CN116050158A

CN116050158A - Fall behavior identification and judgment method

Info

Publication number: CN116050158A
Application number: CN202310086144.2A
Authority: CN
Inventors: 赵建光; 范晶晶; 刘岩石; 王振岩
Original assignee: Hebei University of Architecture
Current assignee: Hebei University of Architecture
Priority date: 2023-02-08
Filing date: 2023-02-08
Publication date: 2023-05-02

Abstract

The invention provides a method for identifying and judging falling behaviors, which belongs to the technical field of falling judgment and comprises the following steps: contour information of a person and a supporting surface within a certain range is collected, and a corresponding character model and a supporting surface model are generated in an upper computer through the contour information. And setting the strength parameter and the hardness parameter of the supporting surface model according to actual conditions. And determining the gravity center position of each character model and the magnitude and direction of the gravity center supporting force of the supporting surface model. And when the time interval of unbalance of the character model and the acting force of the character model when the character model collides with the supporting surface reach preset requirements, judging that falling occurs. According to the method for identifying and judging the falling behaviors, provided by the invention, the falling judgment conditions are optimized through the unbalanced time interval and the acting force during the fitted expansion, so that the method has a guiding significance, the erroneous judgment is avoided, and the accuracy is improved.

Description

Fall behavior identification and judgment method

Technical Field

The invention belongs to the technical field of fall judgment, and particularly relates to a fall behavior recognition and judgment method.

Background

In cases where global aging is exacerbated, falls are one of the leading health threats for elderly people. More and more old people live alone, cannot be found in time when accidents happen without looking at the old people, and accordingly great potential safety hazards exist in the life of the old people.

With the continuous development of various constructions such as safe cities and intelligent traffic in China, a method for integrating a machine vision technology into a video monitoring system has become a current hot research problem. At present, most of existing methods are used for identifying falling behaviors by using a traditional machine learning method, and the identification rate is low, so that the old cannot be cured in time. Therefore, how to efficiently, accurately and real-time detect the fall of the old is an urgent problem to be solved at present.

The current methods for detecting falling mainly comprise three types: the method based on the wearable equipment, the method based on the environment sensor and the method based on the computer vision exist excessive judgment or insufficient judgment, and even if people are unbalanced, the people cannot judge that the people fall when the people support the people below, so that the current judgment method is not strong in practicability and low in accuracy due to the reasons.

Disclosure of Invention

The invention aims to provide a method for identifying and judging falling behaviors, which aims to solve the problems of weak practicability and low accuracy in judging falling.

In order to achieve the above purpose, the invention adopts the following technical scheme: the method for identifying and judging the falling behaviors comprises the following steps:

acquiring contour information of a person and a supporting surface within a certain range, and generating a corresponding character model and a supporting surface model in an upper computer through the contour information;

setting the strength parameter and the hardness parameter of the supporting surface model according to actual conditions;

determining the gravity center position of each character model and the magnitude and direction of the gravity center supporting force of the supporting surface model;

and judging that the falling occurs when the time interval of unbalance of the character model and the acting force of the character model when the character model collides with the supporting surface reach preset requirements.

In one possible implementation, the acquiring contour information of the person and the supporting surface within a certain range includes:

scanning all people and supporting surfaces within a certain range through acquisition equipment; the support surface includes a ground surface and an outer peripheral surface of each object.

In one possible implementation manner, the generating, in the host computer, the corresponding character model and the supporting surface model through the contour information includes:

and enabling the character model and the supporting surface model in the upper computer to be synchronous with reality through the contour information transmitted back by the acquisition equipment in real time.

and determining the difference between the two profile information of the adjacent time periods, and adjusting the character model and the supporting surface model by the upper computer according to the difference.

In one possible implementation manner, the setting the strength parameter and the hardness parameter of the supporting surface model according to the actual situation includes:

and determining the hardness and strength information of the supporting surface, and setting the same and corresponding hardness parameters and strength parameters for the supporting surface model in the upper computer according to the hardness and strength information.

setting a reference object, and determining the proportional relation between the actual size of the reference object and the size of a model generated in the upper computer;

and determining the required actual size through the character model and the supporting surface model according to the proportional relation.

