Pollution source monitoring system based on unmanned aerial vehicle multi-source sensing

Abstract

The invention relates to the technical field of data processing, in particular to a pollution source monitoring system based on multi-source sensing of an unmanned aerial vehicle, which comprises a processor and a memory, and is characterized in that the processor executes a computer program of the memory; the method comprises the following steps: recording a pollution source monitoring index corresponding to a sensor of which the sampling frequency is smaller than a preset target sampling frequency as a target monitoring index, and obtaining an interpolation moment of the target monitoring index in a preset time period; dividing all interpolation moments into linear interpolation moments, continuous interpolation moments and non-continuous interpolation moments according to data change characteristics of a target monitoring index in a local time range of each interpolation moment, and then adaptively allocating an optimal interpolation method for each interpolation moment; according to the method, the re-sampling data sequence of the target monitoring indexes in the preset time period is obtained, and according to the re-sampling data sequence of each target monitoring index, monitoring and early warning of the pollution source are realized by using a multi-source data fusion algorithm, so that the monitoring accuracy of the pollution source is improved.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of data processing, and in particular to a pollution source monitoring system based on multi-source sensing of unmanned aerial vehicles. BACKGROUND

[0002] Pollution source monitoring is to timely find abnormal emissions, master emission rules, accurately trace and quantitatively analyze, support environmental risk assessment and scientific management, so as to protect environmental quality and public health. At present, with the vigorous development of unmanned aerial vehicle technology, more comprehensive sampling data can be obtained through various sensors carried by unmanned aerial vehicles, and higher pollution source monitoring accuracy and efficiency can be achieved.

[0003] When collecting data by using various sensors, due to the inherent physical and chemical response time of different sensors, the sampling interval of the sensor must be greater than the response time of the sensor, in order to obtain accurate data. For example, if a slow sensor (such as a gas sensor) is forced to sample at a high frequency, a large amount of low-quality, high-noise redundant data will be generated. Therefore, the sampling frequency of different sensors cannot be forcibly consistent, so that various pollution source data cannot be synchronized at the same time scale. At present, fixed frequency resampling + linear interpolation is usually used to time align multi-source asynchronous data, and then a multi-source data fusion algorithm (such as Kalman filtering or other state estimation methods) is used to combine wind speed, wind direction, environmental parameters and position data to construct the spatial distribution state of pollutants. Finally, based on the fusion result, a pollution diffusion model is used for pollution source positioning and emission intensity inversion to back-propagate the specific position of the pollution source, the emission rate and the influence range, and generate a monitoring report to trigger an early warning when abnormal emissions are detected.

[0004] However, various pollution source data are affected by air flow, unmanned aerial vehicle flight path and pollution source working conditions, resulting in large local variation differences in data collected by various sensors. For example, in areas where air flow is stable and pollutant diffusion is sufficient, gas concentration or particulate matter concentration changes slowly, forming a stable interval, while pollution sources often have instantaneous emission events. When flying into areas with rapidly changing wind speed and direction or close to the edge of a pollution source with instantaneous emissions, pollutant concentration may change rapidly, forming a sudden peak or rapidly changing interval. If only a single interpolation method is used for interpolation, the local characteristics of data under different changing conditions cannot be considered, which may smooth out the peak or produce oscillation, resulting in distorted data after time alignment, which not only weakens the ability to capture short-term pollution events, but also may reduce the accuracy of subsequent pollution source positioning and emission intensity inversion.

[0005] Therefore, how to adaptively obtain the optimal interpolation method according to the local variation differences of various pollution source data to time align various pollution source data and improve the accuracy of pollution source monitoring has become a problem to be solved. SUMMARY

[0006] Therefore, the embodiment of the present application provides a pollution source monitoring system based on unmanned aerial vehicle multi-source sensing to solve the problem of how to adaptively obtain an optimal interpolation method according to local variation differences of various pollution source data, time-align various pollution source data, and improve the accuracy of monitoring pollution sources.

[0007] The embodiment of the present application provides a pollution source monitoring system based on unmanned aerial vehicle multi-source sensing, comprising a memory, a processor, and a computer program stored in the memory and running on the processor, characterized in that the processor implements the following steps when executing the computer program: According to the sampling frequency of the sensor corresponding to each pollution source monitoring index, a pollution source monitoring index corresponding to a sensor with a sampling frequency less than a preset target sampling frequency is marked as a target monitoring index, for any target monitoring index, according to the difference between the sampling frequency of the sensor corresponding to the target monitoring index and the preset target sampling frequency, an interpolation time of the target monitoring index within a preset time period is obtained; According to the data smoothness within the local time range of the target monitoring index at each interpolation time, all interpolation times are divided into linear interpolation times and nonlinear interpolation times, and according to the continuity and volatility within the local time range of the target monitoring index at each nonlinear interpolation time, all nonlinear interpolation times are divided into continuity interpolation times and discontinuity interpolation times; For each linear interpolation time, continuity interpolation time and discontinuity interpolation time, different optimal interpolation methods are adaptively assigned, and according to the optimal interpolation method corresponding to each interpolation time, interpolation filling is performed on each interpolation time to obtain a resampling data sequence of the target monitoring index within the preset time period, and according to the resampling data sequence of each target monitoring index, a multi-source data fusion algorithm is used to realize monitoring and early warning of the pollution source.

[0008] Preferably, according to the data smoothness within the local time range of the target monitoring index at each interpolation time, all interpolation times are divided into linear interpolation times and nonlinear interpolation times, comprising: For any interpolation time, according to the sampling frequency of the sensor corresponding to the target monitoring index, data of a preset number of sampling times before the interpolation time of the target monitoring index is obtained to obtain a left neighborhood data sequence of the interpolation time, and according to the change speed difference between each two adjacent data in the left neighborhood data sequence, a change speed difference degree of the target monitoring index within the left neighborhood time range of the interpolation time is obtained, which is denoted as a left neighborhood difference degree; Based on the data of any target monitoring index at a preset number of sampling times after any interpolation time, the rate of change difference of any target monitoring index within the right neighbor time range at any interpolation time is obtained and denoted as the right neighbor difference. The negative of the sum of the left neighborhood difference and the right neighborhood difference is used as the independent variable of an exponential function with the natural constant as the base, to obtain the data stationarity characteristic index of any target monitoring index within the local time range at any interpolation time. Obtain the data stationarity characteristic index of any target monitoring indicator within the local time range at each interpolation time. Based on the data stationarity characteristic index of any target monitoring indicator within the local time range at each interpolation time, divide all interpolation times into linear interpolation times and nonlinear interpolation times.

