Indoor person fall detection method, device, equipment and storage medium

Abstract

The present invention relates to the field of signal processing technology and discloses a method, apparatus, device, and storage medium for detecting indoor falls. The method comprises: collecting temperature distribution data of a target person using an infrared array sensor deployed in a target indoor space; extracting the target person's transverse morphological features and longitudinal distance features based on the temperature distribution data, with the transverse morphological features used to characterize spatial variations in human posture, and the longitudinal distance features used to analyze the dynamic evolution of movement trends; and inputting the fused feature set into a classification model to output a fall status determination result. The present invention utilizes infrared sensing technology to achieve contactless data collection, combining transverse and longitudinal feature fusion with a classification algorithm to effectively improve the accuracy of fall detection.

Description

Technical Field

[0001] The present invention relates to the field of signal processing technology, and in particular to a method, device, equipment and storage medium for detecting indoor person falls. Background Art

[0002] Indoor fall detection is becoming increasingly important for life safety. Existing fall detection solutions are primarily categorized as wearable and non-wearable. Wearable devices rely on sensors worn on the body, which can be inconvenient and result in high false alarm rates. Non-wearable solutions typically employ methods based on image or sensor data analysis, but these methods struggle to achieve high-precision, real-time monitoring due to environmental interference and complex data processing.

[0003] Therefore, there is an urgent need for a non-contact fall detection method that can overcome the shortcomings of the existing technology and improve the accuracy of fall detection. Summary of the Invention

[0004] In view of this, the present invention provides a method, device, equipment and storage medium for indoor person fall detection to improve the accuracy of fall detection. The technical solution is as follows.

[0005] In a first aspect, the present invention provides a method for detecting a person falling indoors, the method comprising:

[0006] Acquire temperature distribution data of a target person in a target indoor space; the temperature distribution data is collected by an infrared array sensor deployed in the target indoor space;

[0007] Based on the temperature distribution data, the horizontal morphological features and vertical distance features of the target person are extracted to obtain a fusion feature set; the horizontal morphological features are used to characterize the posture changes of the target person; the vertical distance features are used to characterize the movement trend of the target person;

[0008] The fused feature set is input into the classification model to obtain the fall status determination result of the target person.

[0009] A method for detecting indoor falls is provided. Infrared array sensors deployed in the target indoor space collect real-time temperature distribution data on target individuals. Using non-contact thermal imaging technology, the method replaces traditional wearable devices or microwave radar, eliminating the inconvenience of wearing sensors close to the body. Furthermore, the infrared array sensors only capture low-resolution heat source distribution, eliminating the need for high-definition images or electromagnetic wave reflection signals, simplifying hardware deployment. By extracting lateral morphological features (reflecting human posture) and longitudinal distance features (reflecting movement trends) and constructing a fused feature set, the method comprehensively captures the complex characteristics of fall behavior from both spatial distribution and temporal evolution. Compared to single-dimensional feature description methods (such as analyzing only velocity or acceleration), the fused feature set significantly reduces the probability of misjudgment due to ambiguity in some features. Furthermore, the fused feature set is input into a classification model for intelligent judgment, enabling it to handle complex fall behaviors and overcome the sensitivity issues of traditional threshold rule methods due to environmental interference and individual differences. This improves the accuracy of indoor fall detection while maintaining non-invasive detection.

[0010] In an optional embodiment, before extracting the horizontal morphological features and the vertical distance features of the target person, the method further includes:

[0011] Calculate the variance of temperature distribution data of consecutive frames;

[0012] When it is detected that the variance of temperature distribution data of two or more consecutive frames exceeds the variance threshold, the horizontal morphological features and vertical distance features of the target person are extracted.

[0013] The present invention provides a method for detecting indoor falls. By setting a variance threshold, feature extraction is initiated only when significant temperature changes (such as vigorous human movement) are detected. This avoids false triggering caused by minor ambient temperature fluctuations (such as air conditioning airflow or pet movement), significantly reducing the system's false alarm rate. Continuous multi-frame judgment (two or more frames) filters out transient noise interference (such as brief heat source interference), improving the reliability of feature extraction. Feature extraction is performed only when valid movements occur, reducing the amount of invalid data processing.

[0014] In an optional embodiment, extracting the horizontal morphological features of the target person includes:

[0015] Obtain the number of temperature pixels exceeding the preset room temperature threshold in each frame of temperature distribution data, and compare it with the total number of frames to obtain the average number of valid pixels;

[0016] Perform contour analysis on active pixels in consecutive activation frames to extract morphological features;

[0017] Determine the effective motion area of ​​the target person's body motion based on morphological characteristics;

[0018] Based on the average number of effective pixels and effective action area, the horizontal morphological characteristics of the target person are obtained;

[0019] Among them, the continuous activation frame is the frame sequence between the start frame and the end frame; the start frame is the frame where the variance of the temperature distribution data of two consecutive frames is detected to be greater than the variance threshold for the first time; the end frame is the frame where the variance of the temperature distribution data of two consecutive frames is detected to be less than the variance threshold for the first time.