In one possible implementation manner, the determining the position of the center of gravity of each character model and the magnitude and direction of the supporting force of the supporting surface model on the center of gravity includes:

fitting the posture of the corresponding person according to the sex of the person and the type of the current clothes;

and determining the weight of the person according to the quality parameters of all the parts through the posture.

and according to the speed and the acceleration of the character model relative to the supporting surface model and combining the position of the gravity center, analyzing the magnitude and the direction of the supporting force of the supporting surface facing the gravity center.

In one possible implementation, when the time interval of unbalance of the character model and the acting force of the character model when the character model collides with the supporting surface reach preset requirements, determining that the fall occurs includes:

and determining the time interval from when unbalance of the character model occurs to when the character model is rebalanced in the upper computer.

simulating the acting force of the character model when expanding the supporting surface in the upper computer;

and setting different judging sections according to different ages, judging that the patient falls when the time interval is positioned in the judging section and the acting force is larger than the corresponding standard, otherwise, not carrying out early warning.

The method for identifying and judging the falling behaviors has the beneficial effects that: compared with the prior art, the method for identifying and judging the falling behavior firstly collects the contour information of the person and the supporting surface within a certain range, and can generate the corresponding character model and the supporting surface model in the upper computer through the contour information after the contour information is determined. In order to be more realistic, the strength parameters and hardness parameters of the supporting surface model need to be set according to practical situations. Since the character model and the supporting surface model correspond to reality, the position of the center of gravity of the character model and the magnitude and direction of the supporting force of the supporting surface model on the character model can be determined.

In the upper computer, when the unbalanced time interval of the character model and the acting force of the character model when the character model collides with the supporting surface reach preset requirements, the falling is judged. In the method, the corresponding model is generated according to the actual situation, and the falling judgment conditions are optimized through the unbalanced time interval and the acting force during the fitted expansion, so that the method has guiding significance, misjudgment is avoided, and accuracy is improved.

Drawings

In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the drawings that are needed in the embodiments or the description of the prior art will be briefly introduced below, and it is obvious that the drawings in the following description are only some embodiments of the present invention, and that other drawings can be obtained according to these drawings without inventive effort for a person skilled in the art.

Fig. 1 is a flowchart of a method for identifying and determining a falling behavior according to an embodiment of the present invention.

Detailed Description

In order to make the technical problems, technical schemes and beneficial effects to be solved more clear, the invention is further described in detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are for purposes of illustration only and are not intended to limit the scope of the invention.

Referring to fig. 1, a description will now be given of a method for identifying and determining a fall behavior according to the present invention. A fall behavior recognition and determination method, comprising:

contour information of a person and a supporting surface within a certain range is collected, and a corresponding character model and a supporting surface model are generated in an upper computer through the contour information.

And setting the strength parameter and the hardness parameter of the supporting surface model according to actual conditions.

And determining the gravity center position of each character model and the magnitude and direction of the gravity center supporting force of the supporting surface model.

And when the time interval of unbalance of the character model and the acting force of the character model when the character model collides with the supporting surface reach preset requirements, judging that falling occurs.

A fall refers to a sudden, involuntary, unintended change in position, falling to the ground or a lower plane. The death cause monitoring data of the national disease monitoring system in 2006 show that the fall mortality rate of old people over 65 years old in China is 49.56/10 ten thousand for men and 52.80/10 ten thousand for women. Falls are the fourth cause of injury and death in our country, the first in elderly people over 65 years old. Therefore, monitoring the occurrence of a fall in time to provide a rescue quickly, avoiding secondary injuries caused by falls is a relatively critical issue.

The existing fall detection technology mainly adopts a sensor monitoring system based on acceleration, gyroscopes and the like, but the fall detection technology has the technical problems of low accuracy and high false alarm rate.