[0009] Preferably, the step of obtaining the rate of change difference of any target monitoring indicator within the left neighbor time range at any interpolation time based on the rate of change difference between every two adjacent data in the left neighbor data sequence includes: The rate of change of data between every two adjacent data in the left neighbor data sequence is obtained to obtain the rate of change data sequence. The first-order difference sequence of the rate of change data sequence is obtained, and the absolute value of each data in the first-order difference sequence is accumulated to obtain the rate of change difference of any target monitoring index within the left neighbor time range at any interpolation time.

[0010] Preferably, the step of dividing all interpolation times into linear interpolation times and nonlinear interpolation times based on the data stationarity characteristic index of any target monitoring index within a local time range at each interpolation time includes: The interpolation time corresponding to the data stationarity characteristic index that is greater than or equal to the preset stationarity threshold is recorded as the linear interpolation time, and the interpolation time corresponding to the data stationarity characteristic index that is less than the preset stationarity threshold is recorded as the nonlinear interpolation time.

[0011] Preferably, the step of dividing all nonlinear interpolation moments into continuous interpolation moments and discontinuous interpolation moments based on the continuity and volatility of any target monitoring index within the local time range of each nonlinear interpolation moment includes: For any nonlinear interpolation time, based on the sampling frequency of the sensor corresponding to any target monitoring index, data of any target monitoring index at a preset number of sampling times before any nonlinear interpolation time are obtained to obtain the left neighbor data sequence at any nonlinear interpolation time. Based on the data of any target monitoring index at a preset number of sampling times after any nonlinear interpolation time, the right neighbor data sequence at any nonlinear interpolation time is obtained. Based on the data difference between the left neighbor data sequence and the right neighbor data sequence, the data difference degree between the two sides of any target monitoring index at any nonlinear interpolation time is obtained. The left neighbor data sequence and the right neighbor data sequence are combined according to time sequence to obtain the local neighbor data sequence at any nonlinear interpolation time. Based on the fluctuation direction and fluctuation degree between each two adjacent data in the local neighbor data sequence, the data jump degree of any target monitoring indicator at any nonlinear interpolation time is obtained. Calculate the mean between the difference between the two sides of the data and the degree of data jump to obtain the discontinuous characteristic value of any target monitoring index within the local time range at any nonlinear interpolation time. Obtain the discontinuous characteristic value of any target monitoring index within the local time range at each nonlinear interpolation moment. Based on the discontinuous characteristic value of any target monitoring index within the local time range at each nonlinear interpolation moment, divide all nonlinear interpolation moments into continuous interpolation moments and discontinuous interpolation moments.

[0012] Preferably, obtaining the degree of data difference between the left and right neighboring data sequences for any target monitoring indicator at any nonlinear interpolation time based on the data difference between the left and right neighboring data sequences includes: The average value of all data in the left neighbor data sequence is obtained to obtain the left average value of any target monitoring index within the left neighbor time range at any nonlinear interpolation time. Obtain the right average value of any target monitoring index within the right neighborhood time range at any nonlinear interpolation time, and use the absolute value of the difference between the left average value and the right average value as the independent variable of the hyperbolic tangent function to obtain the data level difference degree of any target monitoring index within the local time range at any nonlinear interpolation time. Calculate the difference between the maximum and minimum values ​​in the left neighbor data sequence to obtain the left neighbor fluctuation degree of any target monitoring index within the left neighbor time range at any nonlinear interpolation time. Obtain the right neighbor fluctuation degree of any target monitoring index within the right neighbor time range at any nonlinear interpolation time. Use the absolute value of the difference between the left neighbor fluctuation degree and the right neighbor fluctuation degree as the independent variable of the hyperbolic tangent function to obtain the fluctuation range difference degree of any target monitoring index within the local time range at any nonlinear interpolation time. Calculate the average value between the data level difference and the fluctuation range difference to obtain the data difference between the two sides of any target monitoring index at any nonlinear interpolation time.

[0013] Preferably, the step of obtaining the data jump degree of any target monitoring indicator at any nonlinear interpolation time based on the fluctuation direction and fluctuation degree between every two adjacent data in the local neighborhood data sequence includes: Obtain the first-order difference sequence of the local neighborhood data sequence. For any two adjacent data in the first-order difference sequence, calculate the absolute value of the difference between the absolute values ​​of the two adjacent data to obtain the degree of fluctuation between the two adjacent data. If the product between any two adjacent data is less than 0, the jump weight of the fluctuation degree between any two adjacent data is set to a constant 1; if the product between any two adjacent data is greater than or equal to 0, the jump weight of the fluctuation degree between any two adjacent data is set to a constant 0. The fluctuation degree and its jump weight between every two adjacent data in the first-order difference sequence are obtained respectively. The fluctuation degree between every two adjacent data in the first-order difference sequence is weighted and summed to obtain the summation result. The summation result is used as the independent variable of the hyperbolic tangent function to obtain the data jump degree of any target monitoring index at any nonlinear interpolation time.

[0014] Preferably, the step of dividing all nonlinear interpolation moments into continuous and discontinuous interpolation moments based on the discontinuous characteristic values ​​of any target monitoring index within the local time range at each nonlinear interpolation moment includes: Nonlinear interpolation times corresponding to discontinuous feature values ​​that are greater than or equal to a preset discontinuity threshold are recorded as discontinuous interpolation times, while nonlinear interpolation times corresponding to data stability values ​​that are less than the preset discontinuity threshold are recorded as continuous interpolation times.

