WO2022041484A1

WO2022041484A1 - Human body fall detection method, apparatus and device, and storage medium

Info

Publication number: WO2022041484A1
Application number: PCT/CN2020/127346
Authority: WO
Inventors: 夏钦展
Original assignee: 歌尔股份有限公司
Priority date: 2020-08-26
Filing date: 2020-11-07
Publication date: 2022-03-03
Also published as: CN111899470A; CN111899470B

Abstract

A human body fall detection method, apparatus and a device, and a storage medium, which belong to the technical field of safety protection monitoring. The method comprises: photographing a human body to be subjected to detection, so as to obtain a human body movement image set of said human body (S10); performing, on the basis of a pre-constructed fall detection model, motion state classification on each human body movement image in the human body movement image set, so as to obtain a motion state type corresponding to each human body movement image (S20); and performing fall detection on said human body according to the motion state types (S30). Motion state classification is performed, on a human body movement image set of a human body to be subjected to detection, by means of a pre-constructed fall detection model, and fall detection is performed on said human body according to a motion state type corresponding to each human body movement image, such that fall detection can be accurately performed on said human body according to a plurality of human body movement images of different motion state types, and the problem of low detection accuracy is solved, thereby improving the accuracy of human body fall detection.

Description

Human fall detection method, device, equipment and storage medium

This application claims the priority of the Chinese patent application with the application number 202010867487.9 and the invention title "Human Fall Detection Method, Apparatus, Equipment and Storage Medium" filed with the China Patent Office on August 26, 2020, the entire contents of which are incorporated by reference in this application.

technical field

The present invention relates to the technical field of security monitoring, and in particular, to a method, device, equipment and storage medium for detecting human body fall.

Background technique

With the development of economy and technology and the improvement of people's living standards, smart home technology has also developed by leaps and bounds. The demand for human health monitoring in the home environment is increasingly urgent, and elderly care is a key direction in the field of smart homes. Especially for the elderly living alone, falls are the main health threat in their indoor activities. Therefore, the research on fall detection has attracted more and more attention. The existing elderly care systems are based on multi-sensors and contour detection. These fall detection methods have simple algorithms, but the accuracy of the detection results is low. In practical applications The false positive rate is high.

The above content is only used to assist the understanding of the technical solutions of the present invention, and does not mean that the above content is the prior art.

SUMMARY OF THE INVENTION

The main purpose of the present invention is to provide a human body fall detection method, device, equipment and storage medium, aiming at solving the technical problems of low detection accuracy and high false alarm rate in the prior art.

In order to achieve the above object, the present invention provides a human body fall detection method, which comprises the following steps:

photographing the human body to be detected to obtain a set of human activity images of the human body to be detected;

Based on the pre-built fall detection model, classifying the motion state of each human body motion image in the human body motion image set to obtain the motion state type corresponding to each human body motion image;

Fall detection is performed on the human body to be detected according to the motion state type.

Optionally, the step of performing fall detection on the human body to be detected according to the motion state type includes:

Screening out a motion state type conforming to the first preset type from the motion state types;

acquiring the number of types corresponding to the motion state type that conforms to the first preset type;

Determine the fall confidence level corresponding to the human body to be detected according to the number of types;

Fall detection is performed on the human body to be detected according to the fall confidence level.

Optionally, the step of performing fall detection on the human body to be detected according to the fall confidence includes:

comparing the fall confidence level with a preset confidence level;

If the fall confidence level does not exceed the preset confidence level, it is determined that the human body to be detected does not have a risk of falling;

If the fall confidence level exceeds the preset confidence level, it is determined that the human body to be detected has a risk of falling.

Optionally, after the step of determining that the human body to be detected has a risk of falling if the fall confidence level exceeds the preset confidence level, the method further includes:

When the human body to be detected has a risk of falling, acquiring a set of images of the falling tendency of the human body to be detected within a preset time;

Based on the pre-built fall detection model, classifying the fall tendency state of each fall tendency image in the fall tendency image set to obtain a fall tendency state type corresponding to each fall tendency image;

Selecting a fall prone state type conforming to the second preset type from the fall prone state types;

obtaining the number of types corresponding to the falling tendency state type conforming to the second preset type;

When the number of the types reaches the number threshold, it is determined that a fall event occurs on the human body to be detected, and a fall warning prompt is output.