At present, there are many studies on fall detection methods based on computer vision at home and abroad, and the fall detection methods can be specifically classified into the following categories according to the adopted algorithm and the implementation method:

the body shape analysis method extracts the human body outline from the image through a background elimination modeling algorithm, then uses the human body as an interested area to frame by a rectangle, and judges whether falling occurs or not by using the aspect ratio. The method is easily influenced by illumination change and background moving objects, has high misjudgment rate, and can not realize the function of falling pre-judgment.

According to the method, the falling and similar actions such as bending and squatting are distinguished according to the characteristic that the old cannot recover for a long time after falling, and the falling identification rate is improved by defining a specific area. It can be seen that the method is typical detection after falling, cannot realize the pre-judging function, and cannot correctly distinguish between lying and falling for a long time.

The head movement tracking method adopts a particle filtering method to track the head of a human body, and detects whether falling occurs or not through the distance from the head to the ground and the descending rate of the head. According to the method, a falling pre-judging mechanism is realized to a certain extent through stable tracking of the head of the human body, but the falling model library is lack of construction, so that the detection misjudgment rate is high. In addition, the method lacks a scene understanding algorithm, is low in robustness to complex environments, is time-consuming in particle filtering algorithm and low in detection rate, and therefore real-time detection is difficult to achieve.

According to the method, a plurality of behavior modes such as walking, squatting, sitting, lying, falling and the like are trained by adopting a convolutional neural network CNN to generate a falling model library, and then classification and identification are carried out on the falling model library, so that falling detection is realized. The method generates a model library of the user, and greatly improves the accuracy of falling detection. However, using the CNN training model is computationally intensive, resulting in lower algorithm efficiency and no predictive function for falls.

Falls are the primary cause of accidental injuries to elderly people, and it is counted that 60% of the accidental injuries to elderly people are head injuries, 90% of hip and wrist injuries are caused by falls, and 30% of solitary old people and 50% of old people in long-term care institutions fall at least once a year. Thus, timely detection of falls is important in the long-term care of elderly people. On the other hand, as the trend of world population aging becomes more and more serious, the cost of long-term care of the aged is also higher and higher, especially in some care institutions such as hospitals and nursing homes, and therefore, the demand for a real-time detection system for the fall of the aged is very large.

The current methods for detecting falling mainly comprise three types: wearable device-based methods, environmental sensor-based methods, and computer vision-based methods.

The method based on the wearable equipment detects the falling behavior of the human body by detecting the acceleration through the sensor, and has the advantages of small calculated amount and simple use, but needs to be worn at any time to influence normal life. The detection method based on the environment sensor, such as a pressure sensor and a sound sensor, has the advantages of small calculation amount, but the method is greatly influenced by the pressure change of the environment and the sound change, and has higher false alarm. The method based on computer vision mainly uses monitored video image information, does not need wearing equipment, is easily influenced by illumination, cannot be used at night, has poor privacy protection and cannot reach higher standards.

In some embodiments of a fall behavior recognition and determination method provided in the present application, collecting contour information of a person and a support surface within a certain range includes:

In order to identify the human body, real-time data acquisition is performed through optical acquisition equipment such as cameras at present, and a mature technology can accurately identify the outline of the human body at present, and according to facial features, information such as the identity of a target person can be determined.

Based on the technical foundation, the joint points of the human body can be determined through other methods such as image acquisition, sensor detection and the like, so that the current condition of relative movement among the joint points of the human body is determined.

However, it should be noted that, because of the difference between the weights of different persons, there may be a large difference in the body shapes of the same height, that is, different weight distribution, which results in a certain difference in the position of the center of gravity of each person, and because of the difference, the degree of change in the center of gravity is different when the fall occurs, so that a specific analysis is required for specific problems.

More importantly, the human body does not always fall under the standing condition, and may fall under other postures such as squat, and if a support exists under the human body, even if unbalance is supported in time by the support, the human body should not be judged to fall, so that the falling is judged to have a certain narrowness only by the movement modes of the joints.

