1 ,y 1 ,x 2 ,y 2 ,x 3 ,y 3 ,x 4 ,y 4 ) As the feature vector of the video frame, calculating the Euclidean distance of the feature vector of the video frame within a preset time threshold, judging that the pedestrian does not lie for a long time if the Euclidean distance of the feature vector does not exceed the threshold, and sending out an alarm; if the threshold value is exceeded, the alarm is released, the feature vector information is cleared, the feature point information is cleared, and the acquisition is resumed. The method is not very sensitive to the change of the illumination condition, and therefore has higher robustness.
In this embodiment, further comprising: and alarming and prompting after detecting abnormal behaviors of different models.
In this embodiment, the rectification detection process for detecting abnormal behavior of the target object is:
step 1: the video stream input of the cameras is used as a data input source, firstly, a plurality of cameras in a monitoring network are accessed through a real-time streaming protocol (RTSP), a plurality of processes are used for processing a plurality of video streams in parallel, each process is responsible for processing one path of RTSP video stream, and the latest video frame is read and is transmitted to a lower target detection network, namely a YOLOV5-CBAM model after being preprocessed;
step 2: detecting whether a target object in a video has a hand lifting action or not through a YOLOV5-CBAM model, wherein the hand lifting action is divided into two situations of 'hand lifting' and 'non-hand lifting';
step 3: if the model in the step 2 detects the 'hand lifting' action, the time threshold value of the target frame is judged, if the duration of the hand lifting action is longer than the time threshold value, the state of 'hand lifting' is judged, the sound alarm and the prompt of the interface popup window are triggered, and otherwise, the state of 'non-hand lifting' is judged;
step 4: and (3) calculating the height-width ratio of all the 'non-lifting hand' target frames in the step (3), further judging by combining the time difference of the two state frames when the height-width ratio is smaller than the set threshold, judging as a falling state when the time interval is smaller than the set threshold, and triggering sound alarm and prompt of an interface popup window.
Step 5: and (3) carrying out target tracking on all the non-lifting target frames in the step (3) by combining with a Deepsort target tracking model, then selecting a template, carrying out characteristic point matching on the subsequent video frames and the template, thereby judging whether the state is in a 'long standing still' state, and if so, sending an audible alarm and prompting an interface popup window.
Example two
As shown in fig. 5, an object of the present embodiment is to provide a system for detecting abnormal behaviors of aged people in a nursing home based on machine vision, including:
the acquisition module is used for: acquiring video data shot by a camera;
and a detection module: after processing the video data containing the target object, detecting abnormal behaviors in different abnormal behavior recognition models; the abnormal behavior recognition model comprises an active hand lifting help recognition model, a falling behavior recognition model and a long standing behavior recognition model.
In this embodiment, the detection module is transplanted to the Django framework system, and when abnormal behavior is detected, the system prompts the staff to go to the scene where the abnormal behavior occurs to view the specific situation in a mode of sound alarm and interface popup. The hardware equipment adopted in the acquisition module is a sea-Kangwei video camera and a server, and video data of the camera are acquired in real time through a real-time streaming protocol (RTSP).
The system also comprises a result display module, and a Web system is built based on the django framework, so that the functions of data transmission and interface display are realized.
The system provided by the embodiment is simpler, a specific AI camera is not required to be installed, most of the existing monitoring systems can be compatible, and the cost is effectively reduced; the public area of the pension institution is subjected to video monitoring through the multi-path camera videos, so that the nursing auxiliary function of daily video inspection is realized, the behavior state of the old people in the public area can be detected in real time, and whether the old people have abnormal behaviors can be accurately judged in real time; caretakers can check the behavior state of the aged people at any time through the browser, intelligent analysis is performed on the behaviors of the people by using a video analysis technology, when abnormal behaviors occur, the system can automatically give an alarm to remind workers, the workload of manual care is greatly reduced, intelligent care service is realized, the aged care institutions are effectively helped to save personnel cost, and the working efficiency is improved.
Example III
It is an object of the present embodiment to provide a computing device comprising a memory, a processor and a computer program stored on the memory and executable on the processor, which processor implements the steps of the method described above when executing the program.
Example IV
An object of the present embodiment is to provide a computer-readable storage medium.
A computer readable storage medium having stored thereon a computer program which when executed by a processor performs the steps of the above method.
The steps involved in the devices of the second, third and fourth embodiments correspond to those of the first embodiment of the method, and the detailed description of the embodiments can be found in the related description section of the first embodiment. The term "computer-readable storage medium" should be taken to include a single medium or multiple media including one or more sets of instructions; it should also be understood to include any medium capable of storing, encoding or carrying a set of instructions for execution by a processor and that cause the processor to perform any one of the methods of the present invention.
It will be appreciated by those skilled in the art that the modules or steps of the invention described above may be implemented in a general purpose computer device, alternatively they may be implemented as program code executable by a computing device, such that they may be stored in a memory device for execution by the computing device, or they may be separately fabricated into individual integrated circuit modules, or a plurality of modules or steps thereof may be fabricated into a single integrated circuit module. The present invention is not limited to any specific combination of hardware and software.
While the foregoing description of the embodiments of the present invention has been presented in conjunction with the drawings, it should be understood that it is not intended to limit the scope of the invention, but rather, it is intended to cover all modifications or variations within the scope of the invention as defined by the claims of the present invention.