[0020] This invention provides a method for detecting indoor falls. It extracts the average number of effective pixels and the effective motion area. The average number of effective pixels reflects the duration and spatial coverage of the action, enhancing the ability to distinguish between different types of actions. The activation frame records the start and end time of the action, improving the accuracy of identifying the temporal features of fall behavior. The effective motion area approximates pixel density and contours, mapping changes in human posture into geometric parameters. This avoids misjudgments caused by traditional methods that rely solely on a single parameter (such as speed) and enhances sensitivity to the magnitude of the fall.

[0021] In an optional embodiment, extracting the longitudinal distance feature of the target person includes:

[0022] According to the maximum value of the variance of each frame of temperature distribution data, the maximum temperature distribution variance is obtained;

[0023] Between the start frame and the end frame of the target person's body movement, the number of pixels when the variance of the temperature distribution data of each frame is greater than the temperature difference threshold is obtained to obtain the maximum number of reaction pixels;

[0024] Based on the maximum temperature distribution variance and the maximum number of reaction pixels, the longitudinal distance characteristics of the target person are obtained.

[0025] This invention provides a method for detecting indoor falls. The maximum temperature variance directly reflects the severity of temperature changes, improving detection sensitivity for sudden movements. The maximum number of pixels represents the size of the area affected by the movement. Combined with the maximum temperature variance, it can distinguish falls from other large-scale movements (such as jumping), reducing false positives. This method addresses the difficulty of traditional methods in describing complex fall behaviors using single-dimensional features (such as velocity or acceleration alone).

[0026] In an optional embodiment, the fused feature set is input into the classification model to obtain the fall status determination result of the target person, including:

[0027] Assign different weight factors to each feature in the fusion feature set to obtain fusion features with different weights;

[0028] The fusion features with different weights are input into the classification model to obtain the fall status determination result of the target person.

[0029] This paper proposes a method for indoor fall detection that dynamically adjusts weighting factors based on the contribution of features to fall recognition. This addresses the classification bias caused by equal weighting of all features in traditional detection methods and improves the model's focus on key features. This weight adjustment allows the algorithm to adapt to different environments and individual differences, enhancing its generalizability.

[0030] In an optional embodiment, the classification model includes a K-nearest neighbor algorithm model or a neural network model.

[0031] This invention provides a method for detecting indoor falls. In small sample scenarios (e.g., limited home user data), KNN uses a similarity metric to quickly identify fall patterns, avoiding the limitation of neural networks requiring large amounts of training data and lowering the deployment threshold. In large data scenarios (e.g., multi-user data from nursing homes), neural networks use deep feature extraction to capture nonlinear patterns in fall behavior (e.g., implicit associations between different falling postures), improving detection accuracy in complex scenarios. The model type is selected based on the application scenario requirements (e.g., KNN for high real-time requirements, neural networks for high accuracy requirements) to meet diverse user needs.

[0032] In an optional embodiment, the infrared array sensor includes a top-view sensor group deployed on the top of the target indoor space and a side-view sensor group deployed on the side of the target indoor space.

[0033] This invention provides a method for detecting indoor falls. A top sensor captures vertical changes in a person's posture (e.g., a sudden drop in height during a fall), while a side sensor monitors horizontal displacement (e.g., the lateral trajectory of a fall). Multi-view fusion eliminates single-sensor blind spots (e.g., the difficulty of detecting ground contact with a top-view sensor). By jointly analyzing top-view and side-view data, the two-dimensional temperature distribution is mapped into a three-dimensional spatial trajectory (e.g., fall angle and contact point location), enhancing the geometric description of the fall posture. Multiple sensors reduce the risk of detection failure caused by occlusion of a single sensor (e.g., furniture blocking a side-view sensor), thereby improving system reliability.

[0034] In summary, the indoor fall detection method provided by the present invention, in the data collection stage, forms a spatial complementary effect through the coordinated deployment of multiple groups of infrared array sensors. The downward-looking sensors capture the sudden drop in vertical height (such as the downward shift of the human center of gravity during a fall), and the side-looking sensors monitor the horizontal displacement trajectory (such as the lateral fall path). Multi-view data fusion provides a three-dimensional spatial modeling foundation for the subsequent extraction of lateral morphological features and longitudinal distance features, solving the problem of missed detection caused by the blind spot of a single sensor perspective. In the feature extraction stage, the variance threshold trigger mechanism and the fusion of horizontal and vertical features form a temporal-spatial association: feature extraction is initiated by the temperature variance exceeding the threshold of consecutive frames. Combined with the spatiotemporal parameters of the effective action area and the maximum temperature variance, the action starting point is identified and the dynamic process is accurately segmented, avoiding the defect of single time point misjudgment in traditional solutions. In the classification stage, the weight factor allocation and the optional classification model form a dynamic adaptive closed loop. The lateral morphological features and longitudinal distance features dynamically adjust the classification boundary through the contribution weight. Combined with the flexible switching of the KNN model (small sample scenario) and the neural network (big data scenario), the algorithm can be generalized and adapted to different scenarios.

