Construction environment intelligent monitoring method and monitoring system based on Internet of Things

Abstract

The invention provides a construction environment intelligent monitoring method and monitoring system based on the Internet of Things. The method comprises the following steps: acquiring a to-be-analyzed environment sensing data set; when the environment sensing data set comprises a first event, a first local relation network is obtained in an event relation network, and the first local relation network comprises a first composition item, composition items in a preset path length range and connecting lines between the composition items; obtaining a first relation embedding array based on a first feature extraction component in the construction monitoring neural network according to the first local relation network; acquiring a sensing embedded array based on a second feature extraction component in the construction monitoring neural network according to the environment sensing data set; generating an integrated embedding array according to the first relation embedding array and the sensing embedding array; and obtaining a construction monitoring result based on a decision component in the construction monitoring neural network according to the integrated embedded array. According to the method and the device, the problems of context missing and inaccurate identification caused by short content of the data stream detected in real time are solved.

Description

Technical Field

[0001] The present application relates to the field of data processing, and specifically, to an intelligent monitoring method and monitoring system for a construction environment based on the Internet of Things. Background Art

[0002] In various construction fields such as buildings, roads, and bridges, accurate monitoring of the construction site environment is a key link to ensure construction safety, project quality, and improve construction efficiency. Traditional construction environment monitoring methods mainly rely on manual inspections and simple sensor data collection. Manual inspections have problems such as low monitoring frequency, strong subjectivity, and difficulty in obtaining comprehensive information in real time, and are unable to promptly detect potential risks and abnormal situations that occur during the construction process. Although simple sensor data collection can provide certain real-time data, the data streams it obtains are often short and lack sufficient contextual information, which limits the understanding and judgment of the construction environment. Summary of the invention

[0003] The purpose of the present invention is to provide a construction environment intelligent monitoring method and monitoring system based on the Internet of Things. This application is implemented as follows: In a first aspect, the present application provides an intelligent monitoring method for a construction environment based on the Internet of Things, comprising: obtaining an environmental sensor data set to be analyzed; when the environmental sensor data set includes a first event, obtaining a first local relationship network in the event relationship network, wherein each component in the event relationship network represents an event, each connecting line in the event relationship network represents an event relationship, the first event belongs to the event represented by the first component in the event relationship network, and the first local relationship network includes the first component, components within a preset path length range obtained based on the first component, and connecting lines between components; based on the first local relationship network, obtaining a first relationship embedding array based on a first feature extraction component in a construction monitoring neural network; based on the environmental sensor data set, obtaining a sensor embedding array based on a second feature extraction component in the construction monitoring neural network; based on the first relationship embedding array and the sensor embedding array, generating an integrated embedding array; based on the integrated embedding array, obtaining a construction monitoring result corresponding to the environmental sensor data set based on a decision component in the construction monitoring neural network.

[0004] In a second aspect, the present application provides a monitoring system, comprising: one or more processors; a memory; one or more computer programs; wherein the one or more computer programs are stored in the memory and configured to be executed by the one or more processors, and when the one or more computer programs are executed by the processors, the method described above is implemented.

[0005] In the method for intelligent monitoring of construction environment based on the Internet of Things provided by the present application, an environmental sensor data set to be analyzed is obtained. If it is identified that the environmental sensor data set includes a first event, a first local relationship network is obtained in the event relationship network, each component in the event relationship network represents an event, and each connecting line in the event relationship network represents an event relationship. The first event belongs to the event represented by the first component in the event relationship network, and the first local relationship network includes the first component, the components within the preset path length range obtained based on the first component, and the connecting lines between the components. Based on this, the present application obtains a first relationship embedding array based on the first feature extraction component in the construction monitoring neural network based on the first local relationship network. At the same time, based on the environmental sensor data set, a sensor embedding array is obtained based on the second feature extraction component in the construction monitoring neural network. Afterwards, an integrated embedding array is generated based on the first relationship embedding array and the sensor embedding array. Finally, based on the integrated embedding array, the construction monitoring result corresponding to the environmental sensor data set is obtained based on the decision component in the construction monitoring neural network. Based on the above process, the solution of the present application combines the events and event relationships in the event relationship network to generate a local relationship network, and improves the context of the environmental sensor data set through the local relationship network, thereby overcoming the problem of missing context and inaccurate recognition caused by the short content of the data stream detected in real time.

[0006] Other features will be described in part in the following description. Those skilled in the art will discover these features in part upon inspection of the following content and drawings, or may learn these features through production or use. The features of the present application may be implemented and obtained by practicing or using various aspects of the methods, tools, and combinations listed in the detailed examples described below. BRIEF DESCRIPTION OF THE DRAWINGS

[0007] In order to more clearly illustrate the technical solutions in the embodiments of the present application, the drawings required for use in describing the embodiments of the present application are briefly introduced below.

[0008] Figure 1 It is a flow chart of an intelligent monitoring method for construction environment based on the Internet of Things provided in an embodiment of the present application.

[0009] Figure 2 It is a schematic diagram of the composition of a monitoring system provided in an embodiment of the present application. DETAILED DESCRIPTION

[0010] The following describes the embodiments of the present application in conjunction with the drawings in the embodiments of the present application. The terms used in the implementation method part of the embodiments of the present application are only used to explain the specific embodiments of the present application, and are not intended to limit the present application.

[0011] The executor of the construction environment intelligent monitoring method based on the Internet of Things in the embodiment of the present application is a monitoring system, including but not limited to a server, a personal computer, a laptop computer, a tablet computer, a smart phone, etc. The monitoring system can be operated alone to implement the present application, or it can be connected to the network and implement the present application through interactive operations with other monitoring systems in the network. Among them, the network where the monitoring system is located includes but is not limited to the Internet, a wide area network, a metropolitan area network, a local area network, a VPN network, etc. The monitoring system can be used to store environmental sensing data collected by a sensor cluster at the construction site.

[0012] The embodiment of the present application provides an intelligent monitoring method for construction environment based on the Internet of Things, such as Figure 1 As shown, the following steps are included: Step S100: Acquire an environmental sensor data set to be analyzed.

[0013] In step S100, the environmental sensor data set to be analyzed includes a variety of data collected by various environmental sensors at the construction site, which can reflect the environmental status and related event information of the construction site. At the construction site, the environmental sensor data set is collected by different types of sensors. For example, the temperature sensor collects the temperature data of the construction site, the humidity sensor obtains the air humidity information, the light sensor records the light intensity, the noise sensor monitors the noise level generated during the construction process, the vibration sensor can detect the vibration of the equipment, and the distance sensor can detect the locations in the construction scene that have distance requirements, such as wall gaps, the distance between personnel and high-risk equipment, etc.

[0014] For sensors with wired connections, the data interface can be used to directly connect to the sensor and receive the data sent by the sensor according to the preset communication protocol. For example, using the RS-485 bus protocol, the sensor encodes the collected data according to the format of the protocol, and then transmits it to the sensor through the bus, and then parses the data according to the corresponding decoding rules. For wireless sensor networks, wireless communication technologies such as Wi-Fi and ZigBee can be used to receive sensor data. Taking Wi-Fi as an example, the sensor encapsulates the collected data into data packets that comply with the Wi-Fi protocol and sends them out through wireless signals. After receiving these data packets, they are unpacked and processed to obtain the environmental sensing data contained therein.

[0015] Step S200: When the environmental sensing data set includes a first event, a first local relationship network is obtained in the event relationship network, wherein each component in the event relationship network represents an event, each connecting line in the event relationship network represents an event relationship, the first event belongs to the event represented by the first component in the event relationship network, and the first local relationship network includes the first component, components within a preset path length range obtained based on the first component, and connecting lines between the components.

[0016] In step S200, when the environmental sensing data set includes the first event, a first local relationship network is obtained in the event relationship network. The core of this step is to use the information of the event relationship network to improve the context of the environmental sensing data set, thereby solving the problem of missing context and inaccurate monitoring of the construction site caused by the short real-time detection data stream. The event relationship network is a structured graph structure, in which each component, i.e., a graph vertex, represents an event, and each connecting line represents an event relationship. These events and event relationships constitute a complex network structure, reflecting the internal connection between various events on the construction site.

[0017] In the actual scene of the construction site, the first event may be a specific construction activity, such as "concrete pouring". First, we need to determine the corresponding component of the event in the event relationship network, that is, find the first component representing the event "concrete pouring". The events in the event relationship network can cover all aspects of the construction process, such as equipment operation, material transportation, personnel activities, etc., and each event may be related to multiple other events.

[0018] When obtaining the first local relationship network, other components related to the first component will be determined based on the preset path length. The preset path length refers to the number of connecting lines from one component to another, which limits the scope of the local relationship network. Taking the "concrete pouring" event as an example, when the preset path length is 2, events directly connected to "concrete pouring" and events connected to these directly connected events will be searched. Directly connected events may include "concrete mixing" and "formwork installation", while events connected to "concrete mixing" may include "cement transportation" and "sand and gravel supply". These events and the connecting lines between them together constitute the first local relationship network.

[0019] In order to obtain the first local relationship network, a graph traversal algorithm can be used. A graph traversal algorithm is an algorithm for accessing all nodes and edges in a graph, such as breadth-first search (BFS) and depth-first search (DFS). Taking the breadth-first search algorithm as an example, the first component is first used as the starting node, marked as visited, and added to the queue. Then, a node is taken out of the queue and all its adjacent nodes are checked. If the adjacent node has not been visited and the path length with the starting node is not greater than the preset path length, it is marked as visited and added to the queue, and the connection line information between the nodes is recorded. Repeat this process until the queue is empty. In this way, all components and connection lines that meet the conditions can be found to generate the first local relationship network.