Optionally, before the step of classifying the motion state of each human body motion image in the human body motion image set based on a pre-built fall detection model, to obtain the motion state type corresponding to each human body motion image, the method further includes:

Obtain the sample fall video of the human body to be detected, and the sample fall time corresponding to the sample fall video;

generating a sample fall image queue according to the sample fall time and the sample fall video;

processing each fall image in the sample fall image queue to obtain a target sample fall image queue;

The target sample fall image queue is trained to obtain a fall detection model.

Optionally, the step of generating a sample fall image queue according to the sample fall time and the sample fall video includes:

According to the time sequence corresponding to the sample fall time, the sample fall video recording is divided into several fall video clips in turn;

Capture the start frame image of each fall video clip and the end frame image of the last fall video clip;

A sample fall image queue is generated according to the start frame image and the end frame image of each fall video clip.

Optionally, the step of processing each fall image in the sample fall image queue to obtain the target sample fall image queue includes:

Perform denoising processing on each fall image in the sample fall image queue according to the human body motion feature to obtain a plurality of human fall images;

classifying a plurality of the human body fall images in sequence according to the time sequence corresponding to the fall time of the sample;

Corresponding category identifiers are set for each classified human fall image to obtain the fall image of the target sample.

In addition, in order to achieve the above object, the present invention also provides a human body fall detection device, the device includes:

an acquisition module, used for photographing the human body to be detected, so as to obtain a set of human activity images of the human body to be detected;

a classification module, used for classifying the motion state of each human body motion image in the human body motion image set based on a pre-built fall detection model, so as to obtain the motion state type corresponding to each human body motion image;

A judgment module, configured to perform fall detection on the human body to be detected according to the motion state type.

In addition, in order to achieve the above object, the present invention also proposes a human body fall detection device, which includes: a memory, a processor, and a human body fall detection device stored in the memory and running on the processor. A program, the human fall detection program is configured to implement the steps of the human fall detection method as described above.

In addition, in order to achieve the above object, the present invention also provides a storage medium, on which a human body fall detection program is stored, and when the human body fall detection program is executed by a processor, the above-mentioned human body fall detection method is realized. step.

In the present invention, the human body to be detected is photographed to obtain a set of human body motion images of the human body to be detected; based on a pre-built fall detection model, the motion state of each human body motion image in the set of human body motion images is classified to obtain each human body motion image set. The motion state type corresponding to the moving image; fall detection is performed on the human body to be detected according to the motion state type, and the motion state is classified according to the human body moving image set of the human body to be detected through a pre-built fall detection model, and the corresponding human body moving images are classified according to the motion state. The fall detection of the human body to be detected can be carried out according to the motion state type of the human body to be detected, and the fall detection of the human body can be accurately performed according to the human body activity images of a plurality of different motion state types, thereby improving the accuracy of the human body fall detection.

Description of drawings

1 is a schematic structural diagram of a human body fall detection device in a hardware operating environment involved in an embodiment of the present invention;

2 is a schematic flowchart of the first embodiment of the human body fall detection method of the present invention;

3 is a schematic flowchart of a second embodiment of a human body fall detection method according to the present invention;

4 is a schematic flowchart of a third embodiment of a human body fall detection method according to the present invention;

5 is a schematic diagram of a human body fall photographing method of the present invention;

FIG. 6 is a structural block diagram of the first embodiment of the human body fall detection apparatus according to the present invention.

The realization, functional characteristics and advantages of the present invention will be further described with reference to the accompanying drawings in conjunction with the embodiments.

detailed description

It should be understood that the specific embodiments described herein are only used to explain the present invention, but not to limit the present invention.

Referring to FIG. 1 , FIG. 1 is a schematic structural diagram of a human body fall detection device in a hardware operating environment involved in an embodiment of the present invention.

As shown in FIG. 1 , the human body fall detection device may include: a processor 1001 , such as a central processing unit (Central Processing Unit, CPU), a communication bus 1002 , a user interface 1003 , a network interface 1004 , and a memory 1005 . Among them, the communication bus 1002 is used to realize the connection and communication between these components. The user interface 1003 may include a display screen (Display), an input unit such as a keyboard (Keyboard), and the optional user interface 1003 may also include a standard wired interface and a wireless interface. Optionally, the network interface 1004 may include a standard wired interface and a wireless interface (such as a wireless fidelity (WIreless-FIdelity, WI-FI) interface). The memory 1005 may be a high-speed random access memory (Random Access Memory, RAM) memory, or may be a stable non-volatile memory (Non-Volatile Memory, NVM), such as a disk memory. Optionally, the memory 1005 may also be a storage device independent of the aforementioned processor 1001 .