Based on the above reasons, the body posture and the surrounding environment need to be analyzed, that is, the environmental factors of the current scene and the conditions of the supporting platforms of the supporters need to be determined, and when the body is at different positions of the current scene, the determination of the falling behavior needs to be correspondingly adjusted, so that whether the falling behavior occurs or not and the severity of the falling can be more accurately and directly determined.

In some embodiments of a fall behavior recognition and determination method provided in the present application, generating, in an upper computer, a corresponding character model and a supporting surface model through contour information includes:

the figure model and the supporting surface model in the upper computer can be synchronized with reality through the contour information transmitted back by the acquisition equipment in real time.

If a person falls on a softer support, even though it is determined by the current state of the art that a fall behavior has occurred, it is not determined in this application as a truly damaging fall because it is not damaging.

Since the final purpose of determining the fall behavior is to be able to early warn of possible injuries in advance, the method of determining the fall behavior is valuable if a fall is made and is damaging, if there is a softer support although the body is out of balance and the body is free to fall for a very short time, or if the imbalance of the centre of gravity by the body is spontaneous, i.e. the body is spontaneous to perform some action, the fall cannot be determined because the above-mentioned behavior is not damaging and has a certain purpose.

However, the existing methods simply analyze the structure of the human body, and do not mention the environment in which the human body is located and the relative positions of the human body and surrounding objects, and the like, so that the existing falling behavior judging methods are low in practicality and have very large misjudgment.

In the present application, it is necessary to recognize a human body by a camera or the like, and to recognize the volume and the relative position of each object within a photographed range. When a person enters the target range, a corresponding model can be generated in the upper computer according to the data acquired by the camera, and then the state of the person is judged according to the position and posture variation of the person.

In the application, the acquisition equipment such as a camera can scan all people and objects in a target range in real time, the acquisition equipment can transmit acquired data to the upper computer, then a model can be generated in the upper computer in real time, and the most visual and accurate judgment can be carried out on the falling behavior through the model.

But in general, there are at least two acquisition devices in order to be able to perform modeling more comprehensively and accurately, and it is required that a plurality of acquisition devices can shoot objects at different angles. For the above reasons, a high requirement is put on the processing capability of the upper computer.

Based on the reasons, after the acquisition equipment transmits data to the upper computer, the upper computer screens the data. The specific mode is that the data determined by the acquisition equipment are divided. Because the data acquired by the acquisition device includes a dynamic portion and a static portion, more people are in the dynamic portion and more objects are in the surrounding environment in the static portion.

Thus, when the data is divided, the outline of the static portion is first defined, and the static portion may be a region where the position reflected by the data does not change during a time interval in general. After the static part is determined, a plurality of points are picked up in the area and used for displaying the outline of the static object, if the plurality of position points at the same position in adjacent data determined by the acquisition equipment are not changed, modeling is not performed any more, namely, the model is replaced according to the previous model, namely, the upper computer only models the changed object, in the same time interval, according to the extraction of the track, outline and the like of the action of the object, the object with the shape which is not changed is not modeled any more, the model is moved correspondingly and the change angle is changed according to the outline, and finally the data amount required to be processed by the upper computer is reduced.

In some embodiments of a method for identifying and determining a falling behavior provided in the present application, setting strength parameters and hardness parameters of a supporting surface model according to actual conditions includes:

When the model is built, the falling behavior cannot be immediately judged, and therefore, the materials are different, and the harmfulness is different. If the support surface is soft, even a fall cannot be determined as a true fall, since there is no harm, for example, to fall on a spring bed. But if the support surface is relatively stiff, there is a difference in the situation.

Therefore, in practical application, related personnel are required to set corresponding parameters such as hardness, strength and the like in the model according to practical conditions, and corresponding quality parameters are also required to be set according to the size of the human body model.