[0015] Preferably, the adaptive allocation of different optimal interpolation methods for each linear interpolation time, continuous interpolation time, and discontinuous interpolation time includes: Linear interpolation is used as the optimal interpolation method for each linear interpolation time step, interpolation method based on the continuity assumption is used as the optimal interpolation method for each continuous interpolation time step, and interpolation method based on the discontinuity assumption is used as the optimal interpolation method for each discontinuous interpolation time step.

[0016] The beneficial effects of the embodiments of the present invention compared with the prior art are as follows: This invention analyzes the stationarity characteristics of data within the local time range of each interpolation moment, filters out interpolation moments with strong linear trends, and records them as linear interpolation moments. It then analyzes the continuity and volatility of data within the local time range of each nonlinear interpolation moment, determines the strength of data continuity within that range, and divides all nonlinear interpolation moments into continuous and discontinuous interpolation moments. For each linear interpolation moment, continuous interpolation moment, and discontinuous interpolation moment, different optimal interpolation methods are adaptively assigned, and interpolation fill is performed on each interpolation moment according to the optimal interpolation method to achieve time alignment. This allows the interpolated data to more realistically reflect the actual abrupt changes of pollutants in space, effectively approximate the actual change path of the data, improve the authenticity and robustness of the time-aligned data, and thus significantly improve the accuracy of monitoring pollution sources based on multi-source monitoring data collected by UAVs. Attached Figure Description

[0017] To more clearly illustrate the technical solutions in the embodiments of the present invention, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0018] Figure 1 This is a flowchart of a pollution source monitoring method based on UAV multi-source sensing provided in Embodiment 1 of the present invention. Detailed Implementation

[0019] Embodiments of this disclosure are described in detail below, with examples of these embodiments illustrated in the accompanying drawings. The embodiments described below with reference to the accompanying drawings are exemplary and intended to explain this disclosure, and should not be construed as limiting it.

[0020] It should be noted that the terms "first," "second," etc., used in this disclosure and the accompanying drawings are used to distinguish similar objects and are not necessarily used to describe a specific order or sequence. It should be understood that such data can be interchanged where appropriate so that the embodiments of this disclosure described herein can be implemented in orders other than those illustrated or described herein. The embodiments described in the following exemplary embodiments do not represent all embodiments consistent with this disclosure. Rather, they are merely examples of apparatuses and methods consistent with some aspects of this disclosure.

[0021] To illustrate the technical solution of the present invention, specific embodiments are described below.

[0022] The specific scenario addressed by this invention is as follows: When using a fixed-frequency resampling + linear interpolation method to time-align asynchronous data from various pollution sources, the data from various sensors exhibit significant local variation differences due to the influence of airflow, drone flight paths, and pollution source operating conditions. If only a single interpolation method is used, it cannot take into account the local characteristics of data under different variations, which may smooth out peaks or cause oscillations, resulting in data distortion after time alignment. Therefore, this invention adaptively obtains the optimal interpolation method based on the local neighborhood data variation characteristics of each interpolation point, so that the time-aligned data can more realistically reflect the actual abrupt changes of pollutants in space, significantly improving the accuracy of monitoring pollution sources based on multi-source monitoring data collected by drones.

[0023] This invention provides a pollution source monitoring system based on UAV multi-source sensing, including a processor and a memory. The processor executes a computer program stored in the memory to implement a pollution source monitoring method based on UAV multi-source sensing, such as... Figure 1 As shown, the method includes the following steps: Step S101: Based on the sampling frequency of the sensors corresponding to each pollution source monitoring indicator, the pollution source monitoring indicators corresponding to sensors with sampling frequencies lower than the preset target sampling frequency are designated as target monitoring indicators. For any target monitoring indicator, based on the difference between the sampling frequency of the sensor corresponding to the target monitoring indicator and the preset target sampling frequency, the interpolation time of the target monitoring indicator within the preset time period is obtained.

[0024] The system acquires various pollution source monitoring indicators, such as the concentrations of gases like sulfur dioxide, nitrogen oxides, methane, and ammonia, and the concentrations of particulate matter with different particle sizes, such as PM1.0 and PM2.5. Sensors corresponding to these pollution source monitoring indicators are mounted on drones to collect time-series data of these indicators. Multi-source data fusion algorithms are then used to monitor and issue early warnings for pollution sources.

[0025] Due to inherent limitations in sensor hardware, different sensors have varying physicochemical response times. The sampling interval of a sensor must be set longer than its response time to obtain accurate data. For example, forcing a slow sensor (such as a gas sensor) to sample at high frequencies will generate a large amount of low-quality, high-noise redundant data. Therefore, the sampling frequencies of different sensors cannot be forced to be identical, preventing data from various pollution sources from synchronizing on the same timescale. Consequently, time alignment of the data collected by each sensor is necessary. Time alignment is divided into upsampling (converting low-frequency data to high-frequency data) and downsampling (converting high-frequency data to low-frequency data). Since upsampling essentially infers unknown data from known data, it is more prone to introducing false information than downsampling; therefore, the choice of upsampling method requires extreme caution.

[0026] First, obtain the pollution source monitoring indicators that need to be time-aligned through upsampling: Based on the drone's flight speed, the sampling frequency of the sensors corresponding to each pollution source monitoring indicator, and the system resource boundary, determine the preset target sampling frequency as the standard for time alignment. Then, mark the pollution source monitoring indicators corresponding to sensors with sampling frequencies lower than the preset target sampling frequency as target monitoring indicators, so as to perform time alignment of the data of each target monitoring indicator through upsampling.