Those skilled in the art can understand that the structure shown in FIG. 1 does not constitute a limitation on the human body fall detection device, and may include more or less components than the one shown, or combine some components, or arrange different components.

As shown in FIG. 1 , the memory 1005 as a storage medium may include an operating system, a network communication module, a user interface module and a human body fall detection program.

In the human body fall detection device shown in FIG. 1, the network interface 1004 is mainly used for data communication with the network server; the user interface 1003 is mainly used for data interaction with the user; the processor 1001, the memory in the human body fall detection device of the present invention 1005 may be set in a human body fall detection device, the human body fall detection device invokes the human body fall detection program stored in the memory 1005 through the processor 1001, and executes the human body fall detection method provided by the embodiment of the present invention.

An embodiment of the present invention provides a human body fall detection method. Referring to FIG. 2 , FIG. 2 is a schematic flowchart of a first embodiment of a human body fall detection method of the present invention.

In this embodiment, the human body fall detection method includes the following steps:

Step S10 : photographing the human body to be detected to obtain a set of human motion images of the human body to be detected.

It should be noted that the execution body of this embodiment may be an image data processing device, which is used for photographing an object to obtain a moving image of the object, and analyzing and processing the moving image of the object, and may also be an image data processing device with a captured image. The terminal device with the data and image data processing function is not limited in this embodiment.

In this embodiment, the human body to be detected is photographed by a camera. The camera includes an analog camera, a digital camera, a high-definition camera, a charge coupled device (CCD) camera, and a spherical camera. In this embodiment, the transmission signal of the camera is , resolution, sensor model and shape are not limited, you can use the appropriate camera according to the actual situation. In addition, when the camera collects the human activity image of the human body to be detected, it can collect at a speed of 24 frames per second, and can also collect at other speeds.

It should be noted that when the camera is shooting the human body to be detected, the image collected is actually the regional image of the moving area where the human body to be detected is located, and there may be other moving objects in the regional image. Obtained, in this embodiment, the area image is denoised according to the human body motion characteristics, moving objects that do not conform to the human body motion characteristics are removed from the area image, and the remaining area image that only includes the human body is the human body activity image, so A set of human activity images of the human body to be detected can be obtained.

Step S20: Based on the pre-built fall detection model, classify the motion state of each human body motion image in the human body motion image set, so as to obtain the motion state type corresponding to each human body motion image.

It should be noted that each human activity image in the human activity image set contains the motion state of the human body to be detected, and different motion states have corresponding motion state types, which essentially reflect the corresponding differences in the process of human falling. Movement state, by dividing the whole process of the human body from standing to falling into several different stages, and defining the corresponding movement state type for the movement state of the human body in each stage, for example, dividing the entire falling process of the human body into standing state, dumping state A , dumping state B, dumping state C, and falling state, define the motion state type corresponding to standing state as 1, the motion state type corresponding to dumping state A as 2, the motion state type corresponding to dumping state B as 3, and the corresponding motion state type of dumping state C. The motion state type is 4, the motion state type corresponding to the fall state is 5, and the motion state type corresponding to the motion state in this embodiment can be set according to the actual situation and is not limited in this embodiment.

In the specific implementation, the motion state classification of each human body activity image is performed by the fall detection model, and the corresponding motion state type of each human body activity image is obtained, and the fall detection model is pre-built based on the sample fall images of the human body.

Step S30: Perform fall detection on the human body to be detected according to the motion state type.

It is easy to understand that after obtaining the motion state type corresponding to each human body activity image, fall detection can be performed on the human body to be detected according to the motion state type corresponding to each human body activity image. For example, if the motion state type corresponding to the standing state is 1, it can be The motion state type 1 determines that the human body to be detected has not fallen, and assuming that the motion state type corresponding to the motion state is 2, it can be determined that the human body to be detected has a fall event according to the motion state type 2.

In this embodiment, the human body to be detected is photographed to obtain the human body motion image set of the to-be-detected human body; based on a pre-built fall detection model, the motion state of each human body motion image in the human body motion image set is classified to obtain each The motion state type corresponding to the human body activity image; fall detection is performed on the human body to be detected according to the motion state type, and the motion state is classified according to the human body activity image set of the human body to be detected through a pre-built fall detection model, and each human body activity image is classified according to the motion state. The fall detection of the human body to be detected corresponding to the motion state type can accurately perform the fall detection on the human body according to the human body activity images of a plurality of different motion state types, thereby improving the accuracy of the fall detection of the human body.