Claims (10)

1. The machine vision-based method for detecting abnormal behaviors of old people in nursing homes is characterized by comprising the following steps of:
acquiring video data shot by a camera;
detecting abnormal behaviors of video data containing target objects based on different abnormal behavior recognition models respectively; the abnormal behavior recognition model comprises an active hand lifting help recognition model, a falling behavior recognition model and a long standing behavior recognition model.
2. The machine vision-based method for detecting abnormal behaviors of aged people in a nursing home as claimed in claim 1, wherein a Backbone network structure of the active help-seeking identification model is YOLOv5, the YOLOv5 comprises three parts including a backbox, an FPN and a yolhead, a convolution attention mechanism module is added between the backbox part and the FPN part, and an up-sampling result and a down-sampling result are respectively input into the convolution attention mechanism module for feature addition in the FPN; the convolution attention mechanism module comprises a channel attention mechanism module and a spatial attention mechanism module.
3. The machine vision-based method for detecting abnormal behaviors of aged people in a nursing home as claimed in claim 1, wherein the backbone network of the falling behavior recognition model is a YOLOv5 network, an circumscribed rectangular frame containing a target object is obtained based on the YOLOv5 network, whether the aspect ratio of the circumscribed rectangular frame of the target object exceeds a first preset threshold value is judged, if the aspect ratio of the circumscribed rectangular frame of the target object is smaller than the first preset threshold value, whether the time difference of a front state video frame and a rear state video frame containing the target object exceeds a preset second preset threshold value is judged, and if the time difference of the front state video frame and the rear state video frame containing the target object does not exceed the second preset threshold value, the target object is falling behavior.
4. The machine vision-based method for detecting abnormal behavior of aged people in a nursing home as set forth in claim 1, wherein the long-standing stationary behavior recognition model comprises a YOLOv 5-convolution attention mechanism module and a deep target tracking model.
5. The machine vision-based method for detecting abnormal behaviors of aged people in a nursing home according to claim 1 is characterized in that in the long-standing stationary behavior recognition model, four feature point positions of each frame image containing a target object are adopted, eight-dimensional vectors are formed by the four feature point positions and serve as feature vectors of video frames, euclidean distances of the feature vectors of the video frames in a preset time threshold are calculated, whether the Euclidean distances obtained through calculation exceed a third preset threshold is judged, and if the Euclidean distances do not exceed the third preset threshold, the target object is stationary behavior for a long time.
6. The machine vision-based detection method for abnormal behaviors of aged people in a nursing home according to claim 5, wherein a detection frame of a target object is obtained, the length and width of the detection frame of the target object are three quarters, a rectangular frame close to the center of a human body is selected, the rectangular frame is divided into four parts uniformly, the mean square error of each part is calculated, and the center point of a small block with the maximum mean square error is selected as a characteristic point to obtain four characteristic points of the target object.
7. The machine vision-based method for detecting abnormal behaviors of aged people in a nursing home according to claim 1, further comprising the step of giving an alarm prompt when the detection result shows that the target object has abnormal behaviors.
8. Machine vision-based system for detecting abnormal behaviors of aged people in nursing homes is characterized by comprising:
the acquisition module is used for: acquiring video data shot by a camera;
and a detection module: after processing the video data containing the target object, detecting abnormal behaviors in different abnormal behavior recognition models; the abnormal behavior recognition model comprises an active hand lifting help recognition model, a falling behavior recognition model and a long standing behavior recognition model.
9. A computer-readable storage medium, on which a computer program is stored, characterized in that the program, when being executed by a processor, implements the steps of a machine vision-based method for detecting abnormal behavior of elderly people in a nursing home as claimed in any one of claims 1 to 7.
10. A processing device comprising a memory, a processor and a computer program stored on the memory and executable on the processor, characterized in that the processor implements the steps of a machine vision based method for detecting abnormal behaviour of elderly people in a nursing home as defined in any one of claims 1-7 when said program is executed.
CN202211573138.1A 2022-12-08 2022-12-08 Machine vision-based method and system for detecting abnormal behaviors of aged and elderly in nursing homes Pending CN116189232A (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN202211573138.1A CN116189232A (en) 2022-12-08 2022-12-08 Machine vision-based method and system for detecting abnormal behaviors of aged and elderly in nursing homes