[0035] In a second aspect, the present invention provides an indoor person fall detection device, the device comprising:

[0036] An acquisition module is used to acquire temperature distribution data of a target person in a target indoor space; the temperature distribution data is collected by an infrared array sensor deployed in the target indoor space;

[0037] The extraction module is used to extract the target person's transverse morphological features and longitudinal distance features based on the temperature distribution data to obtain a fused feature set; the transverse morphological features are used to characterize the target person's posture changes; the longitudinal distance features are used to characterize the target person's movement trend;

[0038] The judgment module is used to input the fusion feature set into the classification model to obtain the fall status judgment result of the target person.

[0039] In a third aspect, the present invention provides a computer device comprising: a memory and a processor, the memory and the processor being communicatively connected to each other, the memory storing computer instructions, and the processor executing the indoor person fall detection method of the first aspect or any corresponding embodiment thereof by executing the computer instructions.

[0040] In a fourth aspect, the present invention provides a computer-readable storage medium having computer instructions stored thereon, the computer instructions being used to enable a computer to execute the indoor person fall detection method of the above-mentioned first aspect or any corresponding embodiment thereof.

[0041] In a fifth aspect, the present invention provides a computer program product, comprising computer instructions, which are used to enable a computer to execute the indoor person fall detection method of the first aspect or any corresponding embodiment thereof. BRIEF DESCRIPTION OF THE DRAWINGS

[0042] In order to more clearly illustrate the specific embodiments of the present invention or the technical solutions in the prior art, the following briefly introduces the drawings required for use in the specific embodiments or the description of the prior art. Obviously, the drawings described below are some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying any creative work.

[0043] Figure 1 1 is a flow chart of a method for detecting an indoor person falling according to an embodiment of the present invention;

[0044] Figure 2 1 is a flow chart of a method for detecting indoor falls according to a specific embodiment of the present invention;

[0045] Figure 3 This is a structural block diagram of an indoor person fall detection device according to an embodiment of the present invention;

[0046] Figure 4 Schematic diagram of the hardware structure of a computer device according to an embodiment of the present invention. DETAILED DESCRIPTION

[0047] To make the purpose, technical solutions, and advantages of the embodiments of the present invention more clear, the technical solutions in the embodiments of the present invention will be clearly and completely described below in conjunction with the accompanying drawings in the embodiments of the present invention. Obviously, the described embodiments are part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without making creative efforts shall fall within the scope of protection of the present invention.

[0048] To overcome the shortcomings of existing fall detection solutions, an embodiment of the present invention provides an embodiment of a method for detecting indoor person falls. It should be noted that the steps shown in the flowcharts of the accompanying drawings can be executed in a computer system such as a set of computer-executable instructions, and although a logical order is shown in the flowcharts, in some cases, the steps shown or described can be executed in an order different from that shown here.

[0049] The method flow of the indoor person fall detection method of this embodiment is as follows: Figure 1 As shown, the following steps are included:

[0050] S101. Acquire temperature distribution data of a target person in a target indoor space. The temperature distribution data is collected by an infrared array sensor deployed in the target indoor space.

[0051] Specifically, infrared array sensors (non-contact thermal imaging devices) deployed in the target indoor space collect the target person's thermal radiation signals in real time, generating a two-dimensional temperature distribution matrix. Each matrix element (pixel) corresponds to the temperature value at a specific location in the indoor space, forming dynamic data reflecting the heat distribution on the human body surface. This system relies solely on infrared thermal imaging technology, without the use of other sensing methods such as visible light cameras or microwave radar. The temperature distribution data is a low-resolution thermal map (e.g., an 8×8 pixel array) to avoid the privacy issues that may arise with high-definition imaging.

[0052] S102. Based on the temperature distribution data, extract the horizontal morphological features and vertical distance features of the target person to obtain a fusion feature set; the horizontal morphological features are used to characterize the posture changes of the target person; and the vertical distance features are used to characterize the movement trend of the target person.