Referring to FIG. 3 , FIG. 3 is a schematic flowchart of a second embodiment of a human body fall detection method according to the present invention.

Based on the above-mentioned first embodiment, in this embodiment, the step S30 includes:

Step S301: Screening out the motion state type conforming to the first preset type from the motion state types.

In this embodiment, to perform fall detection on the human body to be detected according to the type of motion state, it is necessary to first screen out the motion state type that conforms to the first preset type from the motion state types, and the first preset type is that a fall event has not yet occurred but there is a fall. The type of motion state of the risk, for example, the entire fall process of the human body is divided into standing state, dumping state A, dumping state B, dumping state C, and falling state, define the motion state type corresponding to the standing state as 1, and the motion state corresponding to the dumping state A. The type is 2, the motion state type corresponding to the tipping state B is 3, the motion state type corresponding to the tipping state C is 4, and the motion state type corresponding to the falling state is 5. The first preset type can be set to 1, 2 or 1. , 2, and 3, etc.

Step S302: Acquire the number of types corresponding to the motion state types conforming to the first preset type.

In a specific implementation, after filtering out the motion state types that conform to the first preset type, the corresponding number of types needs to be obtained, and the number of types represents the number of different motion state types. For example, it is assumed that the first preset types are 1, 2, and 3. , based on the pre-built fall detection model, the motion state type corresponding to the human activity image X is 1, the motion state type corresponding to the human activity image Y is 2, and the motion state type corresponding to the human activity image Z is 3. Let the number of types corresponding to the type of motion state type be 3, if the motion state type corresponding to the human body motion image Z is 4, then the number of types corresponding to the motion state type conforming to the first preset type is 2, if the motion state type corresponding to the human body motion image Y is 2. The motion state type of , and the motion state type corresponding to the human body motion image Z are both 2, and the number of types corresponding to the motion state type conforming to the first preset type is 2.

Step S303: Determine the fall confidence level corresponding to the human body to be detected according to the number of types.

It should be noted that a motion state that conforms to the first preset type indicates that a fall of the human body to be detected may be about to occur. The greater the probability of the human body to be detected falling down is represented by the fall confidence level in this embodiment, and the greater the fall confidence level, the greater the possibility that the corresponding human body to be detected is about to fall. The obtained number of types of motion state types can determine the fall confidence level corresponding to the human body to be detected. For example, when the number of types obtained is 1, it can be determined that the corresponding fall confidence level of the human body to be detected is 50%. 2, it can be determined that the fall confidence level corresponding to the human body to be detected is 85%, and the corresponding relationship between the number of types and the fall confidence level can be set according to the actual situation, which is not limited in this embodiment.

Step S304: Perform fall detection on the human body to be detected according to the fall confidence level.

In a specific implementation, the fall state of the human body to be detected can be determined according to the fall confidence degree, thereby realizing the fall detection of the human body to be detected. In this embodiment, the steps of performing fall detection on the human body to be detected according to the fall confidence degree include: Compare the fall confidence level with a preset confidence level; if the fall confidence level does not exceed the preset confidence level, it is determined that the human body to be detected does not have a risk of falling; If the reliability is preset, it is determined that the human body to be detected has a risk of falling.

It should be noted that the greater the fall confidence level, the greater the possibility that the human body to be detected is about to fall. The preset reliability indicates that it can be determined that the human body to be detected is about to fall. In this embodiment, the fall confidence level and the preset reliability are determined. For comparison, if the fall confidence level does not exceed the preset confidence level, it means that the possibility of the human body to be detected is about to fall is low, and it can be determined that the human body to be detected does not have a risk of falling. It is very likely that the detected human body is about to fall, and it can be determined that the human body to be detected has a risk of falling.

Further, after it is determined that the human body to be detected has a risk of falling, the human body to be detected continues to be detected, and it can be determined whether a fall event occurs in the human body to be detected. In this implementation, after the step S304, the method further includes: When there is a risk of falling, acquiring a set of fall prone images of the human body to be detected within a preset time period; based on the pre-built fall detection model, perform a fall prone state on each fall prone image in the fall prone image set classifying to obtain the falling tendency state type corresponding to each falling tendency image; screening the falling tendency state type conforming to the second preset type from the falling tendency state types; obtaining the falling tendency state type conforming to the second preset type The number of types corresponding to the type of falling tendency state type; when the number of types reaches the number threshold, it is determined that a fall event occurs on the human body to be detected, and a fall alarm prompt is output.