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN202211573138.1A CN116189232A (en) 2022-12-08 2022-12-08 Machine vision-based method and system for detecting abnormal behaviors of aged and elderly in nursing homes

Publications (1)

Publication Number Publication Date
CN116189232A true CN116189232A (en) 2023-05-30

Family

ID=86439181

Family Applications (1)

Application Number Title Priority Date Filing Date
CN202211573138.1A Pending CN116189232A (en) 2022-12-08 2022-12-08 Machine vision-based method and system for detecting abnormal behaviors of aged and elderly in nursing homes

Country Status (1)

Country Link
CN (1) CN116189232A (en)

Cited By (1)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN118039134A (en) * 2024-04-09 2024-05-14 达州市中心医院(达州市人民医院) Medical information data enhancement method and system based on big data

Cited By (2)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN118039134A (en) * 2024-04-09 2024-05-14 达州市中心医院(达州市人民医院) Medical information data enhancement method and system based on big data
CN118039134B (en) * 2024-04-09 2024-06-04 达州市中心医院(达州市人民医院) Medical information data enhancement method and system based on big data

Similar Documents

Publication Publication Date Title
US11179064B2 (en) Method and system for privacy-preserving fall detection
WO2018188453A1 (en) Method for determining human face area, storage medium, and computer device
CN109271884A (en) Face character recognition methods, device, terminal device and storage medium
Yu et al. An online one class support vector machine-based person-specific fall detection system for monitoring an elderly individual in a room environment
CN111191486B (en) Drowning behavior recognition method, monitoring camera and monitoring system
WO2021139471A1 (en) Health status test method and device, and computer storage medium
CN111507248B (en) Face forehead region detection and positioning method and system based on low-resolution thermodynamic diagram
CN110503081B (en) Violent behavior detection method, system, equipment and medium based on interframe difference
US20200285842A1 (en) Method and apparatus for child state analysis, vehicle, electronic device, and storage medium
CN112926541B (en) Sleeping post detection method and device and related equipment
Shoaib et al. View-invariant fall detection for elderly in real home environment
CN111767823A (en) Sleeping post detection method, device, system and storage medium
CN113378649A (en) Identity, position and action recognition method, system, electronic equipment and storage medium
CN116189232A (en) Machine vision-based method and system for detecting abnormal behaviors of aged and elderly in nursing homes
Tao et al. 3D convolutional neural network for home monitoring using low resolution thermal-sensor array
CN115482485A (en) Video processing method and device, computer equipment and readable storage medium
CN115205581A (en) Fishing detection method, fishing detection device and computer readable storage medium
Wang et al. Detection of early dangerous state in deep water of indoor swimming pool based on surveillance video
CN111382723A (en) Method, device and system for identifying help
CN102867214B (en) Counting management method for people within area range
Liu et al. A video drowning detection device based on underwater computer vision
US11138856B1 (en) Intelligent infant sleep position monitor to avoid SIDS
CN113076799A (en) Drowning identification alarm method, drowning identification alarm device, drowning identification alarm platform, drowning identification alarm system and drowning identification alarm system storage medium
Hartmann et al. Camera-based system for tracking and position estimation of humans
CN114246767B (en) Blind person intelligent navigation glasses system and device based on cloud computing

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