[0027] Currently, fixed interpolation methods are mostly used for upsampling. However, data from various pollution sources are affected by airflow, drone flight paths, and pollution source conditions, leading to significant local variations in the data collected by sensors. For example, when passing through areas with stable airflow and sufficient pollutant dispersion, gas or particulate matter concentrations change slowly, forming a stable range. However, pollution sources often experience instantaneous emission events. When the drone enters an area with drastic changes in wind speed or direction, or approaches the edge of an instantaneous emission pollution source, pollutant concentrations may change rapidly, forming sudden peaks or abruptly changing ranges. If only a single interpolation method is used, it cannot take into account the local characteristics of data under different changes, which may smooth out peaks or cause oscillations, resulting in data distortion after time alignment and weakening the ability to capture short-term pollution events. Therefore, in this embodiment of the invention, the optimal interpolation method is adaptively obtained based on the data change characteristics within the local time range of each interpolation moment. This allows the time-aligned data to more realistically reflect the true abrupt changes in pollutant behavior in space, thereby significantly improving the accuracy of pollution source monitoring.

[0028] Taking the w-th target monitoring indicator as an example, firstly, based on the difference between the sampling frequency of the sensor corresponding to the w-th target monitoring indicator and the preset target sampling frequency, the interpolation time of the w-th target monitoring indicator within the preset time period is obtained. The general process is as follows: A target time axis within the preset time period is constructed based on the preset target sampling frequency, obtaining the target time on the target time axis. Then, it is determined whether the target time on the target time axis is covered by the sampling time of the w-th target monitoring indicator within the preset time period. The uncovered target time is recorded as the interpolation time of the w-th target monitoring indicator within the preset time period. The acquisition of the interpolation time is existing technology and will not be elaborated here. The preset time period is set to 30 minutes, which is not a limitation. Implementers need to set this based on various environmental and physical factors such as the sampling frequency of the w-th target monitoring indicator, the drone's flight speed, the pollution event occurrence cycle, and wind speed, in order to monitor and warn of pollution sources based on data within the preset time period.

[0029] Step S102: Based on the data stability of any target monitoring index within the local time range of each interpolation moment, divide all interpolation moments into linear interpolation moments and nonlinear interpolation moments. Based on the continuity and volatility of any target monitoring index within the local time range of each nonlinear interpolation moment, divide all nonlinear interpolation moments into continuous interpolation moments and discontinuous interpolation moments.

[0030] Since atmospheric pollutants (such as gases or particulate matter) diffuse or transform through processes such as diffusion, sedimentation, and chemical reactions, the data changes caused by these processes are mostly gradual. Furthermore, in areas with stable airflow, the external environment has less interference with pollutant concentrations. Therefore, data collected by drones in such areas typically exhibits slow changes and relatively simple trends. Considering that linear interpolation fills the gaps between two known data points by connecting them with a straight line, for such slowly changing and relatively simple data, linear interpolation can not only effectively approximate the actual data change path but also avoid the computational overhead of overly complex interpolation methods, while ensuring smooth transitions between data points. Therefore, firstly, based on the data stationarity of the w-th target monitoring indicator within the local time range of each interpolation moment within the preset time period, the data stationarity characteristic index of the w-th target monitoring indicator within the local time range of each interpolation moment within the preset time period is obtained to determine the interpolation moment most suitable for linear interpolation. Specifically: Taking the t-th interpolation time of the w-th target monitoring indicator within a preset time period as an example, according to the sampling frequency of the sensor corresponding to the w-th target monitoring indicator, the data of the w-th target monitoring indicator before the t-th interpolation time is obtained by acquiring a preset number of sampling times, and the left neighbor data sequence of the t-th interpolation time is obtained. Here, the preset number is set to 15, which is not limited here. The implementer needs to set it according to factors such as the sampling frequency of the sensor corresponding to the w-th target monitoring indicator, the degree of influence of the w-th target monitoring indicator on changes in meteorological conditions or intermittent release of pollution sources, and the flight speed of the drone. The rate of change of data between every two adjacent data points in the left neighbor data sequence is obtained to form a rate of change data sequence. Taking the i-th data point and the (i+1)-th data point in the left neighbor data sequence as an example, the formula for calculating the rate of change of data between the i-th data point and the (i+1)-th data point is as follows: ,in, This represents the rate of change of data between the i-th data point and the (i+1)-th data point. This represents the (i+1)th data in the left neighbor data sequence. This represents the i-th data in the left neighbor data sequence. This represents the time interval between the i-th and (i+1)-th data points in the left neighbor data sequence, which is also the sampling interval of the sensor corresponding to the w-th target monitoring indicator. Indicates the absolute value symbol; Obtain the first-order difference sequence of the change rate data sequence, accumulate the absolute values ​​of each data point in the first-order difference sequence to obtain the change rate difference degree of the w-th target monitoring index within the left neighbor time range at the t-th interpolation time, denoted as the left neighbor difference degree. Indicates, that is ,in, This represents a preset quantity, which is also the number of data points in the left neighbor data sequence. In this embodiment of the invention, , This indicates the rate of change of data between the (i+1)th and (i+2)th data points in the left neighbor data sequence. This indicates the rate of change of data between the i-th data and the (i+1)-th data in the left neighbor data sequence. That is, the i-th data point in the first-order difference sequence of the change rate data sequence. The smaller, The smaller the value, the smaller the difference in the rate of change between two adjacent data in the left neighbor data sequence, that is, the more similar the rate of change of data, and the more stable the data change trend in the left neighbor data sequence. Similar to obtaining the left neighbor difference, based on the data of the w-th target monitoring index at a preset number of sampling times after the t-th interpolation time, the rate of change difference of the w-th target monitoring index within the right neighbor time range at the t-th interpolation time is obtained, denoted as the right neighbor difference. express; Using the negative of the sum of the left and right neighborhood differences as the independent variable of an exponential function with the natural constant as the base, we obtain the data stationarity characteristic index of the w-th target monitoring index within the local time range at the t-th interpolation time.

[0031] In one embodiment, the formula for calculating the data stationarity characteristic index of the w-th target monitoring index within the local time range at the t-th interpolation time is: in, This represents the data stationarity index of the w-th target monitoring indicator within a local time range at the t-th interpolation time. This represents the rate of change difference of the w-th target monitoring indicator within the left neighborhood time range at the t-th interpolation time, i.e., the left neighborhood variability. This represents the rate of change difference of the w-th monitoring indicator within its right neighborhood over the time range of the t-th interpolation time, i.e., the right neighborhood variability. This represents an exponential function with the natural constant as the base, used to restrict the output to [0, 1].