It should be noted that, from the time when it is determined that the human body to be detected has a risk of falling, a set of falling and tipping potential images of the human body to be detected within a preset time is obtained, and the preset time is the standard falling time in the fall detection model. After obtaining the falling tipping potential image After the collection, it is still necessary to classify each fall prone image in the fall prone image set. The classification process is similar to the process of classifying the motion state of each human activity image in the human activity image set, and it is also based on the pre-built fall detection model. , to obtain the fall tendency state type corresponding to each fall tendency image, and then screen out the fall tendency state type that conforms to the second preset type. 2. The greater the number of preset types of fall prone state types, the greater the possibility of a fall event of the human body to be detected. Therefore, it can be determined whether the human body to be detected has fallen according to the number of types of fall prone state types. event.

It should be noted that the relationship between the second preset type and the first preset type is that the second preset type includes the first preset type, for example, the first preset type is 1, 2, 3, 4, and 5, and the second preset type is Let the types be 1, 2, 3, 4, 5, ..., 20, where the first preset type is used to determine whether the human body to be detected has a risk of falling, and the second preset type is used to determine whether the human body to be detected has a risk of falling. Afterwards, it is determined whether the human body to be detected finally has a fall event. In this embodiment, when the number of types corresponding to the type of falling tendency type reaches the number threshold, it is determined that the human body to be detected has a fall event, and a fall alarm prompt is output. For example, define the first preset type as 1, 2, 3, 4 and 5, the second preset type as 1, 2, 3, 4, 5, ..., 20, the preset reliability is 85% and the quantity threshold is 10. Assuming that the motion state types corresponding to the human body activity images of the human body A to be detected are 1 and 3, and the falling confidence level corresponding to the human body A to be detected is obtained as 90%, it can be determined that the human body A to be detected has a risk of falling, and then continue to obtain the preset The fall prone image set of the human body A to be detected in the time period, the classification results obtained by classifying each fall prone image in the fall prone image set are fall prone state types 1, 3, 6, 7, 8, 10, 12, 14, 16, 17, and 18, it can be obtained that the number of types corresponding to the type of falling tendency state type is 11, and it can be determined that a fall event occurs in the human body A to be detected.

In this embodiment, the motion state types conforming to the first preset type are selected from the motion state types; the number of types corresponding to the motion state types conforming to the first preset type is obtained; the Fall confidence level corresponding to the human body to be detected; fall detection is performed on the human body to be detected according to the fall confidence level, and fall detection is performed according to the number of types corresponding to the motion state type of the human body motion image and the fall confidence level corresponding to the number of types, and at the same time When the human body to be detected has a risk of falling, determine the type of falling tendency state of the falling tendency image based on the second preset type, and further determine whether a fall event occurs. Fall detection is more accurate.

Referring to FIG. 4 , FIG. 4 is a schematic flowchart of a third embodiment of a human body fall detection method according to the present invention.

Based on the above-mentioned first or second embodiment, a third embodiment of a human body fall detection method according to the present invention is proposed.

Taking the above-mentioned first embodiment as an example for description, in this embodiment, the step S20 further includes:

Step S201: Acquire a sample fall video of the human body to be detected, and a sample fall time corresponding to the sample fall video.

It is easy to understand that, before classifying the human activity image or the fall tendency image based on the fall detection model, the fall detection model needs to be constructed in advance.

In this embodiment, the sample fall video of the human body to be detected and the sample fall time corresponding to the sample fall video are obtained. The fall video includes the entire fall process of the human body from the start of the fall until the fall does not move, and the sample fall time is the time when the human body begins to fall. The total duration between the moment of last fall and immobility.

It should also be noted that the standard fall time in the fall detection model is T, T=L _e ±σ, where _Le is the average of all fall times, σ is the standard deviation of the fall time, and σ can be defined according to actual needs. , which is not limited in this embodiment, L _e =(L ₁ +L ₂ +L ₃ +...+_L _n )/n, where L ₁ , L ₂ , L ₃ ,..., L _n represent each fall recording Corresponding sample fall time, n is the number of fall recordings, for example, as shown in Figure 5, N people prepare M fall methods to shoot through 8 cameras, and 8*M*N can be obtained, L ₁ , L ₂ , L ₃ , ..., L _n represents the falling time of each fall recording sample in 8*M*N, and 8*M*N is the number of falling recordings.

Step S202: Generate a sample fall image queue according to the sample fall time and the sample fall video.