[0053] Specifically, the lateral morphological feature is a parameter related to the spatial distribution of human posture extracted from the temperature distribution data. This feature is used to quantify the morphological changes of human posture in the horizontal direction (such as standing upright, leaning, and falling). For example, the geometric shape of the human body contour (such as the approximate rectangular area, aspect ratio), and the spatial clustering of effective temperature pixels (such as the center of mass position offset). The longitudinal distance feature is a parameter related to the temporal evolution of the action extracted from the temperature distribution data. This feature is used to characterize the dynamic trend of human action in the time dimension (such as the speed difference between sitting down slowly and falling quickly). For example, the moving speed of the center point of the temperature distribution; the diffusion rate of the hot area (such as the sudden change in the local temperature field caused by the human body touching the ground when falling). The lateral morphological features (spatial distribution parameters) and the longitudinal distance features (temporal dynamic parameters) are combined into a multidimensional vector to form a fusion feature set.

[0054] S103: Input the fused feature set into the classification model to obtain the fall status determination result of the target person.

[0055] Specifically, a classification model built using a machine learning algorithm (such as KNN, SVM, or a neural network) receives a fused feature set as input, learns the characteristic differences between normal activities and falls through training, and outputs a binary classification result (fall / non-fall). The classification model calculates the probability based on the distribution of the fused feature set in the multidimensional feature space (such as the distance to the fall cluster). If the probability exceeds a preset threshold, it is determined to be a fall. For example, when the contour area increases sharply (horizontally) and the speed exceeds a critical value (longitudinal distance), the model determines it as a fall.

[0056] Optionally, in step S101 above, the infrared array sensor includes a top-view sensor group deployed at the top of the target indoor space and a side-view sensor group deployed at the side of the target indoor space. The top sensor captures vertical posture changes of the human body (such as a sudden drop in height during a fall), while the side sensor monitors horizontal displacement (such as the lateral fall trajectory). Multi-view fusion eliminates single-sensor blind spots (such as the difficulty of the top-view sensor in detecting ground-facing movements). By jointly analyzing the top-view and side-view data, the two-dimensional temperature distribution is mapped into a three-dimensional spatial motion trajectory (such as the fall angle and contact point location), enhancing the geometric description of the fall posture. Multiple sensors can reduce the risk of detection failure caused by occlusion of a single sensor (such as furniture blocking the side-view sensor), thereby improving system reliability.

[0057] Optionally, before step S102, the above-mentioned step also includes a determination of whether to perform feature extraction, specifically including: calculating the variance of the temperature distribution data of consecutive frames; when it is detected that the variance of the temperature distribution data of two or more consecutive frames exceeds the variance threshold, extracting the horizontal morphological features and longitudinal distance features of the target person. By setting the variance threshold, feature extraction is only initiated when a significant temperature change is detected (such as a violent human movement), avoiding false triggering caused by slight fluctuations in ambient temperature (such as air conditioning airflow, pet movement), and significantly reducing the system's false alarm rate. Continuous multi-frame judgment (two or more frames) can filter out transient noise interference (such as short-term heat source interference) and improve the reliability of feature extraction. Feature extraction is only performed when valid actions occur, reducing the amount of invalid data processing.

[0058] Optionally, the above step S102 includes the extraction of transverse morphological features and longitudinal morphological features.

[0059] Extracting lateral morphological features specifically includes: obtaining the number of temperature pixels exceeding a preset room temperature threshold in each frame of temperature distribution data, and comparing it with the total number of frames to obtain an average number of valid pixels; performing contour analysis on active pixels in continuously activated frames to extract morphological features; determining the effective motion area of ​​the target person's body motion based on the morphological features; obtaining the lateral morphological features of the target person based on the average number of valid pixels and the effective motion area; wherein the continuously activated frames are a frame sequence between a start frame and an end frame; the start frame is a frame in which the variance of two consecutive frames of temperature distribution data is detected for the first time to exceed the variance threshold; the end frame is a frame in which the variance of two consecutive frames of temperature distribution data is detected for the first time to be lower than the variance threshold.

[0060] In the above steps, the effective action area extraction includes the body contour area and center of mass offset. The contour area is obtained by binarizing the temperature distribution matrix and counting the number of valid pixels. The center of mass offset is calculated by calculating the distance between the geometric center of the valid pixel area and the reference position. The contour area is obtained through contour analysis, which binarizes the temperature distribution matrix, extracts the valid pixel area, and calculates the body projected area using the number of pixels and the actual area per pixel.

[0061] Extract longitudinal distance features, including the movement speed of the temperature distribution center. The movement speed is calculated as the ratio of the distance between the center points of adjacent frames within a sliding time window to the time interval, and then processed through a moving average filter. Specifically, the maximum temperature distribution variance is obtained based on the maximum value of the variance of each frame's temperature distribution data. Between the start and end frames of the target person's body movement, the number of pixels where the variance of each frame's temperature distribution data exceeds the temperature difference threshold is obtained to obtain the maximum number of pixels. Based on the maximum temperature distribution variance and the maximum number of reaction pixels, the longitudinal distance feature of the target person is obtained.