[0032] It should be noted that, The smaller the value, the more stable the data change of the monitoring index of the w-th target within the local time range at the t-th interpolation time, and thus... The larger the value, the better it matches the linear interpolation method at the t-th interpolation time.

[0033] Similarly, the data stationarity index of the w-th target monitoring index is obtained within the local time range of each interpolation time. Further, the interpolation time corresponding to the data stationarity index greater than or equal to a preset stationarity threshold is recorded as the linear interpolation time, which is more suitable for linear interpolation; the interpolation time corresponding to the data stationarity index less than the preset stationarity threshold is recorded as the nonlinear interpolation time. Regarding the setting of the preset stationarity threshold, since setting it too high may lead to excessively high requirements for data stationarity, thus increasing unnecessary computational overhead; setting it too low may lead to insufficient interpolation accuracy, thereby introducing erroneous or inaccurate data and reducing the reliability of the interpolated data, in this embodiment of the invention, the preset stationarity threshold is set to 0.7. This is not a limitation, and implementers can set it according to their tolerance for the data stationarity of the w-th target monitoring index.

[0034] For nonlinear interpolation moments, the data of the w-th target monitoring index within the local time range of the nonlinear interpolation moment does not exhibit strong stationarity. This typically reflects phenomena such as drastic changes in wind speed and direction, or the drone approaching the edge of a pollution source with instantaneous emissions. When a drone flies into an area of ​​drastic changes in wind speed and direction or approaches the edge of an instantaneous emission source, due to the sudden gradient of pollutant concentration in space, the monitoring data of the w-th target monitoring index may show a rapid but continuous transition, or it may show discontinuous jumps due to the drone crossing a narrow-scale gradient zone, the duration of the pollutant emission event being shorter than the sampling interval, or the formation of multiple small block or strip-shaped high-concentration areas of pollutants in the wind field by airflow with vortex structures (i.e., plume fragmentation caused by turbulence). In other words, the data of the w-th target monitoring index may show a rapid but smooth continuous change without abrupt jumps or breaks, or it may show a sudden jump, with the trend being interrupted or reversed instantly.

[0035] Since interpolation essentially infers data at unknown locations based on data at known locations, it represents an assumption and extension of data change patterns. For continuously changing data, the change process exhibits a traceable and extendable trend within the neighborhood, meaning that there is a clear functional relationship or at least some smoothness between adjacent sampling times. However, for discontinuously changing data, abrupt changes are not caused by smooth evolution but are influenced by factors such as spatial cross-regional changes, turbulent patch structures, and short-term pollution events. In this case, there is no extendable trend or functional relationship between adjacent sampling times. Therefore, the interpolation strategies adapted to these two types of data with different characteristics are completely different. Based on the above idea, we analyze the continuity of the w-th monitoring index within the local time range of each nonlinear interpolation time, and obtain the discontinuous characteristic values ​​of the w-th monitoring index within the local time range of each nonlinear interpolation time, in order to assign the optimal interpolation strategy for each nonlinear interpolation time.

[0036] Taking the c-th nonlinear interpolation time as an example, the steps to obtain the discontinuous characteristic value of the w-th monitoring index within the local time range at the c-th nonlinear interpolation time are as follows: (1) Based on the data difference between the left and right neighboring data sequences at the cth nonlinear interpolation time, obtain the data difference degree between the two sides of the wth target monitoring index at the cth nonlinear interpolation time.

[0037] Specifically, the average value of all data in the left neighbor data sequence at the c-th nonlinear interpolation time is obtained, and the left average value of the w-th target monitoring index within the left neighbor time range at the c-th nonlinear interpolation time is obtained. Obtain the right average value of the monitoring index of the w-th target within the right neighborhood time range at the c-th nonlinear interpolation time. Use the absolute value of the difference between the left average value and the right average value as the independent variable of the hyperbolic tangent function to obtain the data level difference degree of the monitoring index of the w-th target within the local time range at the c-th nonlinear interpolation time. Calculate the difference (range) between the maximum and minimum values ​​in the left neighbor data sequence at the c-th nonlinear interpolation time. This yields the left neighbor fluctuation degree of the w-th target monitoring index within the left neighbor time range at the c-th nonlinear interpolation time, denoted as [missing information]. To obtain the right neighborhood fluctuation degree of the monitoring index of the w-th target within the right neighborhood time range at the c-th nonlinear interpolation time, denoted as The absolute value of the difference between the fluctuation degree of the left neighborhood and the fluctuation degree of the right neighborhood is used as the independent variable of the hyperbolic tangent function to obtain the fluctuation range difference of the w-th target monitoring index within the local time range at the c-th nonlinear interpolation time. Calculate the average value between the data level difference and the fluctuation range difference to obtain the data difference between the two sides of the w-th target monitoring index at the c-th nonlinear interpolation time.

[0038] In one embodiment, the formula for calculating the difference in data between the two sides of the w-th target monitoring index at the c-th nonlinear interpolation time is: in, This represents the difference between the two sides of the monitoring index of the w-th target at the c-th nonlinear interpolation time. This represents the number of data points in the left or right neighboring data sequence at the c-th nonlinear interpolation time. In this embodiment of the invention, , This represents the i-th data point in the left neighbor data sequence at the c-th nonlinear interpolation time. This represents the i-th data point in the right neighbor data sequence at the c-th nonlinear interpolation time. This represents the degree of fluctuation in the left neighborhood of the monitoring index of the w-th target within the time range of the left neighborhood at the c-th nonlinear interpolation time. This represents the degree of right-neighbor fluctuation of the w-th target monitoring index within the right-neighbor time range at the c-th nonlinear interpolation time. This represents the hyperbolic tangent function, used to restrict the output to [0, 1]. Represents the absolute value symbol.