In a specific implementation, the fall videos are divided in the chronological order of the sample fall times, and the fall images in each fall video are intercepted to generate a sample fall image queue. In this embodiment, step S202 includes: falling according to the samples The time sequence corresponding to the time divides the sample fall video into several fall video clips in turn; intercepts the start frame image of each fall video clip and the end frame image of the last fall video clip; The start frame image and the end frame image generate samples that fall into the image queue.

It should be noted that, according to the time sequence corresponding to the time of the sample falling (the sequence from the beginning of the fall to the final fall and not moving), the sample falling video is divided into several falling video segments in turn, for example, the sample falling video V is divided into V ₁ , V ₂ . and V3 _three fall video clips, after obtaining the fall video clip, intercept the start frame image of each fall video clip and the end frame image of the last fall video clip, for example, divide the sample fall video V into V ₁ , V After the _three falling video clips ₂ and _V3 , intercept the starting frame images P1, _P2 and P3 _of V1, _V2 and _V3 , and the end frame image _P3 _' of the last falling recording _V3 . The initial frame images P ₁ , P ₂ , P ₃ and the end frame image P ₃ ′ can construct a sample image queue. In addition, it should be noted that the interception of the start frame and end image of the fall video clip is essentially to decompose the entire fall process of the human body, and the human body state at different stages during the fall process can be obtained. The human body state includes human activity state and fall. Tendency state.

Step S203: Process each fall image in the sample fall image queue to obtain a target sample fall image queue.

In the specific implementation, it is necessary to perform denoising and classification processing on each sample fall image in the sample fall image queue to obtain the target sample fall image queue. In this embodiment, step S203 includes: according to human motion characteristics, the sample fall image Perform denoising processing on each fall image in the queue to obtain multiple human fall images; classify the multiple human fall images in turn according to the time sequence corresponding to the fall time of the samples; set corresponding human fall images for each classified human fall image. The class identities to obtain the target sample fall into the image queue.

It should be noted that there may be other moving objects in the fall image, which may affect the fall detection. Therefore, it is necessary to denoise each fall image according to the human motion characteristics to obtain a human fall image containing only the human body. Then, according to the time sequence of the fall time of the samples, the fall images of the human body are classified in turn. Each fall image of the human body corresponds to a state of the human body, and a corresponding category identifier is set for each classified image of the fall of the human body. For example, set the category identifier of the human fall image X. is 1, and the category of the human fall image Y is identified as 2, and the multiple human fall images after denoising and classification can constitute the target sample fall image queue.

Step S204: Train the target sample fall image queue to obtain a fall detection model.

In this embodiment, training the target sample fall image queue can input the target sample fall image queue after denoising and classification into a preset neural network model, and a fall detection model can be obtained. The preset neural network model includes BP Neural network, Hopfield network, ART network and Kohonen network etc.

In this embodiment, a sample fall video of the human body to be detected and the sample fall time corresponding to the sample fall video are obtained; a sample fall image queue is generated according to the sample fall time and the sample fall video; Each fall image in the image queue is processed to obtain a target sample fall image queue; the target sample fall image queue is trained to obtain a fall detection model, and the human body fall image is obtained by denoising the sample fall image queue, and Classify the human fall images to obtain the target sample fall image queue, and train the fall detection model based on the target sample image queue, which makes the pre-built fall detection model more accurate and further improves the accuracy of human fall detection.

In addition, an embodiment of the present invention also provides a storage medium, where a human body fall detection program is stored, and the human body fall detection program is executed by a processor to implement the steps of the above-mentioned method for human fall detection.

Referring to FIG. 6 , FIG. 6 is a structural block diagram of the first embodiment of the human body fall detection apparatus according to the present invention.

As shown in FIG. 6 , the human body fall detection device proposed by the embodiment of the present invention includes:

The acquisition module 10 is used for photographing the human body to be detected, so as to obtain a set of human activity images of the human body to be detected.

In this embodiment, the human body fall detection device can collect moving images of the object and analyze and process the moving images of the object. The human body fall detection device is provided with a camera, and the human body to be detected is photographed through the camera. The camera includes an analog camera and a digital camera. , high-definition camera, charge-coupled device CCD camera, spherical camera, etc. In this embodiment, the transmission signal, resolution, sensor model and shape of the camera are not limited, and an appropriate camera can be used according to the actual situation. In addition, when the camera collects the human activity image of the human body to be detected, it can collect at a speed of 24 frames per second, and can also collect at other speeds.

