Multi-sensor data fusion device based on Kalman filter

Abstract

The present invention provides a multi-sensor data fusion device based on Kalman filtering, comprising: a data acquisition system, an edge data processing system, a network transmission system, and an upper-layer application system; the input of the data acquisition system is connected to multiple sensors, the output of the data acquisition system is connected to the input of the edge data processing system, the output of the edge data processing system is connected to the input of the network transmission system, and the output of the network transmission system is connected to the input of the upper-layer application system. The present invention can effectively avoid the problem of systematic error and measurement distortion in multi-sensor data, and conveniently achieve multi-sensor data fusion.

Description

Technical Field

[0001] The present invention relates to the technical field of data processing of electronic equipment, and in particular to a multi-sensor data fusion device based on Kalman filtering. Background Art

[0002] With the development of IoT technology, the variety and number of IoT devices are constantly increasing, and correspondingly, the amount of data requiring analysis is also growing exponentially. Furthermore, the inherent characteristics of the IoT, namely heterogeneous networks and data fusion across multiple systems, present various challenges. The timely extraction of hidden information and valid data poses significant challenges to data processing. Currently, to mitigate the measurement errors of individual IoT sensors, multi-sensor measurements are typically taken and then averaged to reduce the measurement error. However, this simple averaging approach is only effective for sensors with random errors and has limited effectiveness in suppressing systematic errors.

[0003] Patent document CN109655825B discloses a data processing method, apparatus, and multi-sensor fusion method for multi-sensor fusion, comprising: acquiring image data of a target object and at least one set of detection data groups; the image data is detected by an image acquisition sensor, and the detection data groups are detected by other sensors; the other sensors are at least one of the following: microwave radar, ultrasonic radar, lidar, infrared sensor, and terahertz imaging sensor; the image data is used to represent the target image captured by the image acquisition sensor using at least one pixel data matrix; different detection data groups are detection data of different detection dimensions; a multi-dimensional matrix structure is formed; wherein: the multi-dimensional matrix structure includes multiple matrix layers distributed longitudinally, the multiple matrix layers including at least one pixel matrix layer and at least one sensor matrix layer, each pixel matrix layer corresponding to a pixel data matrix, each sensor matrix layer representing a set of detection data groups, the detection data elements in the detection data groups longitudinally corresponding to pixel elements in the pixel matrix layer; and the values ​​of the detection data elements are determined based on the detection data assignments. However, this patent document still has the drawback of limited effectiveness in suppressing systematic errors. Summary of the Invention

[0004] In view of the defects in the prior art, the object of the present invention is to provide a multi-sensor data fusion device based on Kalman filtering.

[0005] According to the present invention, a multi-sensor data fusion device based on Kalman filtering includes: a data acquisition system, an edge data processing system, a network transmission system and an upper-layer application system;

[0006] The input end of the data acquisition system is connected to multiple sensors, the output end of the data acquisition system is connected to the input end of the edge data processing system, the output end of the edge data processing system is connected to the input end of the network transmission system, and the output end of the network transmission system is connected to the input end of the upper-layer application system.

[0007] Preferably, the data acquisition system includes a data temporary storage module, a data amplifier module, an A / D module, a first power supply module and a communication module;

[0008] The input end of the data storage module is connected to the multi-sensor, the output end of the data storage module is connected to the input end of the data amplifier module, the output end of the data amplifier module is connected to the first input end of the A / D module, the first power supply module is connected to the second input end of the A / D module, the output end of the A / D module is connected to the communication module, and the output end of the communication module is connected to the input end of the edge data processing system.

[0009] Preferably, the multiple sensors collect the same variable information of the environment and transmit it to the data temporary storage module, and store the signals of different sensors separately in the form of a grid.

[0010] Preferably, the data temporary storage module is capable of storing time-series data streams, arranging and storing uploaded data values ​​from different sensors according to timestamps;

[0011] The temporary storage time of the data temporary storage module is greater than or equal to the maximum upload time interval of the upload signal.

[0012] Preferably, the data amplifier module processes the data in the register into an analog signal, and the process is that the analog signals in the register are read sequentially according to the timestamp sequence to the data amplifier module to amplify the analog signals.

