阅读说明：本技术 一种基于多传感器检测和gnss定位的溺水检测方法以及设备 (Drowning detection method and equipment based on multi-sensor detection and GNSS positioning ) 是由 纪刚 臧强 李彦 吴庭永 于 2019-11-08 设计创作，主要内容包括：本发明属于溺水施救领域,具体涉及一种基于多传感器检测和GNSS定位的溺水检测方法以及设备,具体检测步骤如下,将海拔高度写入flash；溺水检测便携设备每隔一定时间,使用GNSS模块获取当前坐标,判定当前位置与上一次位置距离,并且与数据表中的坐标数值对比,当前位置在新的单元区域,则重新设置水深传感器基准值；如果位置依旧在上次单位区域内,则不调整基准值；再基于加速度计进行溺水行为识别,识别步骤包括加速度计采集数据预处理、数据加窗、行为特征提取、基于决策树的动作识别；最后进行溺水行为判定。能够及时监测到溺水行为并进行准确预警以报告溺水人员的位置,有效提高了施救的效率,安全性好。(The invention belongs to the field of drowning rescue, and particularly relates to a drowning detection method and equipment based on multi-sensor detection and GNSS positioning, wherein the detection steps are as follows, the altitude is written into flash; the portable equipment for detecting drowning uses a GNSS module to obtain a current coordinate at regular intervals, the distance between the current position and the last position is judged, the current position is compared with a coordinate value in a data table, and the reference value of the water depth sensor is reset when the current position is in a new unit area; if the position is still in the last unit area, the reference value is not adjusted; performing drowning behavior recognition based on an accelerometer, wherein the recognition step comprises the steps of accelerometer data acquisition preprocessing, data windowing, behavior feature extraction and action recognition based on a decision tree; and finally, carrying out drowning behavior judgment. The drowning behavior can be monitored in time, and the drowning behavior can be accurately early warned to report the position of drowning personnel, so that the rescuing efficiency is effectively improved, and the safety is good.)

1. A drowning detection method and equipment based on multi-sensor detection and GNSS positioning are characterized in that the structure of the drowning detection equipment based on multi-sensor detection and GNSS positioning is as follows:

the accelerometer, the water depth measuring sensor and the heart rate measuring sensor are respectively in communication connection with the MCU through the IIC and used for transmitting collected data to the MCU, the Beidou/GPS module and the SIM card module are respectively in communication with the MCU through the UART, the Beidou/GPS module is used for acquiring the position of a person, and the SIM card is used for sending alarm information; the external flash is in communication connection with the MCU through the SPI and is used for storing coordinate information and altitude information; the computing power of chips on equipment including the bracelet and the tag is limited, and a complex computing method cannot be used, different geographical position altitudes are stored in a flash in the equipment, and altitude information is read from the flash according to a positioning result of a GNSS module when the equipment is used; the LED lamp is in communication connection with the MCU through the GPIO to send an alarm signal, and the MCU is controlled through the GPIO

And the LED lamp is turned on and off.

2. The drowning detection method and the drowning detection equipment based on multi-sensor detection and GNSS positioning according to claim 1 are characterized in that when drowning detection is carried out, the concrete operation steps are carried out as follows:

because the higher the altitude is, the lower the air pressure is, the different altitudes and air pressures are caused, when the water pressure sensor is used, the reference 0 value of the water surface of the swimming pool can change, and needs to be adjusted correspondingly according to the altitude, so that different geographical position altitudes need to be stored in a flash;

step 1, writing in flash at altitude

S1, an altitude data table needs to be made, and a domestic area is divided into a plurality of areas of 10 x 10km by taking 10km as a unit;

s2, numbering the divided areas according to a certain sequence (from south to north, from east to west or from north to south and from west to east);

s3, inquiring and recording the average altitude of each 10 x 10km area, and particularly, appropriately inquiring and reducing the area range in areas with dense population and severe terrain change;

s4, writing the inquired altitude data into the flash in sequence, wherein the altitude data does not exceed 10000, and the altitude of each area is written into the flash in a 2-byte or 4-byte mode so as to be convenient for inquiry;

step 2, automatically adjusting the reference value of the water depth sensor based on the position of the monitored person