The classification module 20 is configured to classify the motion state of each human body motion image in the human body motion image set based on the pre-built fall detection model, so as to obtain the motion state type corresponding to each human body motion image.

In the specific implementation, the motion state classification of each human body activity image is performed by the fall detection model, and the corresponding motion state type of each human body activity image is obtained, and the fall detection model is pre-built based on the sample image of the human body to be detected.

The judgment module 30 is configured to perform fall detection on the human body to be detected according to the motion state type.

It is easy to understand that after obtaining the motion state type corresponding to each human body activity image, fall detection can be performed on the human body to be detected according to the motion state type corresponding to each human body activity image. For example, if the motion state type corresponding to the standing state is 1, it can be The motion state type 1 determines that the human body to be detected has not fallen, and assuming that the motion state type corresponding to the falling state is 2, then the human body to be detected can be determined to have a fall event according to the motion state type 2.

In one embodiment, the judging module 30 is further configured to screen out a motion state type conforming to the first preset type from the motion state types; obtain the corresponding motion state type conforming to the first preset type. The number of types; the fall confidence level corresponding to the human body to be detected is determined according to the type number; the fall detection is performed on the human body to be detected according to the fall confidence level.

In one embodiment, the judging module 30 is further configured to compare the fall confidence level with a preset confidence level; if the fall confidence level does not exceed the preset confidence level, determine the to-be-detected The human body does not have a fall risk; if the fall confidence level exceeds the preset confidence level, it is determined that the human body to be detected has a fall risk.

In one embodiment, the judging module 30 is further configured to, when the human body to be detected has a risk of falling, obtain a set of images of the falling tendency of the human body to be detected within a preset time; A detection model for classifying the falling tendency states of each falling tendency image in the set of falling tendency images, so as to obtain a falling tendency state type corresponding to each falling tendency image; Falling tendency state types of the second preset type; obtaining the number of types corresponding to the falling tendency state types conforming to the second preset type; when the number of types reaches the number threshold, it is determined that the human body to be detected has fallen event, and output a fall warning prompt.

In one embodiment, the human body fall detection device further includes: a building block;

The building module is used to obtain the sample fall video of the human body to be detected, and the sample fall time corresponding to the sample fall video; generate a sample fall image queue according to the sample fall time and the sample fall video; Process each fall image in the sample fall image queue to obtain a target sample fall image queue; train the target sample fall image queue to obtain a fall detection model.

The building module is further configured to sequentially divide the sample fall video into several fall video clips according to the time sequence corresponding to the sample fall time; intercept the start frame image of each fall video clip and the last fall video clip. An end frame image; a sample fall image queue is generated according to the start frame image and the end frame image of each fall video clip.

The building module is further configured to perform denoising processing on each fall image in the sample fall image queue according to the human body motion feature, so as to obtain a plurality of human fall images; The fall images of the human body are classified in sequence; corresponding category identifiers are set for each classified fall image of the human body, so as to obtain the fall images of the target sample.

It should be understood that the above are only examples, and do not constitute any limitation to the technical solutions of the present invention. In specific applications, those skilled in the art can make settings as required, which is not limited by the present invention.

It should be noted that the above-described workflow is only illustrative, and does not limit the protection scope of the present invention. In practical applications, those skilled in the art can select some or all of them to implement according to actual needs. The purpose of the solution in this embodiment is not limited here.

In addition, for technical details not described in detail in this embodiment, reference may be made to the human body fall detection method provided by any embodiment of the present invention, and details are not repeated here.

Furthermore, it should be noted that, herein, the terms "comprising", "comprising" or any other variation thereof are intended to encompass non-exclusive inclusion, such that a process, method, article or system comprising a series of elements includes not only those elements, but also other elements not expressly listed or inherent to such a process, method, article or system. Without further limitation, an element qualified by the phrase "comprising a..." does not preclude the presence of additional identical elements in the process, method, article or system that includes the element.

The above-mentioned serial numbers of the embodiments of the present invention are only for description, and do not represent the advantages or disadvantages of the embodiments.

From the description of the above embodiments, those skilled in the art can clearly understand that the methods of the above embodiments can be implemented by means of software plus a necessary general hardware platform, and of course hardware can also be used, but in many cases the former is better implementation. Based on this understanding, the technical solutions of the present invention can be embodied in the form of software products in essence or the parts that contribute to the prior art, and the computer software products are stored in a storage medium (such as a read-only memory (Read Only Memory). , ROM)/RAM, magnetic disk, optical disk), including several instructions to make a terminal device (which may be a mobile phone, a computer, a server, or a network device, etc.) execute the methods described in the various embodiments of the present invention.