[0013] Preferably, the edge data processing system includes a second power supply module, a local server and a local router;

[0014] The input end of the local server is connected to the output end of the communication module, the output end of the second power supply module is connected to the input end of the local server, the output end of the local server is connected to the input end of the local router, and the output end of the local router is connected to the input end of the network transmission system.

[0015] Preferably, the network transmission system includes a Wifi module, a LoRa module, a ZigBee module and a 5G module for data transmission;

[0016] The local router transmits the IoT multi-sensor data to the upper-layer application system in a time series through the Wifi module, the LoRa module, the ZigBee module and the 5G module.

[0017] Preferably, the upper-layer application system includes a computer, a tablet and a mobile phone.

[0018] Preferably, the local server preloads a Kalman filter algorithm, performs Kalman filter processing on the data of the multiple sensors, and stores a portion of the result data locally.

[0019] Preferably, the Kalman filter algorithm is:

[0020] The first step is to establish the discrete state equation of the system. From time k-1 to time k, the system state prediction equation is:

[0021] X k =AX k-1 +Bu k +w k

[0022] The system state observation equation is:

[0023] Z k =HX k +v k

[0024] Among them, A is the state transfer matrix; u k is the system input vector; B is the input gain matrix; w k is the process noise that obeys the normal distribution, with mean 0 and covariance matrix Q; H is the measurement matrix; v k The measurement noise obeys the normal distribution, with mean 0 and covariance matrix O;

[0025] In the second step, the prediction phase, the Kalman filter uses the estimate of the previous state to make a prediction of the current state:

[0026]

[0027] P' k =AP k-1 A T +Q

[0028] in, is the predicted value, is the Kalman estimate, P' k is the prediction error covariance matrix, P k is the Kalman estimation error covariance matrix;

[0029] In the third step, the correction phase, the Kalman filter uses the observation value of the current state to correct the predicted value obtained in the prediction phase to obtain a new estimated value that is closer to the true value:

[0030]

[0031] K k =P k 'H T (HP k H T +O) -1

[0032]

[0033] in, is the measurement margin, K k is the Kalman gain;

[0034] The fourth step is to update the covariance estimate:

[0035] P k =(1-K k H)P k '.

[0036] Compared with the prior art, the present invention has the following beneficial effects:

[0037] 1. The present invention effectively avoids the problem of systematic error and measurement distortion in multi-sensor data, conveniently realizes multi-sensor data fusion, and can more objectively reflect the state index of the physical space. At the same time, when a sensor fails or the data is abnormal, it ensures the normal operation of the system and increases the robustness of the Internet of Things monitoring system.

[0038] 2. The real environmental data melted by this invention can help customers understand the real environmental conditions in real time, and can also help financial institutions obtain risk information;

[0039] 3. The present invention can realize data fusion of multiple data sensors through the designed algorithm and device, thereby more accurately realizing real-time information monitoring of the measured physical environment, avoiding data distortion caused by a single sensor, and reducing system measurement errors. BRIEF DESCRIPTION OF THE DRAWINGS

[0040] Other features, objects and advantages of the present invention will become more apparent upon reading the detailed description of non-limiting embodiments with reference to the following drawings:

[0041] Figure 1 Schematic diagram of the structure of the multi-sensor data fusion device based on Kalman filtering of the present invention. DETAILED DESCRIPTION

[0042] The present invention will be described in detail below with reference to specific embodiments. The following examples will help those skilled in the art to further understand the present invention, but are not intended to limit the present invention in any form. It should be noted that, for those skilled in the art, several changes and improvements can be made without departing from the scope of the present invention. These all fall within the scope of protection of the present invention.

[0043] Example 1:

[0044] like Figure 1 As shown, this embodiment provides a multi-sensor data fusion device based on Kalman filtering, including: a data acquisition system, an edge data processing system, a network transmission system and an upper-layer application system. The input end of the data acquisition system is connected to multiple sensors, the output end of the data acquisition system is connected to the input end of the edge data processing system, the output end of the edge data processing system is connected to the input end of the network transmission system, and the output end of the network transmission system is connected to the input end of the upper-layer application system.