The portable drowning detection equipment acquires the current coordinate by using the GNSS module at regular intervals, and although the indoor positioning error is large, the final result is not influenced by the error within fifty meters;

s1, judging the distance between the current position and the last position, comparing the distance with the coordinate value in the data table, and resetting the reference value of the water depth sensor when the current position is in a new unit area; if the position is still in the last unit area, the reference value is not adjusted;

step 3, drowning behavior recognition is carried out based on accelerometer

S1 preprocessing accelerometer acquisition data

The accelerometer can measure the motion acceleration and the gravity acceleration signals of a human body, but the measured data of the accelerometer comprises various interference signals and noise signals, and the interference and the noise can bring a lot of negative effects on drowning behavior identification, so that the data measured by the accelerometer needs to be subjected to preprocessing operations including noise filtering, smoothing and windowing;

the data sampled by the accelerometer is processed using a chebyshev low-pass filter as shown in equation (1),

filtering the data sampled by the accelerometer by using the filter obtained by the calculation for further analysis;

s2, data windowing

Because acceleration data obtained by direct measurement is presented in a data stream form in a time domain, the acceleration data is not suitable for direct feature extraction, and window adding pretreatment is usually carried out before feature extraction is carried out on an acceleration signal; in the invention, the data acquisition frequency of the accelerometer is 50Hz, the sampling window is set to be 256 points, and the adjacent windows are overlapped by 50 percent;

s3, behavior feature extraction

Because the windowed accelerometer measurement data in the S2 cannot be directly used for judging the motion state of the human body and cannot be directly identified by the classifier, the windowed accelerometer measurement data in the S2 needs to be subjected to feature extraction, and the standard deviation, the mean value, the peak interval and the peak and trough are selected to form a feature set;

s4 action recognition based on decision tree

The characteristic set obtained by behavior characteristic extraction is used for designing attributes of internal nodes in a decision tree, the calculation amount of a decision tree classification algorithm is relatively small, the model is simple, but the identification precision is high, the decision tree classification method adopts a top-down recursion mode, the attributes of the nodes are compared internally, a top-down path is obtained according to the difference of the attributes, and classification is obtained at leaf nodes;

firstly, generating a tree by using an ID3 algorithm, and pruning the tree according to results of cross validation and test set validation to obtain a decision tree;

step 4, drowning behavior judgment

S1, after the drowning detection equipment for multi-sensor detection and GNSS positioning starts to work, reading a GNSS positioning result and determining the current positioning position;

s2, calculating the sequence number of the current area;

s3, searching the altitude of the current area, and modifying the reference value;

s4, reading accelerometer data, and analyzing the current drowning behavior according to an accelerometer-based drowning behavior recognition algorithm

Whether the behavior is at rest;

s5, if the current state of motion is identified, whether the swimming is normal or drowned is further judged, if the person is in the drowned state, an alarm signal is sent to a safety person and a nursing person through an SIM card, and the LED lamp flashes and gives an alarm;

s6, if the current state of standstill is identified, determining that the interval time from the last GNSS positioning use needs to be determined, if the interval time exceeds the set time (for example, half an hour), repositioning, and determining whether the altitude reference value needs to be adjusted; if the time is less than half an hour, the adjustment is not carried out by mistake;

s7, reading water depth data, further reading heart rate data if the water depth is larger than a threshold value, judging whether the heart rate of a current person is in a drowned state, sending an alarm signal to a safety worker and a nursing worker through an SIM card if the heart rate of the current person is in the drowned state, and giving an alarm through LED and other flash frequencies; otherwise, continuing to read the accelerometer data.

3. The method and apparatus for drowning detection based on multi-sensor detection and GNSS positioning of claim 1, wherein the accelerometer comprises single but not limited to various types of MEMS-based 3-axis, 6-axis, 9-axis accelerometers.

4. The drowning detection method and apparatus based on multi-sensor detection and GNSS positioning according to claim 1, characterized in that the water depth sensor includes but is not limited to pressure-based water depth sensor.

5. The method and device for drowning detection based on multi-sensor detection and GNSS positioning as claimed in claim 2, characterized in that in the pre-processing of the accelerometer, low-pass filtering can also be used methods including but not limited to Chebyshev low-pass filtering, Butterworth low-pass filtering, FIR low-pass filtering and Kalman filtering.

6. The method and apparatus for drowning detection based on multi-sensor detection and GNSS positioning as claimed in claim 2, wherein the analysis is performed only by decision tree method, and the behavior classification method can be used, including but not limited to KNN, naive Bayes, SVM, etc., and their combination.

The technical field is as follows:

the invention belongs to the field of drowning rescue, and particularly relates to a drowning detection method and equipment based on multi-sensor detection and GNSS positioning.