[0020] The generation of the first local relationship network can provide richer context information and help the system better understand the first event and other events related to it.

[0021] When detecting an event, taking the detection of the event of concrete pouring from the environmental sensor data set as an example, different types of environmental sensors can be used to obtain relevant data, and the event can be identified by analyzing and processing these data. First, vibration sensors can be used to detect vibration during concrete pouring. When pouring concrete, the concrete is transported from the mixer truck to the pouring site, and the vibration equipment vibrates the concrete. Operations such as vibration will generate vibration. Vibration sensors can convert these vibration signals into electrical signals and record them. For example, vibration sensors are installed on the formwork, support structure and other parts of the construction site. When the concrete starts to be poured and the vibration equipment is turned on, the vibration sensor will detect a vibration signal of a specific frequency and intensity. By performing spectrum analysis on these vibration signals, their main frequency components and intensity ranges are determined. Generally speaking, the vibration frequency during concrete pouring will be concentrated in a certain range, and the intensity will also have a certain change pattern as the vibration progresses. A vibration intensity threshold can be set. When the detected vibration intensity exceeds the threshold and the vibration frequency is within the preset frequency range, it can be preliminarily determined that concrete pouring may be in progress.

[0022] Secondly, pressure sensors can also provide important detection basis. During the concrete pouring process, as the concrete continues to accumulate, pressure changes will occur on the formwork, supporting structure, etc. Pressure sensors are installed at key locations such as the bottom of the formwork and support columns to monitor the changes in pressure in real time. When concrete begins to be poured, the pressure will gradually increase, and the rate of increase is related to factors such as the pouring speed and pouring volume of the concrete. A pressure change model can be established to calculate the pressure change curve over time under normal pouring conditions based on the concrete pouring process and design requirements. The actual detected pressure change data is compared with the model. If the change trends of the two are similar and the pressure value is within a reasonable range, the occurrence of the concrete pouring event can be further confirmed. For example, for a formwork of a specific size and shape, the theoretical change value of the pressure at the bottom of the formwork during the pouring process can be calculated based on parameters such as the density of the concrete and the pouring speed. When the actual detected pressure value is close to this theoretical value, it can be determined that the concrete is being poured as expected.

[0023] Temperature sensors can also be used to detect concrete pouring events. During the concrete pouring process, the hydration reaction of cement releases heat, causing the temperature of the concrete to rise. Temperature sensors are placed inside the concrete and in the surrounding environment to monitor temperature changes in real time. When the concrete begins to be poured, the internal temperature will gradually rise, and the magnitude and speed of the rise are related to factors such as the type and amount of cement, and the ambient temperature. A temperature change threshold can be set. When it is detected that the internal temperature of the concrete rises above the threshold in a short period of time, it can be inferred that concrete pouring may be in progress. For example, under normal temperature conditions, the internal temperature of ordinary Portland cement concrete may rise by 5-10 degrees Celsius within a few hours after pouring. If it is detected that the internal temperature of the concrete has a similar increase in this time period, it can be used as a basis for judging concrete pouring.

[0024] Finally, the data collected by the above-mentioned various sensors are fused and comprehensively analyzed. Multi-sensor data fusion technology can be used to process different types of data such as vibration, pressure, temperature, etc., remove noise and interference, and extract effective information related to the concrete pouring event. By establishing a comprehensive judgment model, the concrete pouring event is finally judged according to the weights and combination rules of different sensor data. For example, a fuzzy logic algorithm can be used to assign corresponding weights according to the credibility and importance of different sensor data, and fuzzy reasoning can be performed on these data to obtain a comprehensive judgment result. When this result exceeds a preset threshold, it can be determined that the concrete pouring event is occurring.

[0025] Step S300: According to the first local relationship network, a first relationship embedding array is obtained based on a first feature extraction component in the construction monitoring neural network.

[0026] In step S300, the events and their relationship information contained in the first local relationship network are converted into a numerical form that can be processed and analyzed, so as to make more in-depth calculations and decisions. The first local relationship network is composed of the components related to the first event in the event relationship network and the connecting lines between them, which reflects the association and context information between the various events in the construction site. The graph embedding technology can be used to obtain the first relationship embedding array, and the components and connecting lines in the graph structure (such as the first local relationship network) are mapped to a low-dimensional vector space, so that nodes with similar structures and relationships in the graph also have similar representations in the vector space. For example, graph embedding algorithms include DeepWalk, Node2Vec, etc.

[0027] Taking the DeepWalk algorithm as an example, it generates node sequences in the first local relationship network by random walks, and then uses these node sequences as text sequences to learn the vector representation of the components using word embedding algorithms (such as the Skip-Gram model). The specific steps are as follows: First, a random walk is performed from each component in the first local relationship network to generate a series of component sequences. The random walk process is similar to randomly selecting a path in a graph, walking from one component to an adjacent component, and repeating this process until the preset walk length is reached. For example, starting from the "tower crane operation" component, it may randomly walk to the "material lifting" node, and then walk from the "material lifting" component to the "equipment inspection" component, forming a component sequence "tower crane operation-material lifting-equipment inspection". Then, these component sequences are used as sensor data sequences, and the Skip-Gram model is used to learn the vector representation of the components. The Skip-Gram model predicts the context node of a component and learns the vector representation of the component by maximizing the probability of this prediction. Specifically, for each component in a sensor data sequence, the model predicts the nodes within a certain range before and after the component. By continuously optimizing the model parameters, the prediction probability is maximized, and finally the vector representation of each component is obtained.

[0028] After obtaining the vector representation of each component item, these vectors are integrated to obtain a first relation embedding array, for example, averaging or weighted averaging the vectors of all component items.

[0029] The first relation embedding array contains important information about events and their relations in the first local relation network. It converts the graph structure information into a numerical vector to facilitate subsequent calculation and analysis.

[0030] Step S400: Acquire a sensor embedding array based on the second feature extraction component in the construction monitoring neural network according to the environmental sensor data set.

[0031] In step S400, valuable feature information is extracted from the environmental sensing data set and converted into a vector form suitable for processing and analysis, so as to be subsequently integrated with the first relational embedding array, thereby providing a basis for accurate construction monitoring results. The environmental sensing data set is composed of data collected by various environmental sensors at the construction site, which contains various environmental information of the construction site, such as temperature, humidity, light, noise, vibration, etc.

[0032] The second feature extraction component is part of the construction monitoring neural network, which can be a convolutional neural network (CNN) or a recurrent neural network (RNN) in deep learning. When using a convolutional neural network for feature extraction, the environmental sensing dataset is preprocessed to convert it into a format suitable for the convolutional neural network input. For example, if the environmental sensing dataset is a series of time series data, it can be organized into a two-dimensional matrix, where each row represents the sensor data at a time point and each column represents the measurement value of a sensor. The convolutional neural network then performs a convolution operation on the input data through a convolutional layer, and the convolution kernel in the convolutional layer slides on the input data to extract features of different scales and locations. Each convolution kernel generates a feature map that reflects the response of the input data on different feature dimensions. After the convolutional layer, a pooling layer can be cascaded to downsample the feature map to reduce the dimension of the data while retaining important feature information.

[0033] After being processed by the convolutional layer and the pooling layer, a series of feature maps are obtained, which contain important feature information in the environmental sensing data set. Finally, these feature maps are converted into a vector through the fully connected layer, which is the sensor embedding array. The fully connected layer connects all neurons in each feature map with neurons in the next layer, and integrates the information of the feature map into a vector through linear transformation and nonlinear activation function.

[0034] As mentioned above, in addition to convolutional neural networks, recurrent neural networks (RNNs) can also be used to extract feature information from environmental sensing data sets. The principle can be referred to in the prior art and will not be described here.

[0035] The sensor embedding array contains the key feature information in the environmental sensor data set. It converts the environmental sensor data into a numerical vector to facilitate subsequent calculation and analysis.

[0036] Step S500: Generate an integrated embedded array based on the first relation embedded array and the sensor embedded array.

[0037] Step S500 aims to effectively integrate the event association information contained in the first relationship embedding array extracted from the event relationship network with the construction site environmental information contained in the sensor embedding array extracted from the environmental sensor data set, thereby laying the foundation for obtaining more accurate construction monitoring results in the future. The first relationship embedding array reflects the relationship and context information between construction site events, while the sensor embedding array reflects various environmental status data of the construction site. By generating an integrated embedding array, the two aspects of information can be comprehensively utilized to improve the accuracy and reliability of construction environment monitoring.

[0038] One implementation method is to generate an integrated embedded array through matrix operations. The sensor embedded array is multiplied by the first weight array to obtain a first product result. The first weight array is a two-dimensional array (matrix), which is used to weight each element in the sensor embedded array to highlight the importance of different environmental characteristics. For example, at a construction site, if the temperature has a greater impact on the construction quality, the weight of the corresponding temperature feature in the first weight array may be set higher. Then, the result of the product of the first relationship embedded array and the second weight array is flipped to obtain a flipped result. The second weight array is also used to weight the elements in the first relationship embedded array to reflect the importance of different event relationships. Next, the first product result is multiplied by the flip result to obtain a second product result. Finally, the second product result is added to the bias array to obtain an integrated embedded array. Among them, the first weight array and the second weight array are obtained by extracting the third-order tensor, the dimension of the third-order tensor is i×s×t, the dimension of the first weight array is i×1×t, and the dimension of the second weight array is 1×s×t, i is the dimension of the sensor embedding array, s is the dimension of the first relational embedding array, and t is the dimension of the integrated embedding array.