[0045] The edge data processing system includes a second power supply module, a local server and a local router. The input end of the local server is connected to the output end of the communication module, the output end of the second power supply module is connected to the input end of the local server, the output end of the local server is connected to the input end of the local router, and the output end of the local router is connected to the input end of the network transmission system.

[0046] The local server preloads the Kalman filter algorithm, performs Kalman filter processing on the data from multiple sensors, and stores part of the result data locally. The Kalman filter algorithm is:

[0047] The first step is to establish the discrete state equation of the system. From time k-1 to time k, the system state prediction equation is:

[0048] X k =AX k-1 +Bu k +w k

[0049] The system state observation equation is:

[0050] Z k =HX k +v k

[0051] Among them, A is the state transfer matrix; u k is the system input vector; B is the input gain matrix; w k is the process noise that obeys the normal distribution, with mean 0 and covariance matrix Q; H is the measurement matrix; v k The measurement noise obeys the normal distribution, with mean 0 and covariance matrix O;

[0052] In the second step, the prediction phase, the Kalman filter uses the estimate of the previous state to make a prediction of the current state:

[0053]

[0054] P' k =AP k-1 A T +Q

[0055] in, is the predicted value, is the Kalman estimate, P' k is the prediction error covariance matrix, P k is the Kalman estimation error covariance matrix;

[0056] In the third step, the correction phase, the Kalman filter uses the observation value of the current state to correct the predicted value obtained in the prediction phase to obtain a new estimated value that is closer to the true value:

[0057]

[0058] K k =P k 'H T (HP k H T +O) -1

[0059]

[0060] in, is the measurement margin, K k is the Kalman gain;

[0061] The fourth step is to update the covariance estimate:

[0062] P k =(1-K k H)P k '.

[0063] The data acquisition system includes a data storage module, a data amplifier module, an A / D module, a first power supply module and a communication module. The input end of the data storage module is connected to the multi-sensor, the output end of the data storage module is connected to the input end of the data amplifier module, the output end of the data amplifier module is connected to the first input end of the A / D module, the first power supply module is connected to the second input end of the A / D module, the output end of the A / D module is connected to the communication module, and the output end of the communication module is connected to the input end of the edge data processing system.

[0064] Multiple sensors collect information about the same environmental variable and transmit it to the data storage module, which stores the signals from each sensor separately in a grid format. The data storage module stores time-series data streams, arranging and storing uploaded data values ​​from different sensors according to timestamps. The data storage module's storage duration is equal to or greater than the maximum upload interval for the uploaded signal. The data amplifier module processes the data in the register into analog signals. The analog signals are read sequentially from the register in timestamp order and then amplified by the data amplifier module.

[0065] The network transmission system includes Wi-Fi, LoRa, ZigBee, and 5G modules for data transmission. Local routing uses these modules to transmit IoT multi-sensor data in a time series to upper-layer application systems, including computers, tablets, and mobile phones.

[0066] Example 2:

[0067] Those skilled in the art may understand this embodiment as a more specific description of Embodiment 1.

[0068] like Figure 1 As shown, this embodiment provides a multi-sensor data fusion method and device based on Kalman filtering. The designed algorithm and device can realize data fusion of multiple data sensors, thereby more accurately realizing real-time information monitoring of the measured physical environment, avoiding data distortion caused by a single sensor, and reducing system measurement errors.

[0069] This embodiment provides a multi-sensor data fusion method and device based on Kalman filtering, including three parts: a data acquisition system, an edge data processing system, a network transmission system, and an upper-layer application system. The data acquisition system collects ambient data through multiple sensors, and the communication system transmits the data collected by multiple sensors to the application layer through networks such as LoRa, Wifi, and ZigBee. The application system filters and fuses the raw data by constructing a multi-sensor data fusion algorithm based on Kalman filtering, and stores the data to facilitate the use of the data in subsequent application scenarios. This solution mainly includes the following steps:

[0070] The data acquisition system includes a multi-sensor data temporary storage module, a data amplifier module, an A / D module, a power supply module, and a communication module.