[0062] Among the features extracted above, the average number of effective pixels is defined as the total number of temperature pixels exceeding the room temperature threshold divided by the total number of frames, that is, the average number of pixels in a single frame exceeding the room temperature threshold; the activation frame represents the duration of the action. If there is no one in the indoor environment, the feature value is very small and basically does not change, but when there is someone and movement occurs, the feature value will increase; the maximum number of reaction pixels represents the maximum number of pixels when the variance of the number of single pixels in each frame is greater than the temperature difference threshold between the start frame and the end frame of the human body movement. This feature represents the number of pixels collected by the infrared sensor when the corresponding action occurs; the maximum temperature distribution variance represents the maximum temperature distribution variance between the start frame and the end frame of the human body movement. This feature represents the degree of change in the movement trend; the effective action area is to approximate the outline of the human body action into a regular rectangle, and the approximate area size of the action is calculated as the effective action area. By measuring the number of active pixels in each frame exceeding the activation frame and dividing it by the number of pixels per square, the approximate area where the actual action captured in the frame occurred can be obtained. This data feature can not only be used as a training feature, but also as verification data to verify whether the algorithm is accurate. By comparing the effective action area with the actual estimated action area, it can be determined what kind of action occurred.

[0063] Optionally, in step S103, different weighting factors are first assigned to each feature in the fused feature set to obtain fused features with different weights. These fused features with different weights are then input into the classification model to determine the target person's fall status. The weighting factors are dynamically adjusted based on the feature's contribution to fall detection, addressing the classification bias caused by equal weighting of all features in traditional detection methods and improving the model's focus on key features. This weighting adjustment allows for adaptation to different environments and individual differences, enhancing the algorithm's generalizability.

[0064] In the above steps, the classification model can use either a K-nearest neighbor algorithm or a neural network model. In small-sample scenarios (e.g., limited home user data), KNN uses similarity metrics to quickly identify fall patterns, avoiding the large amount of training data required by neural networks and lowering the deployment threshold. In large-scale scenarios (e.g., multi-user data from nursing homes), neural networks use deep feature extraction to capture nonlinear patterns in fall behavior (e.g., implicit correlations between different falling postures), improving detection accuracy in complex scenarios. The model type should be selected based on the application scenario (e.g., KNN for high real-time requirements, neural networks for high accuracy) to meet diverse user needs.

[0065] In summary, the indoor fall detection method provided by the embodiments of the present invention, during the data collection phase, achieves a spatial complementarity effect through the coordinated deployment of multiple infrared array sensor groups. The downward-looking sensors capture vertical height drops (e.g., the downward shift of a person's center of gravity during a fall), while the side-looking sensors monitor horizontal displacement trajectories (e.g., the lateral fall path). Multi-view data fusion provides a three-dimensional spatial modeling foundation for the subsequent extraction of lateral morphological features and longitudinal distance features, thus addressing the problem of missed detection caused by blind spots in the perspective of a single sensor. During the feature extraction phase, a variance threshold trigger mechanism is combined with the fusion of lateral and longitudinal features to form a temporal-spatial correlation. Feature extraction is initiated when the temperature variance of consecutive frames exceeds the threshold. Combined with the spatiotemporal parameters of the effective action area and the maximum temperature variance, this enables the identification of the action starting point and the precise segmentation of the dynamic process, avoiding the flaw of single-time point misjudgment in traditional solutions. During the classification phase, the weight factor allocation and the optional classification model form a dynamic adaptive closed loop. The lateral morphological features and longitudinal distance features are dynamically adjusted through contribution weights to adjust the classification boundaries. The flexible switching between the KNN model (for small sample scenarios) and the neural network (for large data scenarios) enables the algorithm to be generalized and adapted to different scenarios.

[0066] For example, a specific example is used below to illustrate the process of the above-mentioned indoor person fall detection method. The method process of this example is as follows: Figure 2 As shown, the following steps are included.

[0067] First, infrared array sensors deployed on the top and sides of the room acquire each frame of array data and perform a maximum variance calculation on the temperature pixels. If the temperature distribution variance of two or more consecutive frames exceeds the temperature variance threshold, sensitive data features of human posture are extracted, including the average number of valid pixels, activation frames, maximum number of reaction pixels, maximum temperature distribution variance, and effective motion area. Finally, the extracted feature data is preprocessed according to weighting factors before KNN classification is performed to determine whether a fall has occurred. In addition to the KNN algorithm, other learning algorithms with similar classification results, such as the random forest algorithm, can also be used.

[0068] By replacing microwave radar technology with infrared technology using the target human body as the source, a highly private low-resolution camera is used to collect human body information in real time. Multi-dimensional parameter fusion is used for feature extraction, and artificial intelligence algorithms are used to identify and determine the extracted features, thereby achieving accurate fall monitoring of the target human body. The contribution of each dimension of the feature to fall recognition varies. Within the feature space of the classification algorithm, a weight factor corresponding to the fall contribution is assigned when measuring the distance between each feature parameter and the nearest neighbor, thereby completing the adaptive optimization of the classification algorithm model in multi-dimensional fusion scenarios.