Background art:

at present, swimming pools and seaside swimming are very big entertainment modes, which can achieve the purpose of entertainment and can also exercise the body through swimming, but the entertainment fitness mode also has the situations of drowning or cramping in water or other accidents, so that rescue needs to be timely carried out, due to the particularity of drowning, rescue needs to be completed in a very short time, otherwise life danger exists, the main protection measures to the accidents of swimming, drowning and the like are to arrange life-saving personnel in swimming places, the drowning person is found and timely rescued by observing or hearing a help calling mode after the swimming person drowns, but the life-saving means is usually limited by the number of life-saving personnel in the swimming pools, and the drowning person cannot be accurately found in time. Every second delay, the danger is increased by one minute for drowning personnel, so the invention seeks to design and provide a drowning detection method and equipment based on multi-sensor detection and GNSS positioning, the measurement data of an accelerometer, a water depth sensor and a heart rate sensor are used for judging whether a wearer is in a drowning state, and a GPS/Beidou/Glonass/Galileo positioning module (hereinafter referred to as GNSS positioning module) is used for positioning the position of the personnel to correct the measurement reference value of the water depth sensor, so that the measurement accuracy is improved. The safety of swimming is improved.

The invention content is as follows:

the invention aims to overcome the defects in the prior art, and provides a drowning detection method and equipment based on multi-sensor detection and GNSS positioning, which can accurately analyze and identify drowning behavior.

In order to achieve the purpose, the drowning detection method and the drowning detection equipment based on multi-sensor detection and GNSS positioning are realized by the following technical scheme: the invention relates to a method based on multi-sensor

The drowning detection method and equipment based on the GNSS positioning are realized by the following technical scheme:

wherein drowning check out test set's of multisensor detection and GNSS location structure as follows:

the accelerometer, the water depth measuring sensor and the heart rate measuring sensor are respectively in communication connection with an MCU (micro control unit) through an IIC (Inter-Integrated Circuit) for transmitting the collected data to the MCU,

the Beidou/GPS module and the SIM card module are respectively communicated with the MCU through UART (Universal Asynchronous Receiver/Transmitter), the Beidou/GPS module is used for acquiring the position of personnel, and the SIM card is used for sending alarm information;

the external flash is in communication connection with the MCU through the SPI and is used for storing coordinate information and altitude information; the computing power of chips on equipment including the bracelet and the tag is limited, and a complex computing method cannot be used, different geographical position altitudes are stored in a flash in the equipment, and altitude information is read from the flash according to a positioning result of a GNSS module when the equipment is used;

the LED lamp is in communication connection with the MCU through the GPIO and used for sending an alarm signal, and the MCU controls the LED lamp to be on or off through the GPIO.

When the drowning detection equipment based on multi-sensor detection and GNSS positioning performs drowning detection, the specific operation steps are performed as follows:

because the higher the altitude is, the lower the air pressure is, the different altitudes and air pressures are caused, when the water pressure sensor is used, the reference 0 value of the water surface of the swimming pool can change, and needs to be adjusted correspondingly according to the altitude, so that different geographical position altitudes need to be stored in a flash;

step 1, writing in flash at altitude

S1, an altitude data table needs to be made, and a domestic area is divided into a plurality of areas of 10 x 10km by taking 10km as a unit;

s2, numbering the divided areas according to a certain sequence (from south to north, from east to west or from north to south and from west to east);

s3, inquiring and recording the average altitude of each 10 x 10km area, and particularly, appropriately inquiring and reducing the area range in areas with dense population and severe terrain change;

s4, writing the inquired altitude data into the flash in sequence, wherein the altitude data does not exceed 10000, and the altitude of each area is written into the flash in a 2-byte or 4-byte mode so as to be convenient for inquiry;

step 2, automatically adjusting the reference value of the water depth sensor based on the position of the monitored person

The method comprises the steps that a GNSS module is used for obtaining current coordinates of drowning detection portable equipment at regular intervals, wherein the time interval is set according to needs and ranges from 10 minutes to 120 minutes, although indoor positioning errors are large, the final result is not influenced by errors within fifty meters;

s1, judging the distance between the current position and the last position, comparing the distance with the coordinate value in the data table, and resetting the reference value of the water depth sensor when the current position is in a new unit area; if the position is still in the last unit area, the reference value is not adjusted;

step 3, drowning behavior recognition is carried out based on accelerometer

S1 preprocessing accelerometer acquisition data

The accelerometer can measure the motion acceleration and the gravity acceleration signals of a human body, but the measured data of the accelerometer comprises various interference signals and noise signals, and the interference and the noise can bring a lot of negative effects on drowning behavior identification, so that the data measured by the accelerometer needs to be subjected to preprocessing operations including noise filtering, smoothing and windowing;

the data sampled by the accelerometer is processed using a chebyshev low-pass filter as shown in equation (1),