[0071] Furthermore, the power module is the system's power supply, providing the required energy for other modules and the main control chip. This module utilizes the W800 power chip, which boasts low power consumption and the ability to switch between multiple modes. In standby mode, it consumes only 10μA and also features current slope compensation, undervoltage, and overvoltage protection. Compared to other linear regulated power supplies, this power module offers higher energy efficiency, thereby improving the overall battery life of the sensor device.

[0072] The multiple sensors collect the same variable information of the environment and transmit it to the data temporary storage module, and store the different sensor signals separately in the form of a grid.

[0073] Furthermore, the temporary register has the ability to store time-series data streams, arranging and storing data values ​​uploaded by different sensors according to timestamps, and the temporary storage time of the module must meet the maximum upload time interval of the upload signal.

[0074] The data amplifier module processes the data in the temporary register into an analog signal. The process is that the analog signals in the temporary register are read into the data amplifier module in sequence according to the timestamp order to amplify the analog signals.

[0075] The A / D module is a digital-to-analog converter, which converts analog data from all sensors into digital signals in sequence through time stamps.

[0076] Furthermore, the digital-to-analog sensor utilizes a TLC0831 successive approximation A / D, which includes an internal sampling data comparator responsible for performing differential processing and comparison on the input analog signal before converting it into a digital signal. Furthermore, its analog voltage differential input method helps suppress common-mode signals and reduce or eliminate conversion offset errors. The adjustable voltage reference input allows for the conversion of small-scale analog voltage signals into digital signals with higher resolution.

[0077] The communication module forwards the converted IoT multi-sensor digital signals to the edge data processing system.

[0078] The edge processing system includes: a power module, a local server, and a local router.

[0079] Furthermore, the edge processing system power module adopts PC / 104POWER, which has the characteristics of small size, high reliability, and wide input voltage adaptability, making it suitable for IoT application scenarios.

[0080] Furthermore, the local server uses PMC-1618, which uses a Cortex A53 quad-core processor, a 1.5GHz main frequency, and 128G of memory. It can meet the real-time access and processing of IoT multi-sensor data. Its built-in Docker environment completely decouples software and hardware development, which is conducive to the subsequent development and loading of the Kalman filter algorithm.

[0081] The local server will receive the multi-sensor collected data transmitted from the data collection system in time sequence.

[0082] The local server preloads the Kalman filter algorithm, performs Kalman filter processing on the multi-sensor data, and stores part of the result data locally. The storage time is set to 7 days in case of data transmission loss or later data retrieval.

[0083] Furthermore, the Kalman filter algorithm is designed as follows:

[0084] The first step is to establish the discrete state equation of the system. From time k-1 to time k, the system state prediction equation is:

[0085] X k =AX k-1 +Bu k +w k (1)

[0086] The system state observation equation,

[0087] Z k =HX k +v k (2)

[0088] Among them, A is the state transfer matrix, u k is the system input vector, B is the input gain matrix, w k is the process noise that obeys the normal distribution (mean is 0, covariance matrix is ​​Q), H is the measurement matrix, v k The measurement noise follows a normal distribution (mean 0, covariance matrix O).

[0089] The second step is the prediction phase, in which the Kalman filter uses the estimate of the previous state to make a prediction of the current state.

[0090]

[0091] P' k =AP k-1 A T +Q (4)

[0092] in, is the predicted value, is the Kalman estimate, P' k is the prediction error covariance matrix, P k is the Kalman estimation error covariance matrix.

[0093] The third step is the correction phase, in which the Kalman filter uses the observation value of the current state to correct the predicted value obtained in the prediction phase to obtain a new estimated value that is closer to the true value.

[0094]

[0095] K k =P' k H T (HP k H T +O) -1 (6)

[0096]

[0097] in, is the measurement margin, K k is the Kalman gain.

[0098] Step 4: Update the covariance estimate

[0099] P k =(1-K k H)P k ' (8)

[0100] After the IoT multi-sensor data is processed by the Kalman filter fusion loaded on the local server, part of it is forwarded to the local storage, and part of it is forwarded to the local router to prepare for subsequent data transmission.

[0101] The local routing terminal transmits the processed IoT multi-sensor data to the platform layer in time series through wireless networks such as Wifi, 5G, and LoRa for data application analysis and data query.

[0102] The present invention can effectively avoid the problem of systematic error measurement value distortion in multi-sensor data and conveniently realize multi-sensor data fusion.