When the parameters are set, the force of the models when the models collide with each other can be approximately calculated, so that the injury condition when the human body falls can be predicted.

setting a reference object, and determining the proportional relation between the actual size of the reference object and the size of a model generated in the upper computer.

And determining the required actual size through the character model and the supporting surface model according to the proportion relation.

In the application, modeling of a human body and objects in an environment needs to be performed in an upper computer, and in order to perform real-time scanning so as to finally generate a model, acquisition equipment in at least two positions is required to scan people and objects in a certain range from different angles.

In reality, however, the farther the acquisition device is from the object to be measured, the smaller the size of the model that is ultimately generated, and if the acquisition device is closer to the object, the larger the volume of the model that is ultimately generated.

In order to make the size of the generated model be the same as that of the actual model, a reference object is arranged in the range covered by the acquisition equipment, the reference object is scanned together in the scanning process of the acquisition equipment, and the actual size of the reference object is stored in the upper computer, so that the upper computer marks the generated model according to the actual size of the reference object, namely, the proportion between the unit model length and the actual length is determined in the upper computer, and the actual object size and the height of a person can be judged through the model according to the proportion.

In some embodiments of a fall behavior recognition and determination method provided in the present application, determining a position of a center of gravity of each character model and a magnitude and a direction of a supporting surface model to a center of gravity supporting force includes:

fitting the posture of the corresponding person according to the sex of the person and the type of the current clothes.

And determining the weight of the person through the posture according to the quality parameters of all the parts.

After the dynamic part and the static part are modeled by the acquisition device, the main object of the dynamic part is usually a person, so that in practical application, the time for unbalance of the center of gravity of the person needs to be determined, and the time from unbalance of the center of gravity of the person until the person contacts the support needs to be determined.

For the reasons mentioned above, the center of gravity of the person needs to be determined, the model of the person can be determined by the collecting device, and after the model of the person is determined, the volume of the model is the total volume of the person and the clothes, and because the thicknesses of the clothes worn by the person in different seasons are different, if the model is directly regarded as the human body, the effective estimation of the total weight cannot be performed, and the determination of the center of gravity position is deviated.

Based on the above reasons, the upper computer needs to primarily judge the dressing of the current person according to the current air temperature and weather conditions, and judge the sex of the target person through the acquisition equipment and the face recognition, primarily judge the body state of the person according to the information such as the sex on the basis of the model according to the information, and then determine the body state of the person according to the general model of the clothes and the established model after the body state is determined, and determine the weight of the person through the body state, and the gravity center condition of the person can be judged in the upper computer due to the fact that the model is determined.

and analyzing the magnitude and direction of the supporting force of the supporting surface facing the gravity center according to the speed and the acceleration of the character model relative to the supporting surface model and combining the position of the gravity center.

How to judge the unbalance of the gravity center of the human body is the key of the application, firstly, the body state of the human body is judged in an upper computer according to the model, the sex of the human body, the external environment and other conditions, then the gravity center of the human body is judged through the body state and the corresponding density of the human body at each part through the model.

The model in the upper computer is changed in real time, and the position of the center of gravity is also changed in real time. In order to conveniently judge whether the person is in an unbalanced state, the direction and the magnitude of the gravity force of the support surface are needed to be judged, and the relative positions of the human body and other objects are needed to be combined.

The center of gravity is inside the human body before there is no fall, and if there is no object around, the center of gravity is supported by legs or the like below this center of gravity. When unbalance occurs, there is no human body part for supporting below the center of gravity, and surrounding objects cannot provide supporting force, the supporting force of the supporting face to the person is smaller than the gravity, and unbalance is determined at this time.

For the above reasons, it is necessary to determine the direction of the supporting force of the supporting surface to the person by the upper computer based on the determined center of gravity and the supporting condition below the center of gravity, and for detailed description, if the supporting force of the supporting surface to the person is smaller than the gravity of the person, it is determined that the supporting force is unbalanced at this time.

