Multi-dimensional data fusion method and system for intelligent construction site water condition monitoring

Abstract

The present application relates to the technical field of multi-source data fusion, and proposes a multi-dimensional data fusion method and system for intelligent construction site water condition monitoring, comprising: collecting water level data, flow rate data and rainfall at a monitoring point of the construction site water condition; screening the water level data and forming a high-frequency water level data sequence; using wavelet transform to decompose the high-frequency water level data sequence into a first preset number of layers, and obtaining the instantaneous energy intensity of each detail coefficient in each layer of the detail coefficient sequence; processing the detail coefficients of the first second preset number of layers, combining the detail coefficients of all other layers, obtaining the denoised high-frequency water level data sequence, and obtaining the denoised water level data; and performing multi-dimensional data fusion on the flow rate data, rainfall and denoised water level data. The present application can obtain an accurate denoised water level measurement signal, and improve the accuracy of the construction site water condition monitoring result.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of multi-source data fusion, and particularly relates to a multi-dimensional data fusion method and system for intelligent construction site water condition monitoring. BACKGROUND

[0002] When the water condition of a construction site is monitored, in order to more comprehensively show the water condition of the construction site, the water condition of the construction site is often comprehensively monitored according to information provided by multi-dimensional data, so as to improve the accuracy of water condition change monitoring, realize fine management, and enhance the emergency response capability. In order to ensure the accuracy of the water condition monitoring of the construction site, the accuracy of the multi-dimensional data is crucial. When the precipitation is large, the water flow is turbulent, the water flow and water quantity in the upstream river area increase rapidly, the flow velocity increases rapidly, and turbulent water flow is formed. When the high-speed water flow mixes with air or collides with obstacles, a large amount of air bubbles will be mixed into the water, which affects the accuracy of the water level measurement signal.

[0003] When the collected water level measurement signal is directly denoised, in order to sufficiently suppress the high-frequency bubble pulse, the water level measurement signal needs to be subjected to high-intensity noise suppression. However, high-intensity noise suppression is easy to cause loss of the true signal, and cannot take into account the difference between the characteristics of the sudden interference and the slowly varying noise. The quality of the denoised water level measurement signal is poor, which often leads to inaccurate construction site water condition monitoring results based on multi-dimensional data. SUMMARY

[0004] The present application provides a multi-dimensional data fusion method and system for intelligent construction site water condition monitoring, to solve the problem of poor denoising effect of the water level measurement signal affected by the air bubbles in the water flow, which leads to inaccurate construction site water condition monitoring results based on multi-dimensional data. The technical solution adopted is as follows:

[0005] In a first aspect, an embodiment of the present application provides a multi-dimensional data fusion method for intelligent construction site water condition monitoring, which comprises the following steps:

[0006] At the monitoring point of the water condition to be monitored of the construction site, water level data, flow velocity data and rainfall in different collection time periods are collected respectively;

[0007] According to the difference between the water level data of the same monitoring point in the same collection time period, part of the water level data is selected to form a high-frequency water level data sequence;

[0008] The high-frequency water level data sequence is decomposed into a first preset number of layers using wavelet transform, and the instantaneous energy intensity of each detail coefficient in each detail coefficient sequence of each layer is obtained; the detail coefficients of the first second preset number of layers are processed according to the difference between the values of the detail coefficients and the instantaneous energy intensity of the detail coefficients; the denoised high-frequency water level data sequence is obtained according to the processing result of the detail coefficients of the first second preset number of layers and the detail coefficients of all other layers; and all water level data collected at the same monitoring point in the same period are combined to obtain all denoised water level data of the same monitoring point in the same period.

[0009] The flow rate data, rainfall and denoised water level data collected at the same monitoring point in the same period are subjected to multi-dimensional data fusion.

[0010] Further, the method for screening part of the water level data to form the high-frequency water level data sequence according to the difference between the water level data collected at different collection times of the same monitoring point in the same period comprises the following specific steps:

[0011] Any one of the water level data collected at the same monitoring point in the same period is recorded as target water level data, and a local window of the target water level data is established with the target water level data as the center;

[0012] The energy fluctuation degree of the target water level data is determined according to all the water level data in the local window of the target water level data;

[0013] The water level data is screened according to the energy fluctuation degree, and a high-frequency water level data sequence is formed.

[0014] Further, the method for determining the energy fluctuation degree comprises the following specific steps:

[0015] The variance of the instantaneous amplitude of each water level data in the local window of the target water level data is recorded as the energy fluctuation degree of the target water level data.

[0016] Further, the method for screening the water level data according to the energy fluctuation degree and forming the high-frequency water level data sequence comprises the following specific steps:

[0017] The energy fluctuation degrees of all the water level data in the same period are subjected to threshold segmentation, the water level data with an energy fluctuation degree greater than a threshold value is recorded as high-frequency water level data, and a sequence composed of adjacent high-frequency water level data is recorded as a high-frequency water level data sequence.

[0018] Further, the method for processing the detail coefficients of the first second preset number of layers according to the difference between the values of the detail coefficients and the instantaneous energy intensity of the detail coefficients comprises the following specific steps:

[0019] The high-frequency water level data sequence is decomposed into any one layer of the first second preset number of layers, and the layer is recorded as a target layer; any one detail coefficient in the detail coefficient sequence of the target layer is recorded as a target detail coefficient;

[0020] The dynamic threshold of the target detail coefficient is determined according to the difference between the instantaneous energy intensity of all the detail coefficients of the target layer and the instantaneous energy intensity of all the detail coefficients.

[0021] The processing of the detail coefficients of the target layer is realized according to the dynamic threshold of all the detail coefficients of the target layer.

[0022] Further, the method for determining the dynamic threshold is:

[0023] The first ratio of the target detail coefficient is determined according to the difference between the instantaneous energy intensity of all the detail coefficients of the target layer.

[0024] The kurtosis of all the detail coefficients in a local window with the target detail coefficient as the center and a length of is recorded as the kurtosis of the target detail coefficient, wherein represents the second preset length.

[0025] The positive correlation processing result of the first ratio and the kurtosis of the target detail coefficient is recorded as the dynamic threshold of the target detail coefficient.

[0026] Further, the method for determining the first ratio of the target detail coefficient is:

[0027] The ratio of the instantaneous energy intensity of the target detail coefficient to the median value of the instantaneous energy intensity of all the detail coefficients of the target layer is recorded as the first ratio of the target detail coefficient.

[0028] Further, the method for realizing the processing of the detail coefficients of the target layer according to the dynamic threshold of all the detail coefficients of the target layer includes:

[0029] The square of the difference between the dynamic thresholds of different detail coefficients is used as the distance between the different detail coefficients, the dynamic thresholds of all the detail coefficients of the target layer are clustered, three clustering clusters are obtained, and the detail coefficients in the clustering cluster with the maximum and minimum mean values of all the dynamic thresholds are assigned a value of 0.

[0030] Further, the method for obtaining all the water level data of the same period of the same monitoring point after denoising includes:

[0031] The values in the denoised high-frequency water level data sequence are used to replace the corresponding high-frequency water level data, and all the water level data of the same period of the same monitoring point after denoising are obtained.

[0032] In a second aspect, the embodiments of the present application also provide a multi-dimensional data fusion system for intelligent construction site water regime monitoring, comprising a memory, a processor, and a computer program stored in the memory and running on the processor, and the processor implements the steps of the method according to any one of the preceding aspects when executing the computer program.

[0033] The present application has the following beneficial effects:

[0034] The present application considers that when the precipitation is large, the water flow is relatively turbulent, which can entrain air into the water to form a large amount of bubbles, resulting in a deviation between the collected water level data and the actual water level data. The water level data is denoised, and specifically: according to the characteristics that the energy of the water level data presents a mutation, and the energy of the water level data affected by the bubbles is obviously larger, the water level data that may be affected by the bubbles is screened from the water level data, and a high-frequency water level data sequence is formed; further, considering that the bubble noise mainly presents as high-frequency pulses and is mainly located in the shallow wavelet, the high-frequency water level data sequence is decomposed into a first preset number of layers using wavelet transform, the water level data affected by the bubble noise has a larger value of instantaneous energy intensity, the water level data affected by the environmental noise has a smaller value of instantaneous energy intensity, and the actual water level data has a stable and moderate value of instantaneous energy intensity, according to the value of the instantaneous energy intensity of the water level data in the shallow wavelet, the detail coefficients affected by the bubble noise and the environmental noise are screened out, and the screened detail coefficients are processed to obtain the denoised high-frequency water level data sequence and the water level data; finally, the flow velocity data, the rainfall and the denoised water level data collected at the same monitoring point in the same period are subjected to multi-dimensional data fusion; the problem of poor denoising effect of the water level measurement signal affected by the bubbles in the water flow, resulting in inaccurate construction site water regime monitoring results based on multi-dimensional data is solved. BRIEF DESCRIPTION OF DRAWINGS

[0035] In order to more clearly illustrate the technical solutions in the embodiments of the present application or the prior art, the following will briefly introduce the drawings needed to be used in the embodiments or the prior art description. Obviously, the drawings in the following description are only some embodiments of the present application, and other drawings can also be obtained by those skilled in the art without creative labor.

[0036] Figure 1 A flowchart of a multi-dimensional data fusion method for intelligent construction site water regime monitoring provided by an embodiment of the present application;

[0037] Figure 2 A flowchart of a high-frequency water level data sequence acquisition process provided by an embodiment of the present application. DETAILED DESCRIPTION

[0038] With reference to the drawings of the embodiments of the present application, the technical solutions in the embodiments of the present application will be clearly and completely described. Obviously, the described embodiments are only a part of the embodiments of the present application, but not all the embodiments of the present application. Based on the embodiments of the present application, all the other embodiments obtained by a person of ordinary skill in the art without creative work fall within the scope of the present application.

[0039] Please refer to Figure 1 which shows a flow chart of a multi-dimensional data fusion method for intelligent construction site water condition monitoring provided by an embodiment of the present application. The method comprises the following steps:

[0040] Step S001, water level data, flow rate data and rainfall at different collection time points in each cycle are collected at the monitoring points of the water condition monitoring position of the construction site.

[0041] Monitoring points are set at the water condition monitoring position of the construction site, and ultrasonic water level monitors, flow rate monitors and rainfall sensors are arranged at each monitoring point. The ultrasonic water level monitors are used to collect water level data, the flow rate monitors are used to collect flow rate data of water flow, and the rainfall sensors are used to collect rainfall.

[0042] Among them, the monitoring points can be set at positions where water level changes are prone to occur in the upstream area of the construction site, downstream drainage pipes and other positions according to the setting of the monitoring points of the water condition monitoring position of the construction site by a person skilled in the art.

[0043] Preferably, in an embodiment of the present application, when collecting water level data, flow rate data and rainfall, water level data, flow rate data and rainfall are collected every 1 second in the embodiment, each collection cycle is set to 1 minute, and the time stamps of all collected water level data, flow rate data and rainfall are standardized. In actual application, as other implementation manners, the implementer can determine the sampling frequency and the length of a collection cycle according to actual conditions, and the present application does not make special limitations.

[0044] The water level data, flow rate data and rainfall are respectively denoised using low-pass filtering.

[0045] Among them, using low-pass filtering for denoising is a known technology and will not be described again.

[0046] At this point, the water level data, flow rate data and rainfall at different collection time points of the monitoring points of the water condition monitoring position of the construction site are obtained.

[0047] Step S002, according to the differences between the water level data at different collection time points of the same monitoring point in the same cycle, part of the water level data is screened to form a high-frequency water level data sequence.

[0048] When the precipitation is large, the water flow is turbulent, the water flow and flow rate of the upstream river area increase rapidly, forming turbulent water flow. When the high-speed water flow mixes with air or collides with obstacles, air is entrained into the water to form a large number of bubbles, resulting in a deviation between the collected water level data and the actual water level data.

[0049] Wavelet denoising has a good effect on separating high-frequency bubble noise. In this embodiment, wavelet denoising is used to remove high-frequency bubble noise in the water level data. When the water level data is processed using wavelet denoising, a too small threshold value will cause the high-frequency bubble pulse to be insufficiently suppressed, and a too large threshold value will cause the loss of real water level data, reducing the effectiveness and reliability of the processed water level data. Therefore, it is necessary to reasonably denoise the water level data to improve the data quality.

[0050] Any one of the water level data collected at the same monitoring point in the same period is recorded as target water level data. A local window with a length of is established around the target water level data and recorded as the local window of the target water level data.

[0051] wherein, represents the first preset length, and the value of the first preset length in this embodiment is 10; when there is a spare position in the local window of the target water level data, the data is filled using the mean value filling method; the data filling using the mean value filling method is a known technology and will not be described again.

[0052] The local window of the target water level data is processed using Hilbert transform to obtain the instantaneous amplitude of each water level data in the local window. The variance of the instantaneous amplitude of all water level data in the local window of the target water level data is recorded as the energy fluctuation degree of the target water level data.

[0053] The energy fluctuation degrees of all water level data in the same period can be obtained in the same way.

[0054] The energy fluctuation degrees of all water level data in the same period are threshold segmented, the water level data with an energy fluctuation degree greater than the threshold value is recorded as high-frequency water level data, and the sequence composed of adjacent high-frequency water level data is recorded as a high-frequency water level data sequence.

[0055] wherein, the maximum inter-class variance method is used to implement threshold segmentation in this embodiment, and the threshold segmentation using the maximum inter-class variance method is a known technology and will not be described again. In actual application, as other embodiments, on the basis of achieving the purpose of threshold segmentation, the implementer can use other methods such as the adaptiveThreshold function in OpenCV to implement threshold segmentation, and the present application does not make special limitation.

[0056] Thus, the high-frequency water level data sequence has been obtained. The flowchart for obtaining the high-frequency water level data sequence is as follows. Figure 2 As shown.

[0057] Step S003: Use wavelet transform to decompose the high-frequency water level data sequence into a first preset number of layers, and obtain the instantaneous energy intensity of each detail coefficient in the detail coefficient sequence of each layer. Based on the difference between the value of the detail coefficient and the instantaneous energy intensity of the detail coefficient in the first second preset number of layers, process the detail coefficients of the first second preset number of layers. Based on the processing results of the detail coefficients of the first second preset number of layers and the detail coefficients of all other layers, obtain the denoised high-frequency water level data sequence. Combine all water level data collected at the same monitoring point in the same period to obtain all denoised water level data of the same monitoring point in the same period.

[0058] The high-frequency water level data sequence is decomposed into a first preset number of layers using wavelet transform, and the instantaneous energy intensity of each detail coefficient within the detail coefficient sequence of each layer is obtained. Based on the differences between the values ​​of the detail coefficients and the instantaneous energy intensities of the detail coefficients in the first and second preset number of layers, the detail coefficients of the first and second preset number of layers are processed. Based on the processing results of the detail coefficients of the first and second preset number of layers and the detail coefficients of all other layers, a denoised high-frequency water level data sequence is obtained. Combining all water level data collected from the same monitoring point in the same period, all denoised water level data from the same monitoring point in the same period are obtained. The high-frequency water level data sequence is then processed using wavelet transform, and the wavelet basis db4 function is used to decompose the high-frequency water level data sequence into... The algorithm extracts the detail coefficient sequence for each layer and the instantaneous energy intensity of each detail coefficient within that sequence. Considering that bubble noise primarily manifests as high-frequency pulses and is mainly located in shallow wavelets, it extracts... The front of the layer The layer is used to eliminate bubble noise.

[0059] in, This represents the first preset quantity, which is 7 in this embodiment; The second preset quantity is indicated, and in this embodiment, the value of the second preset quantity is 3; the detail coefficient sequence is composed of detail coefficients, which correspond to the high-frequency components of different frequency bands.

[0060] Decompose the high-frequency water level data sequence into the following parts: Any layer in the layer is denoted as the target layer, and any detail coefficient in the detail coefficient sequence of the target layer is denoted as the target detail coefficient.

[0061] The ratio of the instantaneous energy intensity of the target detail factor to the median of the instantaneous energy intensities of all detail factors in the target layer is denoted as the first ratio of the target detail factor; a length of [missing information] is established centered on the target detail factor of the target layer. the kurtosis of all detail coefficients in the local window centered on the target detail coefficient is denoted as the kurtosis of the target detail coefficient; and the dynamic threshold of the target detail coefficient is determined according to the first ratio of the target detail coefficient and the kurtosis.

[0062] wherein, denotes the second preset length, and the value of the second preset length is 2 in the embodiment.

[0063] the positive correlation processing result of the first ratio and the kurtosis of the target detail coefficient is denoted as the dynamic threshold of the target detail coefficient.

[0064] It can be understood that the first ratio and the kurtosis of the target detail coefficient are positively correlated, that is, the first ratio and the kurtosis of the target detail coefficient are positively correlated with the dynamic threshold of the target detail coefficient, respectively. It can be understood that the positive correlation in the application refers to the relationship between the independent variable and the dependent variable, the independent variable is the first ratio and the kurtosis of the target detail coefficient, the dependent variable is the dynamic threshold of the target detail coefficient, and the positive correlation is that the dependent variable increases (decreases) as the independent variable increases (decreases), which can be an addition relationship, a multiplication relationship, etc.

[0065] Preferably, as an embodiment of the application, the sum of the normalized value of the kurtosis of the target detail coefficient and the number 1 is denoted as the first sum value of the target detail coefficient, and the product of the first sum value of the target detail coefficient and the first ratio is denoted as the dynamic threshold of the target detail coefficient.

[0066] It should be noted that the Z-Score standard normalization method is used to calculate the normalized value in the embodiment, and other methods such as the maximum and minimum value normalization method, the sigmoid function and other methods in the prior art can be used to calculate the normalized value in actual application, which is not limited herein.

[0067] wherein, the first ratio of the target detail coefficient is calculated according to the median of the instantaneous energy intensity of all detail coefficients of the target layer, which can avoid the influence of local maximum in the target layer on the dynamic threshold calculation process. Generally, the instantaneous energy intensity of real water level data is stable and moderate, the instantaneous energy intensity of water level data affected by bubble noise is large, and the instantaneous energy intensity of water level data affected by environmental noise is small, so the dynamic threshold of the target detail coefficient determined according to the median of the instantaneous energy intensity of all detail coefficients of the target layer is more accurate.

[0068] Further, the kurtosis of the water level data affected by the bubble noise is large, the first ratio is large, and the dynamic threshold of the corresponding detail coefficient is large.

[0069] The high-frequency water level data sequence decomposed into the first The dynamic threshold of each detail coefficient of each layer.

[0070] The square of the difference between the dynamic thresholds of different detail coefficients is taken as the distance between the different detail coefficients, the dynamic thresholds of the same layer of detail coefficients are clustered, three cluster clusters are obtained, and the detail coefficients corresponding to the dynamic thresholds in the three cluster clusters are marked as bubble noise, effective signal and environmental noise in order from large to small according to the mean of the dynamic thresholds contained in the cluster cluster. The detail coefficients marked as bubble noise and environmental noise are assigned a value of 0, and the processing of the detail coefficients of the first layer of the high-frequency water level data sequence decomposed is realized.

[0071] In this embodiment, the cluster cluster is obtained using k-means clustering.

[0072] The detail coefficients of the last layer of the high-frequency water level data sequence decomposed are processed using wavelet denoising, and the high-frequency water level data sequence decomposed after processing the detail coefficients is inverse wavelet transformed to obtain the denoised high-frequency water level data sequence.

[0073] In this embodiment, the threshold processing of the detail coefficients and the processing of the detail coefficients using wavelet denoising, and the inverse wavelet transform to obtain the denoised data are all known technologies and will not be described in detail.

[0074] The denoised high-frequency water level data sequence is used to replace the corresponding high-frequency water level data to obtain all the water level data of the same period after denoising.

[0075] At this point, the denoising of all the water level data of the same period is realized.

[0076] Step S004, the flow rate data, rainfall and denoised water level data collected at the same period for the same monitoring point are multi-dimensional data fusion.

[0077] The flow rate data, rainfall and denoised water level data collected at the same period for the same monitoring point are multi-dimensional data fusion using a fuzzy logic system and a Mamdani model, and the aggregated fuzzy set is obtained. The aggregated fuzzy set is defuzzified, the centroid method is used to calculate the centroid coordinates of the aggregated fuzzy set, the multi-dimensional data fusion using the fuzzy logic system is completed, and the fusion result of the multi-dimensional data collected at the same period for the same monitoring point is obtained.

[0078] The multi-dimensional data of the embodiment specifically includes flow rate data, rainfall data and denoised water level data. In the process of multi-dimensional data fusion, specifically, the embodiment uses a triangular membership function as a fuzzy method of input variables, sets IF-THEN fuzzy rules by experts in the field, constructs a rule base, and uses a Mamdani model for fuzzy reasoning.

[0079] Thus, the fusion of multi-dimensional data of the water regime monitoring position of a construction site is realized.

[0080] Based on the same inventive concept as the above method, the embodiment of the present application also provides a multi-dimensional data fusion system for intelligent water regime monitoring of a construction site, which comprises a memory, a processor and a computer program stored in the memory and running on the processor, and the processor implements the steps of any one of the above-mentioned methods for multi-dimensional data fusion for intelligent water regime monitoring of a construction site.

[0081] The above description is only the preferred embodiment of the present application, and is not intended to limit the present application. Any modification, equivalent replacement, improvement, etc. made within the principles of the present application shall be included in the protection scope of the present application.

Claims

1. A multi-dimensional data fusion method for water condition monitoring in smart construction sites, characterized in that, The method includes the following steps: At the monitoring points at the construction site where water conditions are to be monitored, water level data, flow velocity data, and rainfall data are collected at different times within each cycle. Based on the differences in water level data at different collection times within the same period at the same monitoring point, a portion of the water level data is selected to form a high-frequency water level data sequence. The high-frequency water level data sequence is decomposed into a first preset number of layers using wavelet transform, and the instantaneous energy intensity of each detail coefficient in the detail coefficient sequence of each layer is obtained. Based on the difference between the instantaneous energy intensity of the detail coefficients in the first preset number of layers and the values ​​of the detail coefficients, the detail coefficients of the first and second preset number of layers are processed. Based on the processing results of the detail coefficients of the first and second preset number of layers and the detail coefficients of all other layers, a denoised high-frequency water level data sequence is obtained. Combined with all water level data collected from the same monitoring point in the same period, all denoised water level data of the same monitoring point in the same period are obtained. Multidimensional data fusion was performed on flow velocity data, rainfall data, and denoised water level data collected from the same monitoring point in the same period; The specific method for processing the detail coefficients of the first second preset number of layers based on the difference between the value of the detail coefficients and the instantaneous energy intensity of the detail coefficients includes: Decompose the high-frequency water level data sequence into any one of the first two preset number layers and denote it as the target layer. Denote any one detail coefficient in the detail coefficient sequence of the target layer as the target detail coefficient. The dynamic threshold of the target detail factor is determined based on the differences in the instantaneous energy intensity of all detail factors in the target layer. Based on the dynamic thresholds of all detail coefficients in the target layer, the detail coefficients of the target layer are processed. The method for determining the dynamic threshold is as follows: The first ratio of the target detail coefficients is determined based on the difference in instantaneous energy intensity of all detail coefficients in the target layer; Centered on the target detail coefficient, with a length of The kurtosis of all detail coefficients within a local window is denoted as the kurtosis of the target detail coefficient, where... Indicates the second preset length; The positive correlation between the first ratio of the target detail coefficient and the kurtosis is recorded as the dynamic threshold of the target detail coefficient. The method for determining the first ratio of the target detail coefficients is as follows: The ratio of the instantaneous energy intensity of the target detail factor to the median of the instantaneous energy intensity of all detail factors in the target layer is denoted as the first ratio of the target detail factor.

2. The multi-dimensional data fusion method for water condition monitoring in smart construction sites according to claim 1, characterized in that, The method for selecting a portion of water level data to form a high-frequency water level data sequence based on the differences between water level data collected at different times within the same period from the same monitoring point includes the following specific methods: Any water level data collected from the same monitoring point within the same period is recorded as the target water level data. A local window of the target water level data is established with the target water level data as the center. The degree of energy fluctuation of the target water level data is determined based on all water level data within a local window of the target water level data. Water level data are filtered based on the degree of energy fluctuation and a high-frequency water level data sequence is constructed.

3. The multi-dimensional data fusion method for water condition monitoring in smart construction sites according to claim 2, characterized in that, The method for determining the degree of energy fluctuation is as follows: The variance of the instantaneous amplitude of each water level data point within a local window of the target water level data is denoted as the energy fluctuation degree of the target water level data.

4. The multi-dimensional data fusion method for water condition monitoring in smart construction sites according to claim 3, characterized in that, The specific method for filtering water level data based on the degree of energy fluctuation and constructing a high-frequency water level data sequence includes: The energy fluctuation of all water level data in the same period is divided by a threshold. Water level data with energy fluctuation greater than the threshold is recorded as high-frequency water level data. The sequence of adjacent high-frequency water level data is recorded as high-frequency water level data sequence.

5. The multi-dimensional data fusion method for water condition monitoring in smart construction sites according to claim 1, characterized in that, The specific method for processing the detail coefficients of the target layer based on the dynamic threshold of all detail coefficients of the target layer includes: The square of the difference between the dynamic thresholds of different detail coefficients is used as the distance between the different detail coefficients. The dynamic thresholds of all detail coefficients in the target layer are clustered to obtain three clusters. The detail coefficients in the clusters with the largest and smallest mean of all dynamic thresholds are assigned a value of 0.

6. The multi-dimensional data fusion method for water condition monitoring in smart construction sites according to claim 1, characterized in that, The specific method for obtaining all water level data of the same monitoring point and the same period by combining all water level data collected at the same monitoring point in the same period includes: Replace the corresponding high-frequency water level data with each value in the denoised high-frequency water level data sequence to obtain all water level data of the same monitoring point and the same period after denoising.

7. A multi-dimensional data fusion system for water condition monitoring at smart construction sites, 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 implements the steps of the method as claimed in any one of claims 1-6.

Images