[0103] The above describes specific embodiments of the present invention. It should be understood that the present invention is not limited to the specific embodiments described above, and those skilled in the art may make various changes or modifications within the scope of the claims, which do not affect the essence of the present invention. The embodiments of this application and the features in the embodiments may be combined with each other in any manner unless there is a conflict.

Claims

1. A multi-sensor data fusion device based on Kalman filtering, characterized in that: include: Data acquisition system, edge data processing system, network transmission system and upper-layer application system; The input end of the data acquisition system is connected to multiple sensors, the output end of the data acquisition system is connected to the input end of the edge data processing system, the output end of the edge data processing system is connected to the input end of the network transmission system, and the output end of the network transmission system is connected to the input end of the upper-layer application system; The edge data processing system includes a second power supply module, a local server and a local router; The input end of the local server is connected to the output end of the communication module, the output end of the second power supply module is connected to the input end of the local server, the output end of the local server is connected to the input end of the local router, and the output end of the local router is connected to the input end of the network transmission system; The local server preloads a Kalman filter algorithm, performs Kalman filter processing on the data of the multi-sensor, and locally stores a portion of the result data; The Kalman filter algorithm is: The first step is to establish the discrete state equation of the system. From time k-1 to time k, the system state prediction equation is: X k =AX k-1 +Bu k +w k The system state observation equation is: Z k =HX k +v k Among them, A is the state transfer matrix; u k is the system input vector; B is the input gain matrix; w k is the process noise that obeys the normal distribution, with mean 0 and covariance matrix Q; H is the measurement matrix; v k The measurement noise obeys the normal distribution, with mean 0 and covariance matrix O; In the second step, the prediction phase, the Kalman filter uses the estimate of the previous state to make a prediction of the current state: P' k =AP k-1 From T +Q in, is the predicted value, is the Kalman estimate, P ' k is the prediction error covariance matrix, P k is the Kalman estimation error covariance matrix; In the third step, the correction phase, the Kalman filter uses the observation value of the current state to correct the predicted value obtained in the prediction phase to obtain a new estimated value that is closer to the true value: K k =P k ' H T (HP k H T +O) -1 in, is the measurement margin, K k is the Kalman gain; The fourth step is to update the covariance estimate: P k =(1-K k H)P k '。 2. The multi-sensor data fusion device based on Kalman filtering according to claim 1, characterized in that: The data acquisition system includes a data temporary storage module, a data amplifier module, an A / D module, a first power supply module and a communication module; The input end of the data storage module is connected to the multi-sensor, the output end of the data storage module is connected to the input end of the data amplifier module, the output end of the data amplifier module is connected to the first input end of the A / D module, the first power supply module is connected to the second input end of the A / D module, the output end of the A / D module is connected to the communication module, and the output end of the communication module is connected to the input end of the edge data processing system.

3. The multi-sensor data fusion device based on Kalman filtering according to claim 2, characterized in that: The multiple sensors collect the same variable information of the environment and transmit it to the data temporary storage module, and store the different sensor signals separately in the form of a grid.

4. The multi-sensor data fusion device based on Kalman filtering according to claim 3, characterized in that: The data temporary storage module can store time-series data streams, arranging and storing uploaded data values ​​from different sensors according to timestamps; The temporary storage time of the data temporary storage module is greater than or equal to the maximum upload time interval of the upload signal.

5. The multi-sensor data fusion device based on Kalman filtering according to claim 4, characterized in that: The data amplifier module processes the data in the temporary register into an analog signal. The process is that the analog signals in the temporary register are read into the data amplifier module in sequence according to the timestamp order to amplify the analog signals.

6. The multi-sensor data fusion device based on Kalman filtering according to claim 1, characterized in that: The network transmission system includes a Wifi module, a LoRa module, a ZigBee module and a 5G module for data transmission; The local router transmits the IoT multi-sensor data to the upper-layer application system in a time series through the Wifi module, the LoRa module, the ZigBee module and the 5G module.

7. The multi-sensor data fusion device based on Kalman filtering according to claim 6, characterized in that: The upper-layer application system includes computers, tablets and mobile phones.

Images