[0069] This embodiment also provides an indoor person fall detection device, which is used to implement the above-mentioned embodiments and preferred implementations. Details already described will not be repeated here. As used below, the term "module" may refer to a combination of software and / or hardware that implements a predetermined function. Although the devices described in the following embodiments are preferably implemented in software, implementation using hardware, or a combination of software and hardware, is also possible and contemplated.

[0070] The structure of an indoor person fall detection device provided in this embodiment is as follows Figure 3 As shown, including:

[0071] An acquisition module 301 is used to acquire temperature distribution data of a target person in a target indoor space; the temperature distribution data is collected by an infrared array sensor deployed in the target indoor space;

[0072] Extraction module 302 is used to extract the target person's transverse morphological features and longitudinal distance features based on the temperature distribution data to obtain a fused feature set; the transverse morphological features are used to characterize the target person's posture changes; the longitudinal distance features are used to characterize the target person's movement trend;

[0073] The judgment module 303 is used to input the fusion feature set into the classification model to obtain the fall status judgment result of the target person.

[0074] In an optional embodiment, the device further comprises:

[0075] The threshold verification module 304 is used to calculate the variance of the temperature distribution data of consecutive frames; when it is detected that the variance of the temperature distribution data of two or more consecutive frames exceeds the variance threshold, the horizontal morphological features and vertical distance features of the target person are extracted.

[0076] In an optional implementation, the extraction module 302 is specifically configured to:

[0077] Obtain the number of temperature pixels exceeding the preset room temperature threshold in each frame of temperature distribution data, and compare it with the total number of frames to obtain the average number of valid pixels;

[0078] Perform contour analysis on active pixels in consecutive activation frames to extract morphological features;

[0079] Determine the effective motion area of ​​the target person's body motion based on morphological characteristics;

[0080] Based on the average number of effective pixels and effective action area, the horizontal morphological characteristics of the target person are obtained;

[0081] Among them, the continuous activation frame is the frame sequence between the start frame and the end frame; the start frame is the frame where the variance of the temperature distribution data of two consecutive frames is detected to be greater than the variance threshold for the first time; the end frame is the frame where the variance of the temperature distribution data of two consecutive frames is detected to be less than the variance threshold for the first time.

[0082] In an optional implementation, the extraction module 302 is specifically configured to:

[0083] According to the maximum value of the variance of each frame of temperature distribution data, the maximum temperature distribution variance is obtained;

[0084] Between the start frame and the end frame of the target person's body movement, the number of pixels when the variance of the temperature distribution data of each frame is greater than the temperature difference threshold is obtained to obtain the maximum number of pixels;

[0085] Based on the maximum temperature distribution variance and the maximum number of reaction pixels, the longitudinal distance characteristics of the target person are obtained.

[0086] In an optional implementation, the judgment module 303 is specifically configured to:

[0087] Assign different weight factors to each feature in the fusion feature set to obtain fusion features with different weights;

[0088] The fusion features with different weights are input into the classification model to obtain the fall status determination result of the target person.

[0089] In an optional embodiment, the classification model includes a K-nearest neighbor algorithm model or a neural network model.

[0090] The further functional description of each of the above modules and units is the same as that of the above corresponding embodiments and will not be repeated here.

[0091] The indoor person fall detection device in this embodiment is presented in the form of a functional unit, where the unit refers to an ASIC (Application Specific Integrated Circuit) circuit, a processor and memory that executes one or more software or fixed programs, and / or other devices that can provide the above functions.

[0092] The embodiment of the present invention also provides a computer device having the above Figure 3 The indoor personnel fall detection device shown.

[0093] See also Figure 4 , Figure 4 is a structural diagram of a computer device provided by an optional embodiment of the present invention, such as Figure 4 As shown, the computer device includes: one or more processors 10, memory 20, and interfaces for connecting various components, including high-speed interfaces and low-speed interfaces. Various components utilize different buses to communicate with each other and can be installed on a common mainboard or installed in other ways as needed. The processor can process the instructions executed in the computer device, including instructions stored in the memory or on the memory to display the graphical information of the GUI on an external input / output device (such as, a display device coupled to the interface). In some optional embodiments, if necessary, multiple processors and / or multiple buses can be used together with multiple memories and multiple memories. Equally, multiple computer devices can be connected, and each device provides part of the necessary operations (for example, as a server array, a group of blade servers, or a multi-processor system). Figure 4 A processor 10 is taken as an example.