[0039] Another implementation method is to combine the first relational embedding array and the sensor embedding array to obtain an integrated embedded array. This method is simple and direct, and the two arrays are connected in a certain order to form a new array. For example, if the dimension of the sensor embedding array is i and the dimension of the first relational embedding array is s, then the dimension of the integrated embedded array is i + s. This method can retain the original information of the two arrays, but may result in a higher dimension of the integrated embedded array, increasing the complexity of subsequent calculations.

[0040] Another implementation method is to obtain the Hadamard product of the first relational embedding array and the sensor embedding array to obtain an integrated embedding array. The Hadamard product is an operation of multiplying corresponding elements, requiring the dimensions of the two arrays to be equal. Through the Hadamard product, the corresponding elements in the first relational embedding array and the sensor embedding array can be multiplied point by point to obtain a new array with the same dimension as the original array. This method can effectively fuse the information in the two arrays and highlight the interaction between the corresponding elements in the two arrays.

[0041] When the environmental sensing data set also includes a second event, it is necessary to obtain a second local relationship network in the event relationship network, and obtain a second relationship embedding array based on it. At this time, the process of generating an integrated embedding array will be more complicated. It is necessary to comprehensively consider the information of the first relationship embedding array, the second relationship embedding array, and the sensor embedding array. For example, the first relationship embedding array and the second relationship embedding array can be processed first, such as performing a Hadamard product with a preset parameter array respectively to obtain a first intermediate result and a second intermediate result, and then the two intermediate results are fused to obtain a target relationship embedding array. Then, the above-mentioned matrix operation, combination or Hadamard product method can be used to integrate the target relationship embedding array with the sensor embedding array to generate the final integrated embedding array.

[0042] Step S600: According to the integrated embedded array, the construction monitoring result corresponding to the environmental sensor data set is obtained based on the decision component in the construction monitoring neural network.

[0043] The integrated embedding array is generated by fusing the first relation embedding array and the sensor embedding array. It integrates the event association information in the event relation network and the field environment information in the environmental sensor dataset, providing comprehensive and effective input data for the decision-making component.

[0044] When executing step S600, a fully connected neural network can be used as a decision component. The fully connected neural network consists of an input layer, a hidden layer, and an output layer, wherein the input layer receives an integrated embedded array as input data, the hidden layer processes and transforms the input data through a series of neurons, and the output layer outputs the final construction monitoring result.

[0045] In a fully connected neural network, each neuron performs a weighted summation of the input data and performs a nonlinear transformation through an activation function. Assuming that there are n neurons in the input layer, corresponding to the n elements of the integrated embedding array, the input of the jth hidden layer neuron can be expressed as: ;in, is the weight from the i-th neuron in the input layer to the j-th neuron in the hidden layer, is the input value of the ith neuron in the input layer (i.e., the ith element of the integrated embedding array), is the bias term of the jth neuron in the hidden layer.

[0046] Then, the output of the jth neuron in the hidden layer is It can be calculated by the activation function f: , the activation function is, for example, Sigmoid function, ReLU function, etc.

[0047] After being processed by the hidden layer, the neurons in the output layer will perform weighted summation on the output of the hidden layer and output the final construction monitoring results.

[0048] The final construction monitoring result can be judged based on the output value of the output layer neuron. For example, if there is only one neuron in the output layer, and its output value represents the probability of a safety hazard at the construction site, then when the output value is greater than a preset threshold (such as 0.5), it can be judged that there is a safety hazard at the construction site; when the output value is less than the threshold, the construction site is judged to be safe.

[0049] The construction monitoring results can include various types of information, such as whether the construction progress is normal, whether there are hidden dangers in construction safety, whether the construction environment meets the requirements, etc. For example, by analyzing the integrated embedded array, the decision component can determine whether the current construction progress is behind the planned progress. If the environmental parameters such as temperature and humidity exceed the range required by the construction process, the decision component can issue an alarm for environmental anomalies; if some key events in the event relationship network do not occur in the normal order, the decision component can indicate that there may be construction risks.

[0050] The integrated embedded array is analyzed and judged through the decision-making component of the construction monitoring neural network, which provides an important basis for the management and decision-making of the construction site, helps to timely discover problems and risks in the construction process, and ensures the smooth progress of the construction and the safety of personnel.

[0051] As an implementation mode, after obtaining the environmental sensing data set to be analyzed, the method further includes an event detection process, specifically including: Step S101: performing event detection on the environmental sensor data set to obtain event detection results; Step S102: when the event detection result indicates that the environmental sensor data set includes an event, the event in the environmental sensor data set is compared with the event in the event relationship network to obtain an event comparison result; Step S103: When the event comparison result indicates that the event in the environmental sensor data set corresponds to the event in the event relationship network, the corresponding event is used as the first event.

[0052] In step S101, event detection is performed on the environmental sensor data set to obtain event detection results. The environmental sensor data set is composed of multi-source data collected by various sensors at the construction site. These data contain various physical information of the construction site, such as temperature, humidity, noise, vibration, etc. The purpose of event detection is to identify events with specific meanings from these complex data, such as the start-up of construction equipment, the transportation of materials, the activities of personnel, etc.

[0053] One method of event detection is a threshold-based detection method. For certain types of sensor data, such as temperature, noise, etc., a reasonable threshold range can be set. When the data collected by the sensor exceeds this threshold range, it is considered that an event may have occurred. For example, at a construction site, the normal ambient temperature range may be between 20℃-30℃. If the temperature detected by the temperature sensor suddenly rises to 40℃, it can be determined that events such as equipment overheating or fire may have occurred. The setting of the threshold can be obtained based on statistical analysis of historical data, or it can be manually set according to construction technology and safety standards.

[0054] Another method is an event detection method based on pattern recognition. The sensor data patterns corresponding to various events can be learned in advance, and then the environmental sensor data collected in real time can be matched with these patterns. For example, when the construction equipment is started, the vibration sensor will detect a vibration signal of a specific frequency and amplitude. By analyzing the characteristics of the vibration signal, such as frequency distribution, amplitude change, etc., it can be determined whether the equipment is started. Pattern recognition can use machine learning algorithms such as support vector machines (SVM), decision trees, etc., or deep learning algorithms such as convolutional neural networks (CNN), recurrent neural networks (RNN), etc.

[0055] Taking support vector machine as an example, its basic idea is to find an optimal hyperplane in the feature space to separate event data of different categories. Assuming that the environmental sensor data can be represented as a feature vector x, and the event category can be represented as a label y (y = ±1), the goal of the support vector machine is to find a weight vector w and a bias b to maximize the classification interval. The specific optimization problem can be expressed as: ; Here, n is the number of samples.

[0056] In step S102, when the event detection result indicates that the environmental sensor data set includes an event, the event in the environmental sensor data set is compared with the event in the event relationship network to obtain an event comparison result. The event relationship network is a pre-constructed network structure, in which each component represents an event and the connecting line represents the relationship between the events.

[0057] When performing event comparison, a method based on semantic matching can be used. For each detected event, its key information can be extracted, such as the name of the event, the time of occurrence, the related equipment, etc., and then this information can be compared with the event information in the event relationship network. For example, if the detected event is "tower crane start", the components related to "tower crane start" can be found in the event relationship network. The comparison process can use natural language processing technology, such as lexical analysis, syntactic analysis, etc., to convert the event description into a computer-processable form, and then perform similarity calculation. Similarity calculation can use a variety of methods, such as edit distance, cosine similarity, etc. Edit distance refers to the minimum number of operations required to convert two strings into another string through insertion, deletion and substitution operations. Another comparison method is rule-based comparison, which can pre-define some comparison rules. For example, if the detected event involves a specific device, then find events related to the device in the event relationship network.

[0058] In step S103, when the event comparison result shows that the event in the environmental sensor data set corresponds to the event in the event relationship network, the corresponding event is used as the first event.

[0059] As an implementation mode, in step S102, when the event detection result indicates that the environmental sensor data set includes an event, the event in the environmental sensor data set is compared with the event in the event relationship network to obtain an event comparison result, including: Step S1021: when the event detection result indicates that the environmental sensor data set includes multiple events, obtaining the conditional probability corresponding to each of the multiple events; Step S1022: taking the event corresponding to the maximum conditional probability among the multiple events as the event in the environmental sensing data set; Step S1023: Compare the events in the environmental sensor data set with the events in the event relationship network to obtain event comparison results.

[0060] In step S1021, when the event detection result indicates that the environmental sensing data set includes multiple events, the conditional probability corresponding to each event in the multiple events is obtained. Conditional probability refers to the probability of an event occurring under certain known conditions. In the context of environmental sensing data sets, these conditions can be data collected by sensors, time information, context of events, etc. By calculating conditional probabilities, the possibility of each event occurring can be quantified, thereby providing a basis for subsequent event screening.

[0061] One method of calculating the conditional probability of an event is based on Bayes' theorem. Bayes' theorem describes the probability that a hypothesis is true given certain evidence. Assuming that event A represents a specific event and event B represents data collected by a sensor or other related conditions, the formula for Bayes' theorem is: ; Among them, P(A|B) is the conditional probability of event A occurring under condition B, P(B|A) is the probability of condition B occurring when event A occurs, P(A) is the prior probability of event A occurring, and P(B) is the probability of condition B occurring. The prior probability P(A) can be obtained based on historical data statistics, for example, by analyzing the frequency of an event occurring over a period of time to estimate its prior probability. P(B|A) can be obtained by statistical analysis of the relevant condition B when the known event A occurs. P(B) can be calculated using the total probability formula: Among them, A i Represents all possible events. In addition to Bayes' theorem, machine learning algorithms can also be used to calculate the conditional probability of events. For example, the naive Bayes classifier is a classification algorithm based on Bayes' theorem and the assumption of feature conditional independence. The naive Bayes classifier can obtain the probability distribution of each event under different feature conditions by learning the training data. When predicting, the probability of each event is calculated based on the input feature conditions.

[0062] In step S1022, the event corresponding to the maximum conditional probability among the multiple events is used as the event in the environmental sensing data set. The purpose of this step is to select the most likely event from multiple possible events to improve the accuracy of subsequent event comparison and monitoring. After calculating the conditional probability of each event, the event corresponding to the maximum conditional probability is found by comparing the sizes of these probability values.

[0063] In step S1023, the events in the environmental sensor data set are compared with the events in the event relationship network to obtain event comparison results. The event relationship network is a pre-constructed network structure, in which each component represents an event and the connecting lines represent the relationship between events.

[0064] When performing event comparison, a method based on semantic matching can be used. For events in the determined environmental sensor data set, key information can be extracted, such as the name of the event, the time of occurrence, related equipment, etc., and then this information can be compared with the event information in the event relationship network. For example, if the determined event is "tower crane start", the components related to "tower crane start" can be found in the event relationship network. The comparison process can use natural language processing technology, such as lexical analysis, syntactic analysis, etc., to convert the event description into a computer-processable form, and then perform similarity calculation. Similarity calculation can use a variety of methods, such as edit distance, cosine similarity, etc. Another comparison method is rule-based comparison. Some comparison rules can be pre-defined. For example, if the determined event involves a specific device, then find events related to the device in the event relationship network. The definition of rules can be customized according to the characteristics and needs of the construction process.

[0065] As an implementation manner, step S200, when the environmental sensing data set includes a first event, obtaining a first local relationship network in the event relationship network includes: Step S210: when the environmental sensing data set includes a first event, taking a component item corresponding to the first event in the event relationship network as a first component item; Step S220: based on the event relationship network, obtaining a set of components involved with the first component, wherein the path length between each component in the set of components and the first component is not greater than a preset path length; Step S230: generating a first local relationship network based on the first component item, the component item set, and the connection lines between the component items.

[0066] In one implementation of step S200, steps S210-S230 are an important process for obtaining a first local relationship network from an event relationship network when the environmental sensing data set includes a first event. The local relationship network can provide rich contextual information for subsequent intelligent monitoring of the construction environment, which helps to solve the problems of missing context and inaccurate monitoring caused by short real-time detection data streams.

[0067] In step S210, when the environmental sensing data set includes a first event, a component item corresponding to the first event in the event relationship network is used as a first component item.

[0068] In step S220, based on the event relationship network, a set of components that are involved (i.e., associated) with the first component is obtained, and the path length between each component in the set of components and the first component is not greater than the preset path length. The preset path length is a preset parameter that limits the scope of the local relationship network and reflects the closeness of the association between events. The path length refers to the number of connecting lines from one component to another. A graph traversal algorithm can be used to obtain a set of components related to the first component. The graph traversal algorithms include breadth-first search (BFS) and depth-first search (DFS), which have been explained in the above content and will not be repeated here.

[0069] In step S230, a first local relationship network is generated based on the first component, the component set and the connection lines between the components. The first local relationship network is a subgraph of the event relationship network, which contains important information related to the first event and can reflect the association and context relationship between events.

[0070] When generating the first local relationship network, the first component item, the component item set and the connection lines between them are extracted from the event relationship network. The connection lines represent the relationship between events, such as causal relationship, sequential relationship, collaborative relationship, etc. A new graph structure, namely the first local relationship network, can be constructed based on the identification of the component items and the information of the connection lines.

[0071] In practical applications, the implementation of steps S210-S230 may be affected by factors such as the complexity of the event relationship network and the accuracy of the data. In order to improve the accuracy and efficiency of obtaining the first local relationship network, the event relationship network can be optimized and sorted to remove redundant information and improve the connectivity of the graph. At the same time, for the accuracy of the data, a data verification and error correction mechanism can be used to ensure that the information in the event relationship network is accurate.

[0072] By determining the first component in the event relationship network, obtaining the set of related components and generating the first local relationship network, important context information is provided for the subsequent intelligent monitoring of the construction environment. In the complex environment of the construction site, the first local relationship network can help better understand the relationship between events, accurately judge the construction status, and promptly discover potential problems and risks, providing strong support for construction management and decision-making.

[0073] As a second implementation, step S200, when the environmental sensing data set includes a first event, obtaining a first local relationship network in the event relationship network includes: Step S201: when the environmental sensing data set includes a first event, taking a component item corresponding to the first event in the event relationship network as a first component item; Step S202: based on the event relationship network, obtaining a set of components that are involved with the first component, wherein the path length between each component in the set of components and the first component is not greater than a preset path length; Step S203: when the environmental sensing data set includes the first event relationship, based on the event relationship network, obtaining a first partial component item set having the first event relationship with the first component item from the component item set; Step S204: generating a first local relationship network based on the first component item, the first local component item set, and the connection lines between the component items.

[0074] Step S201 and step S202 may refer to the aforementioned steps S210 to S220.

[0075] In step S203, when the environmental sensing data set includes a first event relationship, based on the event relationship network, a first partial component item set having a first event relationship with the first component item is obtained in the component item set. The first event relationship refers to a certain specific association between events, such as a causal relationship, a sequential relationship, a collaborative relationship, etc. When executing this step, the relationship between each component item in the component item set and the first component item is checked to determine whether there is a first event relationship. For example, if the first event relationship is a causal relationship, for the first component item "concrete pouring", a component item having a causal relationship with "concrete pouring" will be searched in the component item set. For example, "concrete mixing" is a preceding causal event of "concrete pouring", then "concrete mixing" will be included in the first partial component item set. The relationship between events can be determined by the marking information of the connecting line in the event relationship network. Each connecting line can have a label indicating the type of event relationship it represents, and the components that meet the first event relationship are screened out by reading these label information.

[0076] In step S204, a first local relationship network is generated based on the first component, the first local component set and the connecting lines between the components. The first local relationship network is a subgraph of the event relationship network, which contains information closely related to the first event and has a specific event relationship, and can more accurately reflect the association and contextual relationship between events. When generating the first local relationship network, it is necessary to extract the first component, the first local component set and the connecting lines between them from the event relationship network. The connecting lines represent various relationships between events. According to the identification of the components and the information of the connecting lines, a new graph structure, i.e., the first local relationship network, is constructed. For example, for the first component "concrete pouring", the first local component set may include "concrete mixing", "formwork installation", etc. These components and the connecting lines between them that represent the relationship of cause and effect, sequence, etc. are extracted to generate a first local relationship network with "concrete pouring" as the core. This local relationship network can intuitively display a series of events and their mutual relationships related to "concrete pouring" and having a specific relationship, which helps to better understand the situation on the construction site.

[0077] Steps S201-S204 provide more targeted and relevant context information for subsequent intelligent monitoring of the construction environment by determining the first component in the event relationship network, obtaining a set of related components, screening out a first set of local components with specific event relationships, and generating a first local relationship network. In the complex environment of the construction site, the first local relationship network can help to better understand the relationship between events, accurately judge the construction status, and promptly discover potential problems and risks, providing strong support for construction management and decision-making. In addition, based on the above method, the coverage of the local relationship network is controlled according to the preset path length. Not only can the problem of the explosion of the local relationship network data caused by the long path length be avoided, but also the information collection effect can be better, and the events in the environmental sensor data set can be better characterized.

[0078] As a third implementation, step S200, when the environmental sensing data set includes a first event, obtaining a first local relationship network in the event relationship network includes: Step S2001: when the environmental sensing data set includes a first event, taking a component item corresponding to the first event in the event relationship network as a first component item; Step S2002: based on the event relationship network, obtaining a set of components related to the first component, wherein the path length between each component in the set of components and the first component is not greater than a preset path length; Step S2003: when the environmental sensing data set includes the first event relationship and the second event relationship, based on the event relationship network, obtaining a second partial component item set in the component item set that has one or all of the first event relationship or the second event relationship with the first component item; Step S2004: Generate a first local relationship network based on the first component item, the second local component item set, and the connection lines between the component items.

[0079] Step S2001 and step S2002 may refer to the aforementioned steps S210 to S220.

[0080] In step S2003, when the environmental sensing data set includes the first event relationship and the second event relationship, based on the event relationship network, a second partial component set that has one or all of the first event relationship or the second event relationship with the first component is obtained in the component set. The first event relationship and the second event relationship refer to specific associations between events, such as causal relationships, sequential relationships, collaborative relationships, and dependency relationships. When executing this step, it is necessary to conduct a detailed check on the relationship between each component in the component set and the first component to determine whether there is a first event relationship or a second event relationship. For example, if the first event relationship is a causal relationship and the second event relationship is a collaborative relationship, for the first component "tower crane lifting large components", the component set will be searched for components that have a causal relationship with "tower crane lifting large components", such as "components are ready" is the pre-causal event of "tower crane lifting large components"; at the same time, components that have a collaborative relationship are searched, such as "ground command personnel cooperation" and "tower crane lifting large components" have a collaborative relationship, and these components will be included in the second partial component set. The relationship between events can be determined by the label information of the connection line in the event relationship network. Each connection line can have a label indicating the type of event relationship it represents, and the components that meet the first or second event relationship can be screened out by reading the label information.

[0081] In step S2004, a first local relationship network is generated based on the first component, the second local component set and the connecting lines between the components. The first local relationship network is a subgraph of the event relationship network, which contains information closely related to the first event and has a specific event relationship, and can more accurately reflect the association and contextual relationship between events. When generating the first local relationship network, it is necessary to extract the first component, the second local component set and the connecting lines between them from the event relationship network. The connecting lines represent various relationships between events. According to the identification of the components and the information of the connecting lines, a new graph structure, i.e., the first local relationship network, is constructed. For example, for the first component "tower crane lifting large components", the second local component set may include "component fixing", "tower crane operator operation", "component ready", "ground command personnel cooperation", etc. These components and the connecting lines between them representing causal, collaborative and other relationships will be extracted to generate a first local relationship network with "tower crane lifting large components" as the core. This local relationship network can intuitively display a series of events related to "tower crane lifting large components" and their interrelationships, helping to better understand the situation at the construction site.

[0082] Steps S2001-S2004 provide more targeted and comprehensive context information for subsequent intelligent monitoring of the construction environment by determining the first component in the event relationship network, obtaining a set of related components, screening out a second set of local components with specific event relationships, and generating a first local relationship network. In the complex environment of the construction site, the first local relationship network can help better understand the relationship between events, accurately judge the construction status, and promptly discover potential problems and risks, providing strong support for construction management and decision-making. In addition, based on the above method, the coverage of the local relationship network is controlled according to the preset path length. Not only can the problem of the explosion of the local relationship network data caused by the long path length be avoided, but also the information collection effect can be better, and the events in the environmental sensor data set can be better characterized.

[0083] As an implementation manner, step S500, based on the first relational embedded array and the sensor embedded array, generates an integrated embedded array, including: Step S510: performing a product of the sensor embedding array and the first weight array to obtain a first product result, where the dimension of the sensor embedding array is i; Step S520: flipping the result of multiplying the first relation embedding array and the second weight array to obtain a flipped result, where the dimension of the first relation embedding array is s; Step S530: performing a product of the first product result and the flip result to obtain a second product result; Step S540: Add the second product result to the bias array to obtain an integrated embedded array; wherein the first weight array and the second weight array are obtained by extracting a third-order tensor, the dimension of the third-order tensor is i×s×t, the dimension of the first weight array is i×1×t, the dimension of the second weight array is 1×s×t, i is the dimension of the sensor embedded array, s is the dimension of the first relation embedded array, and t is the dimension of the integrated embedded array.

[0084] In step S510, the sensor embedding array is multiplied with the first weight array to obtain a first product result, and the dimension of the sensor embedding array is i. The sensor embedding array is a numerical representation of the feature information extracted from the environmental sensor data set. It contains various environmental status data of the construction site, such as temperature, humidity, noise, vibration and other data collected by sensors. The first weight array is a two-dimensional array (matrix), which is used to weight each element in the sensor embedding array to highlight the importance of different environmental characteristics. When executing this step, matrix multiplication operations are performed according to the dimensions of the sensor embedding array and the first weight array. Assuming that the sensor embedding array is a, its dimension is i×1, and the first weight array is W1, its dimension is i×t (t is the dimension of the integrated embedding array), then the calculation method of the first product result b is: , where the rule for matrix multiplication is: the j-th row element b of b j is equal to the sum of the products of the jth row of W1 and the corresponding element of a, that is, ,here express The j-th row and k-th column element of a k Represents the kth element of a.

[0085] For example, if the sensor is embedded in the array , the first weight array , then the first product result is Different weight values ​​can be set according to the importance of different environmental features to construction monitoring. For example, if the temperature has a greater impact on the construction quality at the construction site, the weight of the corresponding temperature feature in the first weight array may be set higher, thereby highlighting the role of temperature information in the integration process.

[0086] In step S520, the result of multiplying the first relation embedding array and the second weight array is flipped to obtain a flipped result, and the dimension of the first relation embedding array is s. The first relation embedding array is a numerical representation of the feature information extracted from the first local relation network of the event relation network, which reflects the relationship and context information between the events at the construction site. The second weight array is also used to weight the elements in the first relation embedding array to reflect the importance of different event relations. First, the first relation embedding array c (dimension s×1) is matrix multiplied with the second weight array W2 (dimension s×t) to obtain the intermediate result d: , which is calculated in a similar way to step S510, the j-th row element d j is equal to the sum of the products of the jth row of W2 and the corresponding elements of c. Then, d is flipped. The flipping operation can be to reverse the order of the vector elements. For example, if , the result after flipping By flipping the operation, the corresponding relationship between the first relation embedding array and the sensing embedding array when they are multiplied can be changed, thereby increasing the diversity of information fusion.

[0087] In step S530, the first product result is multiplied with the flip result to obtain a second product result. In this step, the first product result b (dimension is t×1) and the flip result e (dimension is t×1) are multiplied element by element to obtain the second product result f. The rule of element-by-element multiplication is: the jth element f of f j is equal to the product of the jth element of b and the jth element of e, that is, , for example, if ,but Through this product operation, the information of the sensor embedding array and the first relation embedding array is fused, so that the second product result includes both the environmental information of the construction site and the relation information between events.

[0088] In step S540, the second product result is added to the bias array to obtain an integrated embedded array, where the first weight array and the second weight array are obtained by extracting a third-order tensor, the dimension of the third-order tensor is i×s×t, the dimension of the first weight array is i×t, the dimension of the second weight array is s×t, i is the dimension of the sensor embedded array, s is the dimension of the first relational embedded array, and t is the dimension of the integrated embedded array. The bias array b (the meaning of b here is different from that of b in the previous step) is a vector of dimension t×1, which is used to translate the second product result, adjust the overall numerical level of the integrated embedded array, and increase the flexibility of the model. The second product result f is added element by element to the bias array b to obtain the integrated embedded array g: , that is, the jth element g of g j is equal to the sum of the jth element of f and the jth element of b, g j = f j + b j .

[0089] The values ​​of the first weight array and the second weight array can be determined through the training process. A large number of environmental sensor data set samples and corresponding prior construction monitoring result labels can be used, and an optimization algorithm (such as stochastic gradient descent method) can be used to adjust the parameters of the first weight array, the second weight array and the bias array, so that the generated integrated embedded array can better reflect the actual situation of the construction environment, thereby improving the accuracy of subsequent construction monitoring. Steps S510-S540 effectively fuse the information of the sensor embedded array and the first relationship embedded array through matrix operations, flip operations and element-by-element addition. The generated integrated embedded array integrates the environmental information and event relationship information of the construction site, providing an important data basis for the subsequent decision-making components based on the construction monitoring neural network to obtain accurate construction monitoring results.

[0090] As a second implementation, step S500, based on the first relational embedded array and the sensor embedded array, generates an integrated embedded array, including: Step S501: Combine the first relational embedded array and the sensor embedded array to obtain an integrated embedded array; wherein the dimension of the sensor embedded array is i, the dimension of the first relational embedded array is s, and the dimension of the integrated embedded array is i+s, i≥2, s≥2.

[0091] In step S501 of the second implementation of step S500, the first relationship embedding array and the sensor embedding array are combined to obtain an integrated embedded array. This step is to simply and directly fuse the event association information contained in the first relationship embedding array extracted from the event relationship network with the construction site environment information contained in the sensor embedding array extracted from the environmental sensor data set, thereby providing more comprehensive data support for subsequent construction monitoring.

[0092] The first relation embedding array is obtained by extracting features from the first local relation network in the event relation network, which reflects the relationship and contextual information between construction site events.

[0093] When executing step S501, the combination method adopted is to splice the first relation embedded array and the sensor embedded array in a certain order to form a new array, that is, an integrated embedded array. Assuming that the dimension of the sensor embedded array is i and the dimension of the first relation embedded array is s, the dimension of the integrated embedded array is i + s. The elements of the sensor embedded array can be arranged in the front, and the elements of the first relation embedded array can be arranged in the back, forming a new longer vector. Of course, the arrangement order can also be adjusted according to specific needs, such as arranging the elements of the first relation embedded array in the front and the elements of the sensor embedded array in the back.

[0094] As a third implementation, step S500, based on the first relational embedded array and the sensor embedded array, generates an integrated embedded array, including: Step S5001: Obtain the Hadamard product of the first relational embedding array and the sensor embedded array to obtain an integrated embedded array; the dimensions of the first relational embedding array, the sensor embedded array and the integrated embedded array are equal.

[0095] The Hadamard product, also known as the element-by-element product, is an operation of multiplying corresponding elements, requiring the dimensions of the two arrays involved in the operation to be equal. Through the Hadamard product operation, the corresponding elements in the first relational embedded array and the sensor embedded array are multiplied point by point to obtain a new array. This operation can highlight the interaction between the corresponding elements in the two arrays. For example, in the construction monitoring scenario, an element in the first relational embedded array may represent the strength of the association between "tower crane operation" and "material lifting", while the corresponding element in the sensor embedded array may represent the current wind speed. Through the Hadamard product, the two elements are multiplied, and the result obtained can reflect the actual impact of the association between "tower crane operation" and "material lifting" under the current wind speed conditions. If the wind speed is high and the strength of the association between "tower crane operation" and "material lifting" is high, then the result after multiplication may be large, which may mean that in the current environment, the collaborative operation of these two events needs to be more cautious and there may be certain safety risks.

[0096] In one implementation, the method provided in the embodiment of the present application may further include: Step S200a: when the environmental sensing data set also includes a second event, obtaining a second local relationship network in the event relationship network, wherein the second event belongs to the event represented by the second component item in the event relationship network, and the second local relationship network includes the second component item, components within a preset path length range obtained based on the second component item, and connecting lines between the components; Step S200b: According to the second local relationship network, a second relationship embedding array is obtained based on the first feature extraction component in the construction monitoring neural network.

[0097] At this time, as a fourth implementation, step S500, based on the first relational embedded array and the sensor embedded array, generates an integrated embedded array, including: Step S500a: Generate an integrated embedded array based on the first relation embedded array, the second relation embedded array and the sensor embedded array.

[0098] In step S200a, when the environmental sensing data set also includes a second event, a second local relationship network is obtained in the event relationship network, wherein the second event belongs to the event represented by the second component in the event relationship network, and the second local relationship network includes the second component, components within a preset path length range obtained based on the second component, and connecting lines between the components. The second event is another representative event detected from the environmental sensing data set other than the first event. When executing this step, the second component corresponding to the second event is found in the event relationship network. For example, if the first event is "tower crane operation" and the second event is "concrete pouring", the component representing "concrete pouring" will be found in the event relationship network as the second component. Then, with the second component as the center, according to the preset path length, other components related to the second component and the connecting lines between them are obtained through a graph traversal algorithm (such as breadth-first search or depth-first search), thereby generating a second local relationship network. The preset path length limits the scope of the local relationship network and reflects the closeness of the association between events. Taking the breadth-first search algorithm as an example, starting from the second component, mark it as visited and add it to the queue. Then, take a node from the queue and check all its adjacent nodes. If the adjacent node has not been visited and the path length with the second component is not greater than the preset path length, mark it as visited and add it to the queue, and record the connection line information between the nodes. Repeat this process until the queue is empty, and finally get the second local relationship network.

[0099] In step S200b, according to the second local relationship network, the second relationship embedding array is obtained based on the first feature extraction component in the construction monitoring neural network. The first feature extraction component is a part of the construction monitoring neural network, and its function is to extract representative feature information from the second local relationship network and convert it into a numerical vector, i.e., the second relationship embedding array. Graph embedding technology can be used to achieve this goal. Graph embedding algorithms include DeepWalk, Node2Vec, etc. For details, please refer to the above-mentioned related content introduction.

[0100] Based on steps S200a-S200b, in step S500 of the fourth implementation mode, in step S500a, based on the first relation embedding array, the second relation embedding array and the sensor embedding array, an integrated embedding array is generated. The purpose of this step is to fuse the relation information of multiple events with the environmental sensor information to provide more comprehensive information for construction monitoring.

[0101] When executing step S500a, one method is to first process the first relation embedding array and the second relation embedding array, and then integrate them with the sensor embedding array. For example, the Hadamard product of the first relation embedding array and the preset parameter array (such as the weight vector) can be obtained first to obtain the first intermediate result; the Hadamard product of the second relation embedding array and the preset parameter array can be obtained to obtain the second intermediate result. The preset parameter array is used to weight the elements in the first relation embedding array and the second relation embedding array to highlight the importance of different event relationships. Next, the sensor embedding array is multiplied with the first weight array to obtain a first multiplication result. The first weight array is a two-dimensional array (matrix), and its function is to weight each element in the sensor embedding array to highlight the importance of different environmental features. Then, the target relation embedding array is multiplied with the second weight array and then flipped to obtain a flipped result. The second weight array is also used to weight the elements in the target relation embedding array to reflect the importance of different event relationships. Finally, the first multiplication result is multiplied with the flipping result to obtain a second multiplication result, and then the second multiplication result is added with the offset array to obtain an integrated embedding array. Another method for implementing step S500a is to combine the target relationship embedded array and the sensor embedded array to obtain an integrated embedded array. For example, the target relationship embedded array and the sensor embedded array are connected in a certain order to form a new array. Assuming the dimension of the sensor embedded array is i, and the dimension of the target relationship embedded array is s, the dimension of the integrated embedded array is i + s. Another method is to fuse the target relationship embedded array with the sensor embedded array to obtain an integrated embedded array. The fusion method can be to take the average value, weighted average value, etc. of the two, requiring the dimensions of the target relationship embedded array, the sensor embedded array and the integrated embedded array to be consistent.

[0102] As an implementation manner, after obtaining the environmental sensor data set to be analyzed, the method provided in the embodiment of the present application may further include the above-mentioned determination process of the second event: Step S1001: performing event detection on the environmental sensor data set to obtain event detection results; Step S1002: when the event detection result indicates that the environmental sensor data set includes two events, the two events in the environmental sensor data set are respectively compared with the events in the event relationship network to obtain an event comparison result for each event; Step S1003: when an event comparison result indicates that an event in the environmental sensor data set corresponds to an event in the event relationship network, the corresponding event is used as the first event; Step S1004: when the comparison result of the remaining event indicates that the remaining event in the environmental sensor data set corresponds to an event in the event relationship network, the corresponding remaining event is used as the second event.

[0103] In step S1001, event detection is performed on the environmental sensor data set to obtain event detection results. The environmental sensor data set is composed of multi-source data collected by various sensors at the construction site. These data contain various physical information of the construction site, such as temperature, humidity, noise, vibration, etc. The purpose of event detection is to identify events with specific meanings from these complex data, such as the start-up of construction equipment, the transportation of materials, the activities of personnel, etc.

[0104] There are many technical means that can be used for event detection. One method is a threshold-based detection method. For certain types of sensor data, such as temperature, noise, etc., a reasonable threshold range can be set. When the data collected by the sensor exceeds this threshold range, it is considered that an event may have occurred. For example, at a construction site, the normal ambient temperature range may be between 20℃-30℃. If the temperature detected by the temperature sensor suddenly rises to 40℃, it can be judged that events such as equipment overheating or fire may have occurred. The setting of the threshold can be obtained based on statistical analysis of historical data, or it can be manually set according to construction technology and safety standards.

[0105] Another method is an event detection method based on pattern recognition. The sensor data patterns corresponding to various events can be learned in advance, and then the environmental sensor data collected in real time can be matched with these patterns. For example, when the construction equipment is started, the vibration sensor will detect a vibration signal of a specific frequency and amplitude. By analyzing the characteristics of the vibration signal, such as frequency distribution, amplitude change, etc., it can be determined whether the equipment is started. Pattern recognition can use machine learning algorithms such as support vector machines (SVM), decision trees, etc., or deep learning algorithms such as convolutional neural networks (CNN), recurrent neural networks (RNN), etc.

[0106] In addition to the above methods, event detection methods based on time series analysis can also be used. Since environmental sensor data usually changes over time, time series analysis can capture the dynamic change characteristics of the data. For example, potential events can be discovered by analyzing the trend, periodicity, and seasonality of the data. Time series analysis methods include autoregressive integrated moving average model (ARIMA), seasonal decomposition, etc.

[0107] In step S1002, when the event detection result shows that the environmental sensing data set includes two events, the two events in the environmental sensing data set are compared with the events in the event relationship network respectively to obtain the event comparison result of each event. The event relationship network is a pre-constructed network structure, in which each component represents an event and the connecting line represents the relationship between the events. Similarity calculation can use a variety of methods, such as edit distance, cosine similarity, etc. Edit distance refers to the minimum number of operations required to convert two strings into another string through insertion, deletion and substitution operations. Cosine similarity measures the similarity of two vectors by calculating the cosine value of the angle between them. Another comparison method is rule-based comparison. Some comparison rules can be pre-defined. For example, if the detected event involves a specific device, then find events related to the device in the event relationship network.

[0108] In step S1003, when an event comparison result indicates that an event in the environmental sensor data set corresponds to an event in the event relationship network, the corresponding event is taken as the first event.

[0109] In step S1004, when the comparison result of the remaining event indicates that the remaining event in the environmental sensor data set corresponds to the event in the event relationship network, the corresponding remaining event is used as the second event. The determination of the second event further enriches the event information in the environmental sensor data set and provides a basis for a more comprehensive analysis of the construction environment in the future.

[0110] For example, if "crane start" is determined as the first event, and the "concrete mixing" event also corresponds to a component item in the event relationship network, then "concrete mixing" is taken as the second event. Subsequently, based on the second event, a second partial relationship network can be obtained in the event relationship network, and the relevant information of the first event and the second event can be integrated to more accurately understand the situation at the construction site.

[0111] Steps S1001-S1004 provide an important basis for subsequent intelligent monitoring of the construction environment by detecting events in the environmental sensor data set, comparing it with the event relationship network, and determining the first event and the second event. The first event and the second event can be used to obtain relevant context information in the event relationship network, so as to better understand the situation at the construction site and improve the accuracy and reliability of monitoring. In the complex environment of the construction site, accurate event detection and the determination of the first and second events help to timely discover potential problems and risks, and provide strong support for construction management and decision-making. For example, by determining the two events of "tower crane start" and "concrete mixing", the relationship between them can be analyzed, such as whether the tower crane is used to lift concrete-related materials, so as to make coordination and safety precautions in advance to ensure the smooth progress of construction.

[0112] As an implementation manner, step S500a, based on the first relation embedding array, the second relation embedding array and the sensor embedding array, generates an integrated embedding array, including: Step S500a1: obtaining the Hadamard product of the first relation embedding array and the preset parameter array to obtain a first intermediate result; Step S500a2: Obtain the Hadamard product of embedding the second relation into the array preset parameter array to obtain a second intermediate result; Step S500a3: Merge the first intermediate result and the second intermediate result to obtain a target relation embedding array; Step S500a4: performing a product of the sensor embedding array and the first weight array to obtain a first product result, wherein the dimension of the sensor embedding array is i; Step S500a5: performing a product of the target relationship embedding array and the second weight array and then performing a flipping operation to obtain a flipping result, wherein the dimension of the target relationship embedding array is s; Step S500a6: Perform a product of the first product result and the flip result to obtain a second product result; Step S500a7: Add the second product result to the bias array to obtain an integrated embedded array; wherein the first weight array and the second weight array are obtained by extracting a third-order tensor, the dimension of the third-order tensor is i×s×t, the dimension of the first weight array is i×1×t, the dimension of the second weight array is 1×s×t, and the dimension of the integrated embedded array is t.

[0113] In step S500a1, the Hadamard product of the first relationship embedding array and the preset parameter array is obtained to obtain the first intermediate result. The preset parameter array can be a preset weight vector, which is used to weight each element in the first relationship embedding array to highlight the importance of different event relationships. Let the first relationship embedding array be a, and the preset parameter array be w, and their dimensions are the same, both s. The Hadamard product is an operation of multiplying corresponding elements, so the calculation formula of the first intermediate result c is: ,in represents the Hadamard product. For example, if the first relation embeds the array , preset parameter array , then the first intermediate result Through the Hadamard product operation, the elements in the first relation embedding array can be adjusted according to the preset parameter array, so that the importance of different event relations can be reflected in subsequent calculations.

[0114] In step S500a2, the Hadamard product of the second relation embedding array and the preset parameter array is obtained to obtain a second intermediate result. The second relation embedding array is a numerical representation of feature information extracted from the second local relation network of the event relation network, which also reflects the relationship and context information between related events at the construction site. The process of obtaining the second intermediate result can refer to the example of step S500a1, which will not be described in detail here.

[0115] In step S500a3, the first intermediate result and the second intermediate result are fused to obtain the target relationship embedding array. There are many ways to fuse, such as dot product calculation, sum calculation, average, etc., which are not specifically limited.

[0116] In step S500a4, the sensor embedding array is multiplied with the first weight array to obtain a first product result, wherein the dimension of the sensor embedding array is i. The sensor embedding array is a numerical representation of feature information extracted from the environmental sensor data set, which contains various environmental status data of the construction site, such as data collected by sensors such as temperature, humidity, noise, vibration, etc. The first weight array is a two-dimensional array (matrix), which is used to weight each element in the sensor embedding array to highlight the importance of different environmental features.

[0117] In step S500a5, the target relationship embedding array is multiplied by the second weight array and then flipped to obtain a flip result, wherein the dimension of the target relationship embedding array is s. The second weight array is also used to weight the elements in the target relationship embedding array to reflect the importance of different event relationships. Through this step, the target relationship embedding array is weighted and ordered, increasing the diversity of information fusion.

[0118] In step S500a6, the first product result is multiplied with the flip result to obtain a second product result. In this step, the first product result g (dimension is t×1) and the flip result i (dimension is t×1) are multiplied element by element to obtain the second product result j. Through this product operation, the information of the sensor embedding array and the target relationship embedding array is integrated, so that the second product result contains both the environmental information of the construction site and the relationship information between events.

[0119] In step S500a7, the second product result is added to the bias array to obtain an integrated embedded array. The bias array b is a vector of dimension t×1, and its function is to translate the second product result, adjust the overall numerical level of the integrated embedded array, and increase the flexibility of the model. In practical applications, the values ​​of the first weight array, the second weight array, and the preset parameter array can be determined through a training process. A large number of environmental sensing data set samples and corresponding prior construction monitoring result labels can be used, and an optimization algorithm (such as stochastic gradient descent method) can be used to adjust these parameters, so that the generated integrated embedded array can better reflect the actual situation of the construction environment, thereby improving the accuracy of subsequent construction monitoring.

[0120] Steps S500a1-S500a7 effectively integrate the information of the first relational embedded array, the second relational embedded array and the sensor embedded array through a series of operations. The generated integrated embedded array combines the environmental information of the construction site and the relationship information of multiple events, providing an important data basis for the subsequent decision-making components based on the construction monitoring neural network to obtain accurate construction monitoring results.

[0121] As a second implementation, step S500a, based on the first relation embedding array, the second relation embedding array and the sensor embedding array, generates an integrated embedding array, including: Step S500a01: Obtain the Hadamard product of the first relation embedding array and the preset parameter array to obtain a first intermediate result; Step S500a02: Obtain the Hadamard product of the second relation embedding array and the preset parameter array to obtain a second intermediate result; Step S500a03: merge the first intermediate result and the second intermediate result to obtain a target relation embedding array; Step S500a04: Combine the target relationship embedding array and the sensor embedding array to obtain an integrated embedding array; wherein the dimension of the sensor embedding array is i, the dimension of the target relationship embedding array is s, and the dimension of the integrated embedding array is i+s.

[0122] In step S500a01, the Hadamard product of the first relation embedding array and the preset parameter array is obtained to obtain a first intermediate result. In step S500a02, the Hadamard product of the second relation embedding array and the preset parameter array is obtained to obtain a second intermediate result.

[0123] In step S500a03, the first intermediate result and the second intermediate result are fused to obtain the target relation embedding array. The fusion operation can be performed in a variety of ways, such as dot product calculation, sum calculation, average, etc., which are not specifically limited.

[0124] In step S500a04, the target relationship embedding array and the sensor embedding array are combined to obtain an integrated embedded array. The combination operation is to splice the target relationship embedding array and the sensor embedding array in order. When it is detected that the data set contains the first event and the second event, the relationship information and the sensor information related to the two events can be integrated into an array through the above steps. The first relationship embedding array and the second relationship embedding array respectively represent the local relationship information in the event relationship network related to the two events, and the sensor embedding array represents the actual data collected by each sensor. By processing and combining these arrays, the integrated embedded array obtained can more comprehensively reflect the situation at the construction site. The decision component of the construction monitoring neural network can make more accurate construction monitoring result judgments based on this integrated embedded array, such as judging whether there are construction safety hazards, whether the construction progress is affected, etc.

[0125] As a third implementation, step S500a, based on the first relation embedding array, the second relation embedding array and the sensor embedding array, generates an integrated embedding array, including: Step S500a001: Obtain the Hadamard product of the first relation embedding array and the preset parameter array to obtain a first intermediate result; Step S500a002: Obtain the Hadamard product of the second relation embedding array and the preset parameter array to obtain a second intermediate result; Step S500a003: merge the first intermediate result and the second intermediate result to obtain a target relation embedding array; Step S500a004: Fusing the target relationship embedding array and the sensor embedding array to obtain an integrated embedded array; wherein the dimensions of the target relationship embedding array, the sensor embedding array and the integrated embedded array are consistent.

[0126] In step S500a001, the Hadamard product of the first relation embedding array and the preset parameter array is obtained to obtain a first intermediate result. In step S500a002, the Hadamard product of the second relation embedding array and the preset parameter array is obtained to obtain a second intermediate result. In step S500a003, the first intermediate result and the second intermediate result are fused to obtain a target relation embedding array. The fusion operation can be performed in a variety of ways, such as summation, averaging, etc. In step S500a004, the target relation embedding array and the sensor embedding array are fused to obtain an integrated embedding array, and the dimensions of the target relation embedding array, the sensor embedding array and the integrated embedding array are consistent. The sensor embedding array is obtained based on the environmental sensor data set through the second feature extraction component in the construction monitoring neural network, and it is a numerical representation of the environmental sensor data. There are also many fusion methods, such as weighted summation, averaging, etc.

[0127] Through the above steps, the relationship information and sensor information related to the two events are fused. The first relationship embedding array and the second relationship embedding array respectively reflect the local relationship information in the event relationship network related to the two events, and the sensor embedding array is the actual data collected by each sensor. By processing and fusing these arrays, the integrated embedded array obtained can more comprehensively reflect the actual situation of bridge construction. It is helpful to effectively integrate and utilize multi-source data. The information in the event relationship network and the environmental sensor data themselves have different characteristics and sources. Through the Hadamard product and fusion operation, they can be organically combined. The role of the preset parameter array is to adjust the importance of each relationship information according to different application scenarios and needs. At different construction stages, the importance of different events may change. By adjusting the preset parameter array, the integrated embedded array can more accurately reflect the current construction status.

[0128] As an implementation method, the training process of the construction monitoring neural network provided in this application is as follows: Step S10: Acquire an environmental sensor data set sample library, wherein each environmental sensor training data set in the environmental sensor data set sample library includes a priori event marker, the priori event marker in each environmental sensor training data set belongs to the event represented by the target component item in the event relationship network, and each environmental sensor training data set carries a corresponding priori construction monitoring result marker; Step S20: for each environmental sensing training data set, extracting a training local relationship network corresponding to the environmental sensing training data set in the event relationship network, wherein the training local relationship network includes a target component item, components within a preset path length range obtained based on the target component item, and connecting lines between the components; Step S30: for each environmental sensing training data set, according to the training local relationship network, a training relationship embedding array is obtained based on the first feature extraction component in the construction monitoring neural network; Step S40: for each environmental sensor training data set, obtaining a training sensor embedding array based on the second feature extraction component in the construction monitoring neural network according to the environmental sensor training data set; Step S50: for each environmental sensor training data set, generating a training integrated embedding array based on the training relation embedding array and the training sensor embedding array; Step S60: for each environmental sensor training data set, according to the training integrated embedding array, based on the decision component in the construction monitoring neural network, obtain the construction monitoring prediction result corresponding to the environmental sensor training data set; Step S70: training the construction monitoring neural network according to the construction monitoring prediction results corresponding to each environmental sensing training data set and the errors marked by the priori construction monitoring results.

[0129] Steps S10-S70 in the training process of the construction monitoring neural network are a series of operations performed to enable the construction monitoring neural network to accurately monitor the construction site. The core purpose is to enable the neural network to learn the intrinsic relationship between environmental sensor data and construction monitoring results through a large amount of training data and reasonable training methods, thereby improving the accuracy of construction site monitoring and overcoming the problems of missing context and inaccurate monitoring caused by short real-time detection data streams.

[0130] In step S10, an environmental sensor data set sample library is obtained, wherein each environmental sensor training data set in the environmental sensor data set sample library includes a priori event mark, and the priori event mark in each environmental sensor training data set belongs to the event represented by the target component item in the event relationship network, and each environmental sensor training data set also carries a corresponding priori construction monitoring result mark. The environmental sensor data set sample library is a collection containing a large amount of training data, which is collected from the actual construction site and contains information collected by various environmental sensors, such as temperature, humidity, pressure, vibration, etc. The priori event mark is a pre-marking of the events occurring in the data set, which corresponds to the target component item in the event relationship network. The event relationship network is a network structure that describes the relationship between events, in which the component items represent events and the connecting lines represent event relationships. The priori construction monitoring result mark is a mark of the actual construction monitoring result corresponding to each environmental sensor training data set, such as whether the construction is safe, whether there is a fault, etc.

[0131] In step S20, for each environmental sensing training data set, a training local relationship network corresponding to the environmental sensing training data set is extracted from the event relationship network, wherein the training local relationship network includes a target component, components within a preset path length range obtained based on the target component, and connecting lines between the components. The preset path length refers to the limit on the number of connecting lines passing from one component to another. By extracting the training local relationship network, it is possible to focus on the local event relationship information related to the prior event marker in the current environmental sensing training data set, thereby providing more targeted information for subsequent feature extraction. For example, if the target component corresponding to the prior event marker is "construction equipment failure" and the preset path length is 2, then the training local relationship network will include the component "construction equipment failure" and other components connected to it by no more than 2 connecting lines, such as "untimely equipment maintenance" and "aging of equipment parts" and other components and connecting lines between them. A graph traversal algorithm, such as a breadth-first search algorithm, can be used to traverse the event relationship network starting from the target component, according to the limit of the preset path length, extract the relevant components and connecting lines, and construct a training local relationship network.

[0132] In step S30, for each environmental sensing training data set, the training relation embedding array is obtained based on the first feature extraction component in the construction monitoring neural network according to the training local relation network. The first feature extraction component is an important component of the construction monitoring neural network. Its function is to extract features and numerically represent the information in the training local relation network to obtain the training relation embedding array. The training relation embedding array is a numerical encoding of the event relationship information in the training local relation network, which can reflect the degree of association and characteristics between events. For example, the training local relation network contains events such as "construction equipment failure", "untimely equipment maintenance", and "aging of equipment parts". The first feature extraction component may convert the relationship between these events into a vector, such as [0.2, 0.3, 0.5], where each element represents the characteristic value of different event relationships. The first feature extraction component can be implemented using a graph neural network (GNN) in deep learning. The graph neural network can effectively process graph structure data and extract the features of nodes and edges in the graph.

[0133] In step S40, for each environmental sensor training data set, a training sensor embedding array is obtained based on the second feature extraction component in the construction monitoring neural network according to the environmental sensor training data set. The function of the second feature extraction component is to extract features and numerically represent the sensor information in the environmental sensor training data set to obtain a training sensor embedding array. The training sensor embedding array is a numerical encoding of environmental sensor data, which can reflect the characteristics of various information collected by environmental sensors. For example, the environmental sensor training data set contains readings of sensors such as temperature, humidity, and pressure. The second feature extraction component may convert these readings into a vector, such as [25.5, 60, 101.3], where each element represents the characteristic value of a different sensor. The second feature extraction component can be implemented using a multi-layer perceptron (MLP), which can perform nonlinear transformations on the input sensor data to extract useful features.

[0134] In step S50, for each environmental sensor training data set, a training integrated embedding array is generated based on the training relation embedding array and the training sensor embedding array. The purpose of generating the training integrated embedding array is to fuse the event relationship information and the sensor information so that the subsequent decision-making component can comprehensively consider these two aspects of information to judge the construction monitoring results. There are many methods for generating the training integrated embedding array. For example, the training sensor embedding array can be multiplied with the first weight array, the training relation embedding array can be multiplied with the second weight array and then flipped, and then the two results can be multiplied and added to the bias array to obtain the training integrated embedding array; the training relation embedding array and the training sensor embedding array can also be combined, or their Hadamard product can be obtained. Taking the combination method as an example, if the training sensor embedding array is [25.5, 60, 101.3] and the training relation embedding array is [0.2, 0.3, 0.5], the training integrated embedding array is [25.5, 60, 101.3, 0.2, 0.3, 0.5]. The appropriate method can be selected to generate the training integrated embedding array according to the specific application requirements and experimental results.

[0135] In step S60, for each environmental sensing training data set, according to the training integrated embedded array, the construction monitoring prediction result corresponding to the environmental sensing training data set is obtained based on the decision component in the construction monitoring neural network. The decision component is the last component of the construction monitoring neural network. Its function is to judge and predict based on the information in the training integrated embedded array, and output the construction monitoring prediction result. The decision component can be implemented using a fully connected layer and an activation function. For example, the output is converted into a probability distribution using a softmax activation function to obtain the probabilities of different construction monitoring results. For example, the construction monitoring prediction result may be that the probability of "construction safety" is 0.8, and the probability of "construction has hidden dangers" is 0.2. The forward propagation algorithm can be used to input the training integrated embedded array into the decision component, and after a series of calculations and transformations, the construction monitoring prediction result is obtained.

[0136] In step S70, the construction monitoring neural network is trained based on the construction monitoring prediction results corresponding to each environmental sensing training data set and the errors marked by the prior construction monitoring results. The error refers to the difference between the construction monitoring prediction results and the prior construction monitoring result marks. It is necessary to adjust the parameters of the neural network through an optimization algorithm to minimize the error. The optimization algorithm may be stochastic gradient descent (SGD), adaptive moment estimation (Adam), etc. For example, the prior construction monitoring result is marked as "construction safety", and the probability of "construction safety" in the construction monitoring prediction result is 0.6, and the probability of "construction hazards" is 0.4, then there is a certain error. The back propagation algorithm can be used to calculate the gradient of each parameter based on the error, and then the parameters are updated according to the rules of the optimization algorithm. Taking the stochastic gradient descent algorithm as an example, the parameter update formula is ,in are the parameters of the neural network, is the learning rate, is the gradient of the error function with respect to the parameters.

[0137] During the training process, batch training can be used to divide the environmental sensor training data set into multiple batches for training. One batch of data is trained each time. This can reduce memory usage and improve training efficiency. In addition, in order to prevent overfitting, regularization methods can be used. For example, L1 and L2 regularization, by adding regularization terms to the error function, limit the size of the neural network parameters and avoid over-complexity of the model. The error function of L2 regularization can be expressed as ,in is the original error function, is the regularization coefficient, are the parameters of the neural network.

[0138] The present application embodiment provides a monitoring system, such as Figure 2 As shown, the monitoring system 100 includes: a processor 101 and a memory 103. The processor 101 and the memory 103 are connected, such as through a bus 102. Optionally, the monitoring system 100 may also include a transceiver 104. It should be noted that in actual applications, the transceiver 104 is not limited to one, and the structure of the monitoring system 100 does not constitute a limitation on the embodiments of the present application.

[0139] Processor 101 may be a CPU, a general-purpose processor, a GPU, a DSP, an ASIC, an FPGA or other programmable logic device, a transistor logic device, a hardware component or any combination thereof. It may implement or execute various exemplary logic blocks, modules and circuits described in conjunction with the disclosure of this application. Processor 101 may also be a combination that implements computing functions, such as a combination of one or more microprocessors, a combination of a DSP and a microprocessor, etc.

[0140] The bus 102 may include a path to transmit information between the above components. The bus 102 may be a PCI bus or an EISA bus, etc. The bus 102 may be divided into an address bus, a data bus, a control bus, etc. For ease of representation, Figure 2 Only one thick line is used in the diagram, but this does not mean that there is only one bus or only one type of bus.

[0141] The memory 103 may be a ROM or other type of static storage device that can store static information and instructions, a RAM or other type of dynamic storage device that can store information and instructions, or an EEPROM, a CD-ROM or other optical disk storage, an optical disk storage (including a compressed optical disk, a laser disk, an optical disk, a digital versatile disk, a Blu-ray disk, etc.), a magnetic disk storage medium or other magnetic storage device, or any other medium that can be used to carry or store the desired program code in the form of instructions or data structures and can be accessed by a computer, but is not limited thereto.

[0142] The memory 103 is used to store application code for executing the solution of the present application, and the execution is controlled by the processor 101. The processor 101 is used to execute the application code stored in the memory 103 to implement the content shown in any of the above method embodiments.

[0143] An embodiment of the present application provides a monitoring system. The monitoring system in the embodiment of the present application includes: one or more processors; a memory; one or more computer programs, wherein the one or more computer programs are stored in the memory and configured to be executed by the one or more processors. When the one or more programs are executed by the processor, the construction environment intelligent monitoring method based on the Internet of Things provided in the embodiment of the present application is implemented.

[0144] An embodiment of the present application provides a computer-readable storage medium, on which a computer program is stored. When the computer program runs on a processor, the processor can execute the corresponding content in the aforementioned method embodiment.

[0145] It should be understood that, although the steps in the flowchart of the accompanying drawings are displayed in sequence as indicated by the arrows, these steps are not necessarily executed in sequence in the order indicated by the arrows. Unless otherwise specified herein, there is no strict order restriction on the execution of these steps, and they can be executed in other orders. Moreover, at least a part of the steps in the flowchart of the accompanying drawings may include multiple sub-steps or multiple stages, and these sub-steps or stages are not necessarily executed at the same time, but can be executed at different times, and their execution order is not necessarily sequential, but can be executed in turn or alternately with other steps or at least a part of the sub-steps or stages of other steps.

[0146] The above description is only a partial implementation method of the present application. It should be pointed out that for ordinary technicians in this technical field, several improvements and modifications can be made without departing from the principles of the present application. These improvements and modifications should also be regarded as the scope of protection of the present application.

Claims

1. A construction environment intelligent monitoring method based on the Internet of Things, characterized in that: include: Obtain the environmental sensing data set to be analyzed; When the environmental sensing data set includes a first event, a first local relationship network is obtained in