The above are only preferred embodiments of the present invention, and are not intended to limit the scope of the present invention. Any equivalent structure or equivalent process transformation made by using the contents of the description and drawings of the present invention, or directly or indirectly applied in other related technical fields , are similarly included in the scope of patent protection of the present invention.

Claims

A human body fall detection method, characterized in that the human body fall detection method comprises:

photographing the human body to be detected to obtain a set of human activity images of the human body to be detected;

Based on the pre-built fall detection model, classifying the motion state of each human body motion image in the human body motion image set to obtain the motion state type corresponding to each human body motion image;

Fall detection is performed on the human body to be detected according to the motion state type.
The human body fall detection method according to claim 1, wherein the step of performing fall detection on the human body to be detected according to the motion state type comprises:

Screening out a motion state type conforming to the first preset type from the motion state types;

acquiring the number of types corresponding to the motion state type that conforms to the first preset type;

Determine the fall confidence level corresponding to the human body to be detected according to the number of types;

Fall detection is performed on the human body to be detected according to the fall confidence level.
The human body fall detection method according to claim 2, wherein the step of performing fall detection on the human body to be detected according to the fall confidence level comprises:

comparing the fall confidence level with a preset confidence level;

If the fall confidence level does not exceed the preset confidence level, it is determined that the human body to be detected does not have a risk of falling;

If the fall confidence level exceeds the preset confidence level, it is determined that the human body to be detected has a risk of falling.
The human body fall detection method according to claim 3, wherein, after the step of determining that the human body to be detected has a risk of falling if the fall confidence level exceeds the preset confidence level, the method further comprises:

When the human body to be detected has a risk of falling, acquiring a set of images of the falling tendency of the human body to be detected within a preset time;

Based on the pre-built fall detection model, classify the fall tendency state on each fall tendency image in the fall tendency image set, so as to obtain the fall tendency state type corresponding to each fall tendency image;

Selecting a fall prone state type conforming to the second preset type from the fall prone state types;

obtaining the number of types corresponding to the falling tendency state type conforming to the second preset type;

When the number of the types reaches the number threshold, it is determined that a fall event occurs on the human body to be detected, and a fall warning prompt is output.
The human body fall detection method according to any one of claims 1 to 4, characterized in that, based on a pre-built fall detection model, the motion state classification of each human body movement image in the human body movement image set is performed to obtain Before the step of the motion state type corresponding to each human body activity image, the step further includes:

Obtain the sample fall video of the human body to be detected, and the sample fall time corresponding to the sample fall video;

generating a sample fall image queue according to the sample fall time and the sample fall video;

processing each fall image in the sample fall image queue to obtain a target sample fall image queue;

The target sample fall image queue is trained to obtain a fall detection model.
The human body fall detection method according to claim 5, wherein the step of generating a sample fall image queue according to the sample fall time and the sample fall video comprises:

According to the time sequence corresponding to the sample fall time, the sample fall video recording is divided into several fall video clips in turn;

Capture the start frame image of each fall video clip and the end frame image of the last fall video clip;

A sample fall image queue is generated according to the start frame image and the end frame image of each fall video clip.
The human body fall detection method according to claim 5, wherein the step of processing each fall image in the sample fall image queue to obtain the target sample fall image queue comprises:

Perform denoising processing on each fall image in the sample fall image queue according to the human body motion feature to obtain a plurality of human fall images;

classifying a plurality of the human body fall images in sequence according to the time sequence corresponding to the fall time of the sample;

Corresponding category identifiers are set for each classified human fall image to obtain the fall image of the target sample.
A human body fall detection device, characterized in that the human body fall detection device comprises:

an acquisition module, used for photographing the human body to be detected, so as to obtain a set of human activity images of the human body to be detected;

a classification module, configured to classify the motion state of each human body motion image in the human body motion image set based on a pre-built fall detection model, so as to obtain the motion state type corresponding to each human body motion image;

A judgment module, configured to perform fall detection on the human body to be detected according to the motion state type.
A human body fall detection device, characterized in that the human body fall detection device comprises: a memory, a processor, and a human body fall detection program stored in the memory and running on the processor, the human body fall detection program Steps configured to implement the method for human fall detection according to any one of claims 1 to 7.
A storage medium, characterized in that a human body fall detection program is stored on the storage medium, and when the human body fall detection program is executed by a processor, the human body fall detection method according to any one of claims 1 to 7 is implemented. step.