[0094] The processor 10 may be a central processing unit, a network processor, or a combination thereof. The processor 10 may further include a hardware chip. The hardware chip may be an application-specific integrated circuit, a programmable logic device, or a combination thereof. The programmable logic device may be a complex programmable logic device, a field programmable gate array, a general purpose array logic, or any combination thereof.

[0095] The memory 20 stores instructions that can be executed by at least one processor 10, so that the at least one processor 10 executes the method shown in the above embodiment.

[0096] The memory 20 may include a program storage area and a data storage area, wherein the program storage area may store an operating system and application programs required for at least one function; the data storage area may store data created based on the use of the computer device, etc. In addition, the memory 20 may include a high-speed random access memory, and may also include a non-transient memory, such as at least one disk storage device, a flash memory device, or other non-transient solid-state storage device. In some optional embodiments, the memory 20 may optionally include a memory remotely located relative to the processor 10, and these remote memories may be connected to the computer device via a network. Examples of the above-mentioned network include, but are not limited to, the Internet, an intranet, a local area network, a mobile communication network, and combinations thereof.

[0097] The memory 20 may include a volatile memory, such as a random access memory; the memory may also include a non-volatile memory, such as a flash memory, a hard disk or a solid-state drive; the memory 20 may also include a combination of the above types of memory.

[0098] The computer device also includes an input device 30 and an output device 40. The processor 10, the memory 20, the input device 30 and the output device 40 can be connected via a bus or other means. Figure 4 The bus connection is taken as an example.

[0099] The input device 30 can receive input digital or character information and generate key signal input related to user settings and function control of the computer device. Examples include a touch screen, keypad, mouse, trackpad, touchpad, pointing stick, one or more mouse buttons, trackball, joystick, etc. The output device 40 may include a display device, auxiliary lighting devices (e.g., LEDs), and tactile feedback devices (e.g., vibration motors). Such display devices include, but are not limited to, liquid crystal displays, light emitting diodes, monitors, and plasma displays. In some optional embodiments, the display device may be a touch screen.

[0100] The embodiment of the present invention also provides a computer-readable storage medium. The above-mentioned method according to the embodiment of the present invention can be implemented in hardware, firmware, or implemented as a computer code that can be recorded in a storage medium, or implemented as a computer code that is originally stored in a remote storage medium or a non-temporary machine-readable storage medium and downloaded through a network and will be stored in a local storage medium, so that the method described herein can be stored in such software processing on a storage medium using a general-purpose computer, a dedicated processor, or programmable or dedicated hardware. Among them, the storage medium can be a magnetic disk, an optical disk, a read-only storage memory, a random access memory, a flash memory, a hard disk or a solid-state drive, etc.; further, the storage medium can also include a combination of the above-mentioned types of memory. It can be understood that a computer, a processor, a microprocessor controller or programmable hardware includes a storage component that can store or receive software or computer code. When the software or computer code is accessed and executed by a computer, a processor or hardware, the method shown in the above embodiment is implemented.

[0101] A portion of the present invention may be applied as a computer program product, such as a computer program instruction, which, when executed by a computer, can call or provide the method and / or technical solution according to the present invention through the operation of the computer. Those skilled in the art should understand that the form in which the computer program instruction exists in a computer-readable medium includes, but is not limited to, a source file, an executable file, an installation package file, etc. Accordingly, the way in which the computer program instruction is executed by the computer includes, but is not limited to: the computer directly executes the instruction, or the computer compiles the instruction and then executes the corresponding compiled program, or the computer reads and executes the instruction, or the computer reads and installs the instruction and then executes the corresponding installed program. Here, the computer-readable medium may be any available computer-readable storage medium or communication medium that can be accessed by the computer.

[0102] Although the embodiments of the present invention have been described with reference to the accompanying drawings, those skilled in the art may make various modifications and variations without departing from the spirit and scope of the present invention. Such modifications and variations are all within the scope defined by the appended claims.

Claims

1. A method for detecting indoor falls, characterized in that: The method comprises: Acquiring temperature distribution data of a target person in a target indoor space; the temperature distribution data is collected by an infrared array sensor deployed in the target indoor space; Based on the temperature distribution data, the horizontal morphological features and the vertical distance features of the target person are extracted to obtain a fusion feature set; the horizontal morphological features are used to characterize the posture changes of the target person; and the vertical distance features are used to characterize the movement trend of the target person; Inputting the fused feature set into a classification model to obtain a fall status determination result of the target person; The infrared array sensor includes a top-view sensor group deployed on the top of the target indoor space and a side-view sensor group deployed on the side of the target indoor space; The method further includes: mapping the two-dimensional temperature distribution into a three-dimensional spatial motion trajectory by jointly analyzing the top view and side view data; Before extracting the horizontal morphological features and the vertical distance features of the target person, the method further includes: Calculate the variance of temperature distribution data of consecutive frames; When it is detected that the variance of temperature distribution data of two or more consecutive frames exceeds the variance threshold, the horizontal morphological features and vertical distance features of the target person are extracted; Extract the horizontal morphological features of the target person, including: Obtain the number of temperature pixels exceeding the preset room temperature threshold in each frame of temperature distribution data, and compare it with the total number of frames to obtain the average number of valid pixels; Perform contour analysis on active pixels in consecutive activation frames to extract morphological features; Determining an effective motion area of ​​a target person's body motion based on the morphological features; Obtaining a horizontal morphological feature of the target person based on the average number of effective pixels and the effective action area; The continuous activation frame is a frame sequence between a start frame and an end frame; the start frame is a frame where the variance of two consecutive frames of temperature distribution data is detected to be greater than a variance threshold for the first time; the end frame is a frame where the variance of two consecutive frames of temperature distribution data is detected to be less than a variance threshold for the first time; Extract the longitudinal distance features of the target person, including: According to the maximum value of the variance of each frame of temperature distribution data, the maximum temperature distribution variance is obtained; Between the start frame and the end frame of the target person's body movement, the number of pixels when the variance of the temperature distribution data of each frame is greater than the temperature difference threshold is obtained to obtain the maximum number of reaction pixels; Obtaining a longitudinal distance feature of the target person based on the maximum temperature distribution variance and the maximum number of reaction pixels; The longitudinal distance feature includes the moving speed of the center point of the temperature distribution. The moving speed is calculated by the ratio of the distance between the center points of adjacent frames in the sliding time window to the time interval and is processed by moving average filtering.

2. The method according to claim 1, characterized in that The step of inputting the fused feature set into a classification model to obtain a fall state determination result of the target person includes: Assigning different weight factors to each feature in the fusion feature set to obtain fusion features with different weights; The fusion features with different weights are input into the classification model to obtain the fall status determination result of the target person.

3. The method according to claim 2, characterized in that The classification model includes a K-nearest neighbor algorithm model or a neural network model.

4. An indoor person fall detection device, characterized in that: The device comprises: an acquisition module for acquiring temperature distribution data of a target person in a target indoor space; the temperature distribution data is collected by an infrared array sensor deployed in the target indoor space; the infrared array sensor includes a top-view sensor group deployed at the top of the target indoor space and a side-view sensor group deployed at the side of the target indoor space; an extraction module for extracting, based on the temperature distribution data, transverse morphological features and longitudinal distance features of the target person to obtain a fused feature set; the transverse morphological features are used to characterize the posture changes of the target person; and the longitudinal distance features are used to characterize the movement trend of the target person; A judgment module, configured to input the fused feature set into a classification model to obtain a fall status judgment result of the target person; The extraction module is further configured to: map the two-dimensional temperature distribution into a three-dimensional motion trajectory by jointly analyzing the top-view and side-view data; The extraction module is further used to: Calculate the variance of temperature distribution data of consecutive frames; When it is detected that the variance of temperature distribution data of two or more consecutive frames exceeds the variance threshold, the horizontal morphological features and vertical distance features of the target person are extracted; The extraction module is further used to: Obtain the number of temperature pixels exceeding the preset room temperature threshold in each frame of temperature distribution data, and compare it with the total number of frames to obtain the average number of valid pixels; Perform contour analysis on active pixels in consecutive activation frames to extract morphological features; Determining an effective motion area of ​​a target person's body motion based on the morphological features; Obtaining a horizontal morphological feature of the target person based on the average number of effective pixels and the effective action area; The continuous activation frame is a frame sequence between a start frame and an end frame; the start frame is a frame where the variance of two consecutive frames of temperature distribution data is detected to be greater than a variance threshold for the first time; the end frame is a frame where the variance of two consecutive frames of temperature distribution data is detected to be less than a variance threshold for the first time; The extraction module is further used to: According to the maximum value of the variance of each frame of temperature distribution data, the maximum temperature distribution variance is obtained; Between the start frame and the end frame of the target person's body movement, the number of pixels when the variance of the temperature distribution data of each frame is greater than the temperature difference threshold is obtained to obtain the maximum number of reaction pixels; Obtaining a longitudinal distance feature of the target person based on the maximum temperature distribution variance and the maximum number of reaction pixels; The longitudinal distance feature includes the moving speed of the center point of the temperature distribution. The moving speed is calculated by the ratio of the distance between the center points of adjacent frames in the sliding time window to the time interval and is processed by moving average filtering.

5. A computer device, characterized in that: include: A memory and a processor, wherein the memory and the processor are communicatively connected to each other, the memory stores computer instructions, and the processor executes the indoor person fall detection method according to any one of claims 1 to 3 by executing the computer instructions.

6. A computer-readable storage medium, characterized in that The computer-readable storage medium stores computer instructions, and the computer instructions are used to enable a computer to execute the indoor person fall detection method according to any one of claims 1 to 3.

Images