[0039] It should be noted that, The larger, The larger the value, the greater the difference in the numerical levels and fluctuation ranges of the w-th monitoring indicator at the c-th nonlinear interpolation time. This indicates that the data for the w-th monitoring indicator before and after the c-th nonlinear interpolation time are more likely to belong to two different air plates. The larger the value, the less continuous the data of the w-th target monitoring index is at the c-th nonlinear interpolation time.

[0040] (2) Combine the left and right neighbor data sequences at the cth nonlinear interpolation time according to the time sequence to obtain the local neighbor data sequence at the cth nonlinear interpolation time. Based on the fluctuation direction and fluctuation degree between each two adjacent data in the local neighbor data sequence, obtain the data jump degree of the wth target monitoring index at the cth nonlinear interpolation time.

[0041] Specifically: Obtain the first-order difference sequence of the local neighborhood data sequence; for any two adjacent data in the first-order difference sequence, take the i-th data in the first-order difference sequence... and the (i+1)th data For example, among which, , This represents the i-th data point in the local neighborhood data sequence. This represents the (i+1)th data point in the local neighborhood data sequence. Similarly, ,calculate absolute value and The absolute value of the difference between the absolute values ​​is obtained. and The degree of fluctuation between them; like and The product between them is less than 0, indicating that in and The i-th data in the corresponding local neighborhood data sequence The (i+1)th data and the (i+2)th data If the direction of data change changes, forming a "high-low-high" or "low-high-low" jump, then set... and The jump weight for the degree of fluctuation between them is a constant 1, and conversely, if and The product between them is greater than or equal to 0, indicating that in and The i-th data in the corresponding local neighborhood data sequence The (i+1)th data and the (i+2)th data If the direction of data change between any two adjacent data points does not change, or if no data change occurs, i.e. there is no "high-low-high" or "low-high-low" jump, then the jump weight of the fluctuation degree between any two adjacent data points is set to a constant of 0. The fluctuation degree and its jump weight between every two adjacent data in the first-order difference sequence are obtained respectively. The fluctuation degree between every two adjacent data in the first-order difference sequence is weighted and summed to obtain the summation result. The summation result is used as the independent variable of the hyperbolic tangent function to obtain the data jump degree of the w-th target monitoring index at the c-th nonlinear interpolation time.

[0042] In one embodiment, the formula for calculating the degree of data jump of the w-th target monitoring index at the c-th nonlinear interpolation time is: in, This indicates the degree of data jump in the w-th target monitoring index at the c-th nonlinear interpolation time. This represents the number of data points in the left or right neighbor data sequence at the c-th nonlinear interpolation time. That is, the number of data in the local neighborhood data sequence. In this embodiment of the invention, , This represents the (i+1)th data point in the first-order difference sequence of the local neighborhood data sequence. This represents the i-th data point in the first-order difference sequence of the local neighborhood data sequence. This represents the jump weight indicating the degree of fluctuation between the i-th and (i+1)-th data points in the first-order difference sequence of a local neighborhood data sequence. This represents the hyperbolic tangent function, used to restrict the output to [0, 1]. Represents the absolute value symbol.

[0043] It should be noted that, The larger the value, the more data in the local neighborhood data sequence at the c-th nonlinear interpolation time exhibit "high-low-high" or "low-high-low" jumps, and the more drastic the jumps, indicating more drastic data fluctuations. The larger the value, the less continuous the data of the w-th target monitoring index is at the c-th nonlinear interpolation time.

[0044] (3) Calculate the difference between the two sides of the monitoring index of the w-th target at the c-th nonlinear interpolation time. With the degree of data jump The mean of the values ​​is used to obtain the discontinuous characteristic value of the monitoring index of the w-th target at the c-th nonlinear interpolation time within the local time range, denoted as . ,Right now .

[0045] Similarly, the discontinuous characteristic values ​​of the w-th target monitoring index within the local time range of each nonlinear interpolation moment are obtained. Further, the nonlinear interpolation moments corresponding to discontinuous characteristic values ​​greater than or equal to a preset discontinuity threshold are recorded as discontinuous interpolation moments, and the nonlinear interpolation moments corresponding to discontinuous characteristic values ​​less than the preset discontinuity threshold are recorded as continuous interpolation moments. Regarding the setting of the preset discontinuity threshold, if the threshold is set too high, most nonlinear interpolation moments may be judged as continuous, leading to an overly strong continuity assumption and potentially generating false transitions during interpolation, thus reducing the reliability of the interpolated data. If the threshold is set too low, most nonlinear interpolation moments may be judged as discontinuous, potentially amplifying random fluctuations in measurement noise during interpolation. Considering... The value range is [0, 1]. This solution takes a compromise and sets the preset discontinuity threshold to 0.5. There is no restriction here, and implementers can set it according to the specific scenario.

[0046] Thus, by analyzing the data change characteristics of the w-th target monitoring indicator within the local time range of each interpolation moment in the preset time period, all interpolation moments of the w-th target monitoring indicator in the preset time period are divided into linear interpolation moments, continuous interpolation moments, and discontinuous interpolation moments, which are used to assign the optimal interpolation strategy to each interpolation moment.

[0047] Step S103: For each linear interpolation time, continuous interpolation time, and discontinuous interpolation time, different optimal interpolation methods are adaptively assigned. Based on the optimal interpolation method corresponding to each interpolation time, interpolation is performed to fill in the gaps at each interpolation time to obtain the resampled data sequence of any target monitoring indicator within a preset time period. Based on the resampled data sequence of each target monitoring indicator, a multi-source data fusion algorithm is used to realize the monitoring and early warning of pollution sources.

[0048] For linear interpolation time, since the data of the w-th target monitoring index has strong stability within the local time range of the linear interpolation time, the linear interpolation method is taken as the optimal interpolation method for each linear interpolation time.