In some embodiments of a method for identifying and determining a falling behavior provided in the present application, when a time interval of unbalance of a character model and an acting force of the character model when the character model collides with a supporting surface reach preset requirements, determining that a falling occurs includes:

in the host computer, a time interval is determined from when unbalance of the character model occurs to when the character model is rebalanced.

The model in the upper computer is synchronous with the actual situation, and when the user does not fall down, the user is supported by the legs. In the falling process, the supporting surface is in an unbalanced state because the supporting force of the supporting surface is smaller than the gravity or the supporting legs are not supported. When the human body falls on the ground, namely the supporting surface, the supporting surface provides supporting force, so that the human body can be in a balanced state.

Therefore, when in actual determination, the supporting condition between the center of gravity of the human body and the supporting surface needs to be analyzed in real time through the model, and as the model is synchronous with the actual condition and the material of the model is the same as the actual condition, the magnitude and the direction of acting force between the supporting surface and the human body can be determined in real time in the upper computer, so that whether the human body is in an unbalanced state can be determined.

in the host computer, the acting force when the character model expands the supporting surface is simulated.

In the present application, if the distance between the human body and the support surface is low, even if the human body falls down, the human body is not determined to fall because the injury is not high. However, for some old people, due to the decrease of the bone strength, a series of problems such as fracture may be caused by small falling, for the above reasons, basic information of people needs to be identified through the acquisition device, different judging sections need to be set up according to different age groups, each judging section contains the time for judging gravity unbalance when falling, the time for judging gravity unbalance can be analyzed through the acquisition device and the upper computer, and then, through the time, the age of people and the corresponding judging section are combined, so that whether falling occurs or not is analyzed.

In addition to the time determination of the unbalance, the final falling support surface needs to be analyzed and determined, because if the support surface is soft, the damage caused by even falling is small, and if the falling is determined, the final falling is definitely misdetermined.

For the above reasons, in the present application, the upper computer is used to calculate the acting force when the human body hits the supporting surface, and because the upper computer is provided with the model of the human body and the object and the corresponding material parameters, the upper computer can determine the acting force when the human body hits the supporting surface through the analysis of the software such as finite element, if the acting force is greater than a certain value and the unbalance time is in the corresponding determination interval through the software analysis, the falling is determined.

The foregoing description of the preferred embodiments of the invention is not intended to be limiting, but rather is intended to cover all modifications, equivalents, and alternatives falling within the spirit and principles of the invention.

Claims

1. A fall behavior recognition and determination method, characterized by comprising:

2. A fall behavior recognition and determination method as claimed in claim 1, wherein the acquiring profile information of the person and the support surface over a range comprises:

3. A method of fall behavior recognition and determination as claimed in claim 2, wherein the generating of the corresponding character model and supporting surface model from the profile information in the host computer comprises:

4. A fall behavior recognition and determination method as claimed in claim 3, wherein the generating of the corresponding character model and supporting surface model from the contour information in the host computer comprises:

5. A method of fall behavior recognition and determination as claimed in claim 3, wherein said setting the strength parameters and the hardness parameters of the support surface model according to the actual situation comprises:

6. A method of fall behavior recognition and determination as defined in claim 5, wherein generating the corresponding character model and supporting surface model from the contour information in the host computer comprises:

7. A fall behavior recognition and determination method according to claim 5, wherein the determining the position of the center of gravity of each of the character models and the magnitude and direction of the supporting force of the supporting surface model on the center of gravity comprises:

8. A fall behavior recognition and determination method according to claim 7, wherein the determining the position of the center of gravity of each of the character models and the magnitude and direction of the supporting force of the supporting surface model on the center of gravity comprises:

9. A method of fall behavior recognition and determination as claimed in claim 8, wherein determining that a fall has occurred when the time interval of unbalance of the character model and the force of the character model striking the support surface both meet predetermined requirements comprises:

10. A method of fall behavior recognition and determination as claimed in claim 9, wherein determining that a fall has occurred when the time interval of unbalance of the character model and the force of the character model striking the support surface both meet predetermined requirements comprises: