Non-contact blood oxygen and blood pressure real-time monitoring method and system

Abstract

The invention provides a non-contact blood oxygen and blood pressure real-time monitoring system and method, and belongs to the technical field of biomedical engineering. The method comprises the following steps: extracting RGB signals of a facial ROI region, calculating standardized fluctuation characteristics, estimating blood oxygen saturation based on a linear model of a red-green fluctuation ratio, and calculating a blood pressure value through a multiple regression model in combination with a main frequency f; a sliding window mechanism and a skin color mask technology are adopted to enhance environmental adaptability, and real-time processing is realized on embedded equipment. Standardized fluctuation characteristics are used as unified signal representation, traditional AC / DC components or PPG waveforms are replaced, the illumination interference resistance is improved by 40%, a lightweight regression model is combined, double-physiological-parameter synchronous monitoring is achieved on low-computing-power equipment, the blood oxygen error is smaller than or equal to + / -1.5%, and the static scene blood pressure error is smaller than or equal to + / -5 mmHg; and high robustness and real-time performance can be realized in a complex environment, and the hardware and deployment cost is remarkably reduced.

Description

TECHNICAL FIELD

[0001] The present application belongs to the technical field of biomedical engineering, and particularly relates to a non-contact blood oxygen and blood pressure real-time monitoring system and method based on skin light reflection standardization analysis, which combines image processing, optical modeling and regression algorithm. BACKGROUND

[0002] Currently, the measurement technology of blood oxygen and blood pressure mainly relies on contact devices such as finger clip pulse oximeters and electronic sphygmomanometers. Although these devices have high accuracy, they have the following limitations in practical application: Contact restriction: direct contact with the skin may cause discomfort or inconvenience, especially in specific scenarios such as driving.

[0003] Motion artifact: measurement data is easily disturbed in motion state, leading to inaccurate results.

[0004] High cost: high-precision devices are expensive, limiting their use in large-scale deployment.

[0005] With the development of remote photoplethysmography (rPPG) technology, methods of estimating physiological parameters through skin color changes from camera video signals have received widespread attention. However, existing rPPG methods focus on heart rate and respiration rate estimation, and still face challenges in blood oxygen and blood pressure estimation, especially in the presence of environmental light changes and facial motion conditions, where signal quality deteriorates significantly, and real-time performance and hardware adaptability are not ideal. This method has the following limitations: Strong environmental dependence: changes in lighting conditions or environmental light interference can significantly reduce measurement accuracy, i.e., sensitivity to changes in lighting conditions, leading to a decrease in signal quality; Poor dynamic scene adaptability: in motion or facial occlusion scenarios, signal quality is low, i.e., poor robustness in dynamic scenarios such as motion or facial occlusion.

[0006] High computational complexity: real-time performance cannot meet the high-frequency needs of health monitoring, i.e., real-time computational complexity is high, and cannot efficiently run on low-power hardware.

[0007] The invention patent with publication number CN118986291A discloses a non-contact health monitoring method with privacy protection features, focusing on video privacy compression and facial pixel rearrangement technology, which has innovative significance, but relies on a complex compression encoding process, making it difficult to deploy, and does not form a stable modeling mechanism for blood oxygen and blood pressure indicators, making it difficult to support embedded devices and low-power, high-dynamic environments such as traffic scenarios.

[0008] Therefore, there is an urgent need for a non-contact health monitoring method that can operate with high robustness in complex environments, has simple calculations, clear modeling, and can be deployed in actual traffic and health devices. SUMMARY

[0009] The technical problem solved by the present application is to provide a non-contact blood oxygen and blood pressure real-time monitoring method and system based on skin light reflection standardization analysis, which has high robustness and real-time performance in complex lighting and dynamic scenes, is suitable for embedded deployment of low-power devices, and solves the problems of the prior art, such as the need to contact the skin, poor dynamic scene robustness, and high model complexity.

[0010] The technical solution adopted by the present application is: a non-contact blood oxygen and blood pressure real-time monitoring method, comprising the following steps: Step 1, face video acquisition and ROI region extraction; Step 1.1, using a normal RGB camera to capture a face image sequence in real time; Step 1.2, applying Dlib / Mediapipe algorithm to realize real-time face detection, locking the forehead or cheek region as the region of interest (ROI) region; Step 1.3, after skin color mask processing through YCbCr color space, outputting real-time video data of the ROI region; Step 2, RGB signal extraction and standardization calculation; Step 2.1, calculating the mean value of each frame of the RGB channel of the ROI region to form time series signals R(t), G(t), and B(t): , , ; Wherein, R(t), G(t), and B(t) represent the average pixel value of the red, green, and blue channels in the ROI region of the t-th frame of image respectively; Step 2.2, calculating the standardization fluctuation characteristics of the signals , , and outputting the standardized RGB signals; , ; Wherein, std(R) represents the standard deviation of the red channel signal, and mean(R) represents the average value of the red channel signal; std(G) represents the standard deviation of the green channel signal, and mean(G) represents the average value of the green channel signal; Step 3, blood oxygen detection; Step 3.1, calculating the red-green light fluctuation ratio , ; Step 3.2, based on experimental calibration data, using a linear regression model to estimate the blood oxygen saturation SpO 2; ; in, k 1 ，k 2 These are the parameters for the experimental fitting. Step 4: Blood pressure measurement; Step 4.1: Extract the main frequency from the normalized RGB signal. f Combined with standardized fluctuation characteristics S R , S G Constructing a multiple regression model for blood pressure: Systolic blood pressure ; diastolic blood pressure ; in, These are the experimental calibration coefficients; Step 4.2: Calculate and output the real-time systolic blood pressure (SBP) and diastolic blood pressure (DBP) based on the model; Step 5: Result Display and Anomaly Warning; Step 5.1: Set the normal threshold range for SpO2, SBP, and DBP. When the result exceeds the threshold, trigger a prompt and data storage. Step 5.2: Conduct long-term trend analysis and remote health early warning.

[0011] Preferably, the ordinary RGB camera mentioned in step 1.1 is a visible light camera capable of providing color video images, including a built-in camera in a laptop, an external USB camera, a Raspberry Pi camera module, a dashcam, or a front-facing camera on a smartphone; the image acquisition frame rate is not less than 20 frames / second, and the resolution is not less than 640×480.

[0012] Preferably, the determination criteria for skin color mask processing in step 1.3 are that the Cb component is in [77, 127] and the Cr component is in [133, 173], and the effectiveness of the facial ROI is cross-validated in this region.

[0013] Preferably, the standardized fluctuation features described in step 2.2 are dynamically calculated using a sliding window mechanism, with a window length of 5-10 seconds of continuous frames, and a mean-mode filter is applied to suppress noise.

[0014] Preferably, the main frequency mentioned in step 4.1 f Obtained from the standardized green channel signal, the specific extraction steps are as follows: First, the standardized green channel signal G... norm (t) Perform a 0.6-3Hz bandpass filter; then calculate the power spectrum of the signal within a 5-second window; finally, take the frequency corresponding to the maximum spectral peak as... f .

[0015] A non-contact real-time blood oxygen and blood pressure monitoring system includes a standard RGB camera, an embedded processor, an ROI extraction module, a standardized fluctuation feature calculation module, a blood oxygen and blood pressure regression model calculation module, and a result display and early warning module. The result display and early warning module supports graphical visualization of data, including drawing historical trend curves, fluctuation bar charts, or change rate heatmaps. Users can choose to use the system's default threshold or customize the threshold according to their individual circumstances, and an alarm is triggered when the threshold is exceeded.

[0016] The beneficial effects of this invention are as follows: This invention proposes to use standardized fluctuation characteristics as a unified signal representation to replace the traditional AC / DC component or PPG waveform, which improves the anti-light interference capability by 40%. Combined with a lightweight regression model, it can realize the synchronous monitoring of dual physiological parameters on low computing power devices, achieving blood oxygen error ≤ ±1.5% and static scene blood pressure error ≤ ±5mmHg. It can achieve high robustness and real-time performance in complex environments, and significantly reduce hardware and deployment costs. Attached Figure Description

[0017] Figure 1 This is a flowchart of a non-contact real-time blood oxygen and blood pressure monitoring system.

[0018] Figure 2 This is a schematic diagram of ROI extraction and signal processing.

[0019] Figure 3 This is a flowchart for blood oxygen calculation.

[0020] Figure 4 This is a flowchart of a blood pressure regression model. Detailed Implementation

[0021] To make the objectives, technical solutions, and advantages of this invention clearer, the invention will be further described in detail below with reference to embodiments. It should be understood that the specific embodiments described herein are merely illustrative and not intended to limit the invention.

[0022] This invention provides a non-contact blood oxygen and blood pressure monitoring method based on light reflection characteristics and standardized fluctuation features. By optimizing signal processing and modeling algorithms, it can achieve high robustness and real-time performance in complex environments, significantly reducing hardware and deployment costs. It can be widely applied in the following scenarios: Intelligent driver health management: Monitors the health status of bus, truck, and other drivers to reduce the risk of traffic accidents.

[0023] Elderly and long-term patient care: Providing non-invasive monitoring in hospital or home settings to reduce the burden on healthcare.

[0024] Health management in high-risk environments: such as health protection in high-altitude areas and under extreme weather conditions.

[0025] Mobile healthcare and remote diagnostics: It can serve as a portable health assessment system, providing cross-regional health monitoring services.

[0026] like Figures 1 to 4 As shown, a non-contact real-time blood oxygen and blood pressure monitoring method includes the following steps: Step 1: Facial video capture and ROI region extraction; Step 1.1: Capture facial image sequences in real time using a regular RGB camera; Step 1.2: Apply the Dlib / Mediapipe algorithm to achieve real-time face detection and lock the forehead or cheek area as the region of interest (ROI). Step 1.3: After processing the skin tone mask using the YCbCr color space, output the real-time video data of the ROI region.

[0027] The facial video capture and ROI extraction process is as follows: The video capture device is a visible light camera capable of providing color video images, including but not limited to built-in laptop cameras, USB external cameras, Raspberry Pi camera modules, dashcams, and smartphone front-facing cameras. The image capture frame rate is no less than 20 frames per second, and the resolution is no less than 640×480 to ensure that the acquired facial images have sufficient spatiotemporal resolution for light signal analysis. In the captured video image sequence, the system performs face detection on each frame. Lightweight neural network methods such as the Dlib feature point model or the Mediapipe facial mesh recognition model are preferred, but Haar cascade classifiers or deep learning detection models (such as YOLOv5-face) can also be used for face localization. The system extracts the coordinates of the face bounding box and the positions of key points from the detection results. Based on the face bounding box information, the system extracts fixed regions in the image as the facial region of interest (ROI), which typically includes, but is not limited to, the forehead region and the left and right cheek regions. The area must meet the following conditions: (1) the skin surface is stable and there is no hair covering it; (2) the range of motion is small in a static state; (3) it has obvious light reflection characteristics, which is conducive to color change detection.

[0028] To further enhance the accuracy of skin tone regions and filter out background noise, this step converts the image from the RGB color space to the YCbCr color space. The skin tone determination range is set to [77, 127] for the Cb component and [133, 173] for the Cr component. Only pixel regions meeting this condition are retained as candidate skin tone masks, and the effectiveness of facial ROI regions is cross-validated within this range.

[0029] Furthermore, the system smooths the Regions of Interest (ROIs) in consecutive frames to reduce the impact of jitter. Kalman filtering, mean tracking, or optical flow methods (such as Lucas-Kanade) are preferably used to update the ROI bounding boxes, ensuring regional consistency in the facial area even under slight motion conditions. Finally, the extracted ROIs serve as the base regions for RGB channel mean extraction in each subsequent frame, further used to form time-series signals R(t), G(t), and B(t), laying the data foundation for color fluctuation signal modeling and normalization in step two.

[0030] Step 2: RGB signal extraction and standardization calculation; Step 2.1: Calculate the mean of the RGB channels for each frame in the ROI region to form time series signals R(t), G(t), and B(t): The specific process is as follows: For each frame of image, the system extracts the pixel values ​​of the red, green, and blue channels in the ROI region, and calculates their channel averages to form a color intensity sequence signal that changes over time, as shown below: ; ; ; Where R(t), G(t), and B(t) represent the average pixel values ​​of the red, green, and blue channels in the ROI region of the t-th frame image, respectively, forming an RGB three-channel time series signal to reflect the subtle changes in skin color under blood pulsation.

[0031] Step 2.2: Calculate the standardized fluctuation characteristics of the signal. , It outputs a standardized RGB signal; The ROI regions extracted from the acquired video frames undergo RGB channel signal calculation and standardization. To eliminate interference from factors such as changes in ambient lighting and automatic camera exposure adjustments, the system normalizes the extracted RGB time-series signals. Standard deviation normalization is preferred, and the color fluctuation amplitude indices for the red and green channels are calculated using the following expressions: , ; Where std(R) represents the standard deviation of the red channel signal, and mean(R) represents the average value of the red channel signal; std(G) represents the standard deviation of the green channel signal, and mean(G) represents the average value of the green channel signal.

[0032] The above ratio S R, S G This can be considered a parameter representing color volatility, reflecting the relative change amplitude of different color channels per unit time. It has better resistance to light interference and is suitable for subsequent blood oxygen and blood pressure modeling. Furthermore, to improve the stability and robustness of the calculation, this invention employs a sliding window mechanism to dynamically calculate the mean and standard deviation in practical implementation. The window length can be set to consecutive frames within 5 to 10 seconds, for example, 30 to 100 frames. Median filtering or moving average filtering is combined to denoise the RGB signal, suppressing the impact of sudden interference frames on the volatility index. The final output of this step is the standardized RGB time series signal R(t), G(t), B(t) and its volatility index. S R , S G This will serve as the core input feature for subsequent blood oxygen ratio calculation and multiple regression modeling.

[0033] Step 3: Blood oxygen saturation test; Step 3.1: Calculate the red-green light fluctuation ratio; The specific process is as follows: First, the system uses the color fluctuation amplitude parameter obtained in step 2. S R and S G The relative volatility ratio of the red-green channel is calculated and is called the red-green volatility ratio, or R. ratio Its expression is as follows: ; Among them, R ratio It reflects the relative intensity of tiny fluctuations in red and green light on the skin surface, indirectly revealing the difference in absorption of different wavelengths of light by oxygenated and deoxygenated hemoglobin in the blood. Therefore, it can be used as a core characteristic parameter for estimating blood oxygen saturation.

[0034] Step 3.2: Based on the experimental calibration data, use a linear regression model to estimate blood oxygen saturation. SpO 2 ; The system is based on the true blood oxygen values ​​of multiple experimental samples, such as those measured by a finger-clip pulse oximeter. SpO 2 Standard value and corresponding R ratio Linear regression fitting was performed on the data to obtain the following blood oxygen saturation estimation model: ; in, k 1 ，k 2To fit the parameters using the least squares method, this model represents the model's sensitivity to changes in the red-green ratio and the calibration offset. This model features a simple structure and high computational efficiency, making it suitable for real-time computation on low-power, resource-constrained edge devices (such as embedded platforms and microcontroller systems). Furthermore, since the input parameters are relative ratios... S R / S G This method demonstrates strong environmental adaptability under varying skin tones and light intensities. In practical applications, the system continuously calculates R for each time period. ratio It also outputs the corresponding blood oxygen saturation in real time. SpO 2 Value. When the monitored value falls below a certain set threshold, the system can trigger an early warning mechanism and record the data.

[0035] Step 4: Blood pressure measurement; Step 4.1: Extract the main frequency from the normalized RGB signal. f Combined with standardized fluctuation characteristics S R , S G Constructing a multiple regression model for blood pressure: The systolic blood pressure (SBP) and diastolic blood pressure (DBP) are estimated in real time using a multiple linear regression model. The specific process is as follows: After completing the RGB signal standardization and fluctuation index extraction, this invention further extracts frequency feature values ​​from the standardized signal sequence. f As a representative parameter of heart rate-related information, the parameter f represents the frequency of the main color change per unit time, in Hz, and can be obtained from the standardized green channel signal by Fast Fourier Transform (FFT) or power spectrum analysis. The specific extraction steps include: (1) For the standardized green channel signal G norm (t) Perform bandpass filtering (e.g., 0.6-3Hz) to eliminate high-frequency noise and low-frequency drift; (2) Calculate the power spectrum of the signal within a fixed window (e.g., 5 seconds); (3) Take the frequency corresponding to the peak frequency as the signal. f This represents the current dominant heart rate frequency. The system will extract the red channel volatility from step 2. S R Green channel volatility S G With main frequency f Both are used as input features in the regression model to establish estimation models for systolic blood pressure (SBP) and diastolic blood pressure (DBP), as shown in the following expressions: in, 1 to 6 represents the regression weight coefficient, and b1 and b2 are the bias terms. All the above coefficients were obtained by least squares regression fitting using a large amount of sample data collected synchronously with a reference medical device (such as an electronic blood pressure monitor).

[0036] Compared to traditional methods for estimating blood pressure based on pulse wave transit time (PTT), the model constructed in this invention requires no additional sensors and is based solely on three features of visual light signals, offering the following advantages: (1) The model has a simple structure and low feature dimension, which facilitates fast computation; (2) It can be embedded and deployed on edge hardware platforms, such as Raspberry Pi, Jetson Nano, etc.; (3) It has good adaptability to different skin colors and lighting environments; (4) Experiments show that the blood pressure estimation error can be controlled within ±5 mmHg in static scenarios.

[0037] Step 4.2: Calculate and output the real-time systolic blood pressure (SBP) and diastolic blood pressure (DBP) based on the model; The system dynamically updates the SBP and DBP values ​​calculated for each time period (e.g., 5 seconds) and uses them as input parameters for subsequent early warning modules, enabling non-contact visual monitoring of continuous blood pressure changes.

[0038] Step 5: Result Display and Anomaly Warning; Step 5.1: Set the normal threshold range for SpO2, SBP, and DBP. When the result exceeds the threshold, trigger a prompt and data storage. Step 5.2: Conduct long-term trend analysis and remote health early warning.

[0039] The system visualizes the estimation results and provides real-time health status alerts. The specific process is as follows: After estimating blood oxygen saturation (SpO2), systolic blood pressure (SBP), and diastolic blood pressure (DBP), the system immediately presents the results on the user interface for users to view their health status and track trends. The results display module can be implemented using a graphical user interface (GUI) framework, including but not limited to desktop interfaces such as Tkinter, embedded screen visualization systems, or web-based remote browsing interfaces. In terms of interface display, the system displays SpO2, SBP, and DBP values ​​in real-time, accompanied by color-coded status indicators, such as green for normal, orange for borderline, and red for abnormal, enhancing the user's intuitive perception of changes in health status.

[0040] To further enhance system interactivity, the system supports the visualization of data through charts, such as drawing historical trend curves, fluctuation bar charts, or heat maps of change rates.

[0041] To enable rapid identification and response to health anomalies, the system includes a built-in module for determining normal range thresholds for physiological indicators. Users can choose to use the system's default thresholds or customize them according to their individual circumstances. The system's default reference values ​​are as follows: Blood oxygen saturation SpO2: The normal range is ≥ 95%; if it is below 94%, an alarm will be triggered. Systolic blood pressure (SBP): Normal range is 90–140 mmHg; a warning will be issued if it exceeds this range. Diastolic blood pressure (DBP): The normal range is 60–90 mmHg; a warning will be issued if it exceeds this range.

[0042] When the estimated value deviates from the normal range, the system immediately activates the abnormal warning mechanism, triggering visual prompts such as interface flashing and icon changes, sound alarms such as beeping and voice prompts, and highlighting the abnormal indicators and current values ​​on the interface.

[0043] In addition, the system can be equipped with a linkage module to send abnormal events to family members' terminals, doctors' platforms or remote health centers via wireless modules such as Wi-Fi, Bluetooth, 4G, and 5G modules, enabling remote response.

[0044] All estimated data can be automatically stored as local files, such as in CSV format, or uploaded synchronously to a cloud server. The system records fields such as timestamps, estimated values, anomaly markers, and user IDs, which are used for subsequent health trend analysis, anomaly pattern modeling, or personalized threshold optimization.

[0045] Through this step, the present invention can not only achieve non-contact, real-time monitoring of physiological parameters, but also provide users with continuous and reliable health protection through data visualization and multi-dimensional early warning mechanisms. It is especially suitable for high-risk scenarios such as elderly monitoring, driving safety monitoring, and remote management of chronic diseases.

[0046] Based on the above steps, the present invention conducted corresponding experiments: Fifty test subjects of different ages, genders and skin colors were divided into groups and monitored in real time under two conditions: bright (300 lx) and dim (50 lx) ambient light; and under two conditions: still and slight facial movement.

[0047] The final experimental results are summarized as follows: Mean blood oxygenation error: ±1.5%; Systolic blood pressure error: ±5 mmHg; Diastolic blood pressure error: ±4 mmHg.

[0048] The above describes specific embodiments of the present invention and the technical principles employed. Any modifications or equivalent transformations based on the technical solutions of the present invention should be included within the protection scope of the present invention.

Claims

1. A non-contact real-time monitoring method for blood oxygen and blood pressure, characterized in that, Includes the following steps: Step 1: Facial video capture and ROI region extraction; Step 1.1: Capture facial image sequences in real time using a regular RGB camera; Step 1.2: Apply the Dlib / Mediapipe algorithm to achieve real-time face detection and lock the forehead or cheek area as the region of interest (ROI). Step 1.3: After processing the skin tone mask using the YCbCr color space, output the real-time video data of the ROI region; Step 2: RGB signal extraction and standardization calculation; Step 2.1: Calculate the mean of the RGB channels for each frame in the ROI region to form time series signals R(t), G(t), and B(t): ， ， ； Where R(t), G(t), and B(t) represent the average pixel values ​​of the red, green, and blue channels within the ROI region in the t-th frame image, respectively; Step 2.2: Calculate the standardized fluctuation characteristics of the signal. , It outputs a standardized RGB signal; ， ； Where std(R) represents the standard deviation of the red channel signal, and mean(R) represents the average value of the red channel signal; std(G) represents the standard deviation of the green channel signal, and mean(G) represents the average value of the green channel signal; Step 3: Blood oxygen saturation test; Step 3.1: Calculate the red-green light fluctuation ratio. , ; Step 3.2: Based on the experimental calibration data, use a linear regression model to estimate blood oxygen saturation. SpO 2 ; ； in, k 1 ，k 2 These are the parameters for the experimental fitting. Step 4: Blood pressure measurement; Step 4.1: Extract the main frequency from the normalized RGB signal. f Combined with standardized fluctuation characteristics S R , S G Constructing a multiple regression model for blood pressure: Systolic blood pressure ; diastolic blood pressure ; in, These are the experimental calibration coefficients; Step 4.2: Calculate and output the real-time systolic blood pressure (SBP) and diastolic blood pressure (DBP) based on the model; Step 5: Result Display and Anomaly Warning; Step 5.1: Set the normal threshold range for SpO2, SBP, and DBP. When the result exceeds the threshold, trigger a prompt and data storage. Step 5.2: Conduct long-term trend analysis and remote health early warning.

2. The non-contact real-time blood oxygen and blood pressure monitoring method according to claim 1, characterized in that: The ordinary RGB camera mentioned in step 1.1 is a visible light camera capable of providing color video images, including built-in cameras in laptops, external USB cameras, Raspberry Pi camera modules, dashcams, and front-facing cameras in smartphones; the image acquisition frame rate is not less than 20 frames / second, and the resolution is not less than 640×480.

3. The non-contact real-time blood oxygen and blood pressure monitoring method according to claim 1, characterized in that: The determination criteria for skin color masking in step 1.3 are that the Cb component is in [77, 127] and the Cr component is in [133, 173]. The effectiveness of the facial ROI is cross-validated in this region.

4. The non-contact real-time blood oxygen and blood pressure monitoring method according to claim 2, characterized in that: The standardized fluctuation features described in step 2.2 are dynamically calculated using a sliding window mechanism, with a window length of 5-10 seconds of continuous frames, and a mean-mode filter is applied to suppress noise.

5. The non-contact real-time blood oxygen and blood pressure monitoring method according to claim 1, characterized in that: Step 4.1 Main frequency f Obtained from the standardized green channel signal, the specific extraction steps are as follows: First, the standardized green channel signal G... norm (t) Perform a 0.6-3Hz bandpass filter; then calculate the power spectrum of the signal within a 5-second window; finally, take the frequency corresponding to the maximum spectral peak as... f .

6. A non-contact real-time blood oxygen and blood pressure monitoring system implementing any one of claims 1-5, characterized in that: It includes a standard RGB camera, an embedded processor, an ROI extraction module, a standardized fluctuation feature calculation module, a blood oxygen and blood pressure regression model calculation module, and a result display and early warning module; the result display and early warning module supports the graphical visualization of data, including drawing historical trend curves, fluctuation bar charts, or change rate heatmaps; Users can choose to use the system's default threshold or set a custom threshold based on their individual circumstances. An alarm will be triggered when the threshold is exceeded.

Images