[0049] For continuous interpolation time, since the data of the w-th target monitoring index has strong continuity within the local time range of the continuous interpolation time, the interpolation method based on the continuity assumption is used as the optimal interpolation method for each continuous interpolation time, such as cubic spline interpolation method, polynomial interpolation method, etc.

[0050] For discontinuous interpolation moments, since the data of the w-th target monitoring index does not have strong continuity within the local time range of continuous interpolation moments, the interpolation method based on the discontinuity assumption is used as the optimal interpolation method for each discontinuous interpolation moment, such as the zero-order preservation method. This ensures that the interpolated data, without introducing any additional transitions and only maintaining the interval attributes of local values, is consistent with the actual "jump" behavior of pollutants at the disturbance boundary. This can effectively avoid false smoothing or oscillation, thereby improving the authenticity and robustness of the time-aligned data and significantly enhancing the accuracy of monitoring pollution sources based on multi-source monitoring data collected by UAVs.

[0051] Based on the optimal interpolation method corresponding to each interpolation time, interpolation is performed at each interpolation time to obtain the resampled data sequence of the w-th target monitoring indicator within the preset time period. Similarly, the resampled data sequence of each target monitoring indicator within the preset time period is obtained.

[0052] For pollution source monitoring indicators corresponding to sensors with sampling frequencies greater than the preset target sampling frequency, it is necessary to use aggregation methods (such as taking the average value, maximum value, and minimum value) to downsample the data of these pollution source monitoring indicators within a preset time period to achieve time alignment and obtain the corresponding resampled data sequence.

[0053] After obtaining the resampling sequences of various pollution source monitoring indicators within a preset time period, a spatial distribution map of pollutants is obtained by using a multi-source data fusion algorithm (such as Kalman filtering or other state estimation methods) combined with wind speed, wind direction, environmental parameters, and location data. Finally, based on the spatial distribution map of pollutants, a pre-trained pollution diffusion model is used to locate pollution sources and invert emission intensity, thereby deducing the specific location, emission rate, and impact range of the pollution sources and generating a monitoring report to trigger an early warning when abnormal emissions are detected, thus realizing the monitoring and early warning of pollution sources.

[0054] It is worth noting that the focus of this invention is on how to select the optimal interpolation method to upsample the pollution source monitoring data corresponding to sensors with sampling frequencies lower than the preset target sampling frequency, thereby achieving time alignment and improving the authenticity and robustness of the time-aligned data, significantly enhancing the accuracy of pollution source monitoring. Downsampling the pollution source monitoring index data corresponding to sensors with sampling frequencies higher than the preset target sampling frequency using aggregation methods, and utilizing multi-source data fusion algorithms to achieve pollution source monitoring and early warning are existing technologies and will not be elaborated upon here.

[0055] The above embodiments are only used to illustrate the technical solutions of the present invention, and are not intended to limit it. Although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features. Such modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of the present invention, and should all be included within the protection scope of the present invention.

Claims

1. A pollution source monitoring system based on UAV multi-source sensing, comprising a memory, a processor, and a computer program stored in the memory and running on the processor, characterized in that, When the processor executes the computer program, it performs the following steps: Based on the sampling frequency of the sensors corresponding to each pollution source monitoring indicator, the pollution source monitoring indicators corresponding to sensors with sampling frequencies lower than the preset target sampling frequency are designated as target monitoring indicators. For any target monitoring indicator, the interpolation time of the target monitoring indicator within a preset time period is obtained based on the difference between the sampling frequency of the sensor corresponding to the target monitoring indicator and the preset target sampling frequency. Based on the stability of the data of any target monitoring index within the local time range at each interpolation time, all interpolation times are divided into linear interpolation times and nonlinear interpolation times. Based on the continuity and volatility of any target monitoring index within the local time range at each nonlinear interpolation time, all nonlinear interpolation times are divided into continuous interpolation times and discontinuous interpolation times. Different optimal interpolation methods are adaptively assigned to each linear interpolation time, continuous interpolation time, and discontinuous interpolation time. Based on the optimal interpolation method corresponding to each interpolation time, interpolation is performed to fill in the gaps at each interpolation time to obtain the resampled data sequence of any target monitoring indicator within a preset time period. Based on the resampled data sequence of each target monitoring indicator, a multi-source data fusion algorithm is used to realize the monitoring and early warning of pollution sources.

2. The pollution source monitoring system based on UAV multi-source sensing according to claim 1, characterized in that, The method of dividing all interpolation times into linear interpolation times and nonlinear interpolation times based on the data stability of any target monitoring index within a local time range at each interpolation time includes: For any interpolation time, based on the sampling frequency of the sensor corresponding to any target monitoring index, data of any target monitoring index at a preset number of sampling times before any interpolation time are obtained, and the left neighbor data sequence at any interpolation time is obtained. Based on the difference in the rate of change between each two adjacent data in the left neighbor data sequence, the rate of change difference of any target monitoring index within the left neighbor time range at any interpolation time is obtained, and is denoted as the left neighbor difference. Based on the data of any target monitoring index at a preset number of sampling times after any interpolation time, the rate of change difference of any target monitoring index within the right neighbor time range at any interpolation time is obtained and denoted as the right neighbor difference. The negative of the sum of the left neighborhood difference and the right neighborhood difference is used as the independent variable of an exponential function with the natural constant as the base, to obtain the data stationarity characteristic index of any target monitoring index within the local time range at any interpolation time. Obtain the data stationarity characteristic index of any target monitoring indicator within the local time range at each interpolation time. Based on the data stationarity characteristic index of any target monitoring indicator within the local time range at each interpolation time, divide all interpolation times into linear interpolation times and nonlinear interpolation times.

3. A pollution source monitoring system based on UAV multi-source sensing according to claim 2, characterized in that, The step of obtaining the rate of change difference of any target monitoring indicator within the left neighbor time range at any interpolation time based on the rate of change difference between every two adjacent data in the left neighbor data sequence includes: The rate of change of data between every two adjacent data in the left neighbor data sequence is obtained to obtain the rate of change data sequence. The first-order difference sequence of the rate of change data sequence is obtained, and the absolute value of each data in the first-order difference sequence is accumulated to obtain the rate of change difference of any target monitoring index within the left neighbor time range at any interpolation time.

4. A pollution source monitoring system based on UAV multi-source sensing according to claim 2, characterized in that, The method of dividing all interpolation times into linear interpolation times and nonlinear interpolation times based on the data stationarity characteristic index of any target monitoring index within the local time range at each interpolation time includes: The interpolation time corresponding to the data stationarity characteristic index that is greater than or equal to the preset stationarity threshold is recorded as the linear interpolation time, and the interpolation time corresponding to the data stationarity characteristic index that is less than the preset stationarity threshold is recorded as the nonlinear interpolation time.

5. A pollution source monitoring system based on UAV multi-source sensing according to claim 1, characterized in that, The step of dividing all nonlinear interpolation moments into continuous interpolation moments and discontinuous interpolation moments based on the continuity and volatility of any target monitoring index within the local time range at each nonlinear interpolation moment includes: For any nonlinear interpolation time, based on the sampling frequency of the sensor corresponding to any target monitoring index, data of any target monitoring index at a preset number of sampling times before any nonlinear interpolation time are obtained to obtain the left neighbor data sequence at any nonlinear interpolation time. Based on the data of any target monitoring index at a preset number of sampling times after any nonlinear interpolation time, the right neighbor data sequence at any nonlinear interpolation time is obtained. Based on the data difference between the left neighbor data sequence and the right neighbor data sequence, the data difference degree between the two sides of any target monitoring index at any nonlinear interpolation time is obtained. The left neighbor data sequence and the right neighbor data sequence are combined according to time sequence to obtain the local neighbor data sequence at any nonlinear interpolation time. Based on the fluctuation direction and fluctuation degree between each two adjacent data in the local neighbor data sequence, the data jump degree of any target monitoring indicator at any nonlinear interpolation time is obtained. Calculate the mean between the difference between the two sides of the data and the degree of data jump to obtain the discontinuous characteristic value of any target monitoring index within the local time range at any nonlinear interpolation time. Obtain the discontinuous characteristic value of any target monitoring index within the local time range at each nonlinear interpolation moment. Based on the discontinuous characteristic value of any target monitoring index within the local time range at each nonlinear interpolation moment, divide all nonlinear interpolation moments into continuous interpolation moments and discontinuous interpolation moments.

6. A pollution source monitoring system based on UAV multi-source sensing according to claim 5, characterized in that, The step of obtaining the data difference degree between the left and right neighboring data sequences for any target monitoring indicator at any nonlinear interpolation time based on the data difference between the left and right neighboring data sequences includes: The average value of all data in the left neighbor data sequence is obtained to obtain the left average value of any target monitoring index within the left neighbor time range at any nonlinear interpolation time. Obtain the right average value of any target monitoring index within the right neighborhood time range at any nonlinear interpolation time, and use the absolute value of the difference between the left average value and the right average value as the independent variable of the hyperbolic tangent function to obtain the data level difference degree of any target monitoring index within the local time range at any nonlinear interpolation time. Calculate the difference between the maximum and minimum values ​​in the left neighbor data sequence to obtain the left neighbor fluctuation degree of any target monitoring index within the left neighbor time range at any nonlinear interpolation time. Obtain the right neighbor fluctuation degree of any target monitoring index within the right neighbor time range at any nonlinear interpolation time. Use the absolute value of the difference between the left neighbor fluctuation degree and the right neighbor fluctuation degree as the independent variable of the hyperbolic tangent function to obtain the fluctuation range difference degree of any target monitoring index within the local time range at any nonlinear interpolation time. Calculate the average value between the data level difference and the fluctuation range difference to obtain the data difference between the two sides of any target monitoring index at any nonlinear interpolation time.

7. A pollution source monitoring system based on UAV multi-source sensing according to claim 5, characterized in that, The step of obtaining the data jump degree of any target monitoring indicator at any nonlinear interpolation time based on the fluctuation direction and fluctuation degree between each two adjacent data in the local neighborhood data sequence includes: Obtain the first-order difference sequence of the local neighborhood data sequence. For any two adjacent data in the first-order difference sequence, calculate the absolute value of the difference between the absolute values ​​of the two adjacent data to obtain the degree of fluctuation between the two adjacent data. If the product between any two adjacent data is less than 0, the jump weight of the fluctuation degree between any two adjacent data is set to a constant 1; if the product between any two adjacent data is greater than or equal to 0, the jump weight of the fluctuation degree between any two adjacent data is set to a constant 0. The fluctuation degree and its jump weight between every two adjacent data in the first-order difference sequence are obtained respectively. The fluctuation degree between every two adjacent data in the first-order difference sequence is weighted and summed to obtain the summation result. The summation result is used as the independent variable of the hyperbolic tangent function to obtain the data jump degree of any target monitoring index at any nonlinear interpolation time.

8. A pollution source monitoring system based on UAV multi-source sensing according to claim 5, characterized in that, The step of dividing all nonlinear interpolation moments into continuous and discontinuous interpolation moments based on the discontinuous characteristic values ​​of any target monitoring index within the local time range at each nonlinear interpolation moment includes: Nonlinear interpolation times corresponding to discontinuous feature values ​​that are greater than or equal to a preset discontinuity threshold are recorded as discontinuous interpolation times, while nonlinear interpolation times corresponding to data stability values ​​that are less than the preset discontinuity threshold are recorded as continuous interpolation times.

9. A pollution source monitoring system based on UAV multi-source sensing according to claim 1, characterized in that, The adaptive allocation of different optimal interpolation methods for each linear interpolation time, continuous interpolation time, and discontinuous interpolation time includes: Linear interpolation is used as the optimal interpolation method for each linear interpolation time step, interpolation method based on the continuity assumption is used as the optimal interpolation method for each continuous interpolation time step, and interpolation method based on the discontinuity assumption is used as the optimal interpolation method for each discontinuous interpolation time step.

Images