filtering the data sampled by the accelerometer by using the filter obtained by the calculation for further analysis;

s2, data windowing

Because acceleration data obtained by direct measurement is presented in a data stream form in a time domain, the acceleration data is not suitable for direct feature extraction, and window adding pretreatment is usually carried out before feature extraction is carried out on an acceleration signal; in the invention, the data acquisition frequency of the accelerometer is 50Hz, the sampling window is set to be 256 points, and the adjacent windows are overlapped by 50 percent;

s3, behavior feature extraction

Because the windowed accelerometer measurement data in the S2 cannot be directly used for judging the motion state of the human body and cannot be directly identified by the classifier, the windowed accelerometer measurement data in the S2 needs to be subjected to feature extraction, and the standard deviation, the mean value, the peak interval and the peak and trough are selected to form a feature set;

s4 action recognition based on decision tree

The characteristic set obtained by behavior characteristic extraction is used for designing attributes of internal nodes in a decision tree, the calculation amount of a decision tree classification algorithm is relatively small, the model is simple, but the identification precision is high, the decision tree classification method adopts a top-down recursion mode, the attributes of the nodes are compared internally, a top-down path is obtained according to the difference of the attributes, and classification is obtained at leaf nodes;

firstly, generating a tree by using an ID3 algorithm, and pruning the tree according to results of cross validation and test set validation to obtain a decision tree;

step 4, drowning behavior judgment

S1, after the drowning detection equipment for multi-sensor detection and GNSS positioning starts to work, reading a GNSS positioning result and determining the current positioning position;

s2, calculating the sequence number of the current area;

s3, searching the altitude of the current area, and modifying the reference value;

s4, reading accelerometer data, and analyzing whether the current behavior is static or not according to a drowning behavior recognition algorithm based on an accelerometer;

s5, if the current state of motion is identified, whether the swimming is normal or drowned is further judged, if the person is in the drowned state, an alarm signal is sent to a safety person and a nursing person through an SIM card, and the LED lamp flashes and gives an alarm;

s6, if the current state of standstill is identified, determining that the interval time from the last GNSS positioning use needs to be determined, if the interval time exceeds the set time (for example, half an hour), repositioning, and determining whether the altitude reference value needs to be adjusted; if the time is less than half an hour, the adjustment is not carried out by mistake;

s7, reading water depth data, further reading heart rate data if the water depth is larger than a threshold value, judging whether the heart rate of a current person is in a drowned state, sending an alarm signal to a safety worker and a nursing worker through an SIM card if the heart rate of the current person is in the drowned state, and giving an alarm through LED and other flash frequencies; otherwise, continuing to read the accelerometer data.

Further, the accelerometer in the present invention includes but is not limited to various types of 3-axis, 6-axis, and 9-axis MEMS-based accelerometers.

Further, the water depth sensor in the present invention includes, but is not limited to, a pressure-based water depth sensor.

Further, in the present invention, in the preprocessing of the accelerometer, the low-pass filtering may also be implemented by methods including, but not limited to, chebyshev low-pass filtering, butterworth low-pass filtering, FIR low-pass filtering, and kalman filtering.

Further, in the present invention, only the decision tree method is used for analysis, and the behavior classification method can be used, including but not limited to KNN, naive bayes, SVM, and the like, and combinations thereof.

Compared with the prior art, the invention has the following beneficial effects:

the air pressure is different due to the different altitude. Compared with the existing product, the method can determine the altitude of the current position according to the GNSS positioning result, and further adjust the reference value of the water depth sensor. Therefore, the accuracy of the measuring value of the water depth sensor can be improved, and the drowning behavior can be analyzed more accurately. The drowning monitoring system is simple in main body structure, ingenious in design concept, accurate in monitoring result, capable of monitoring drowning behavior in time and performing accurate early warning to report the position of a drowning person, capable of effectively improving rescuing efficiency, good in safety, friendly in application environment and wide in market prospect.

Description of the drawings:

fig. 1 is a schematic structural diagram of a drowning detection device based on sensor detection and GNSS positioning according to the present invention.

Fig. 2 is a schematic diagram illustrating a flow of writing altitude data into a flash according to the present invention.

Fig. 3 is a schematic diagram of the distance division principle according to the present invention.

Fig. 4 is a schematic diagram illustrating the flow of executing the reference value of the water depth sensor according to the present invention.

FIG. 5 is a schematic diagram illustrating a flow of decision tree action recognition according to the present invention.

Fig. 6 is a schematic diagram of a drowning behavior recognition flow principle according to the present invention.

The specific implementation mode is as follows:

the invention is further illustrated by way of example and with reference to the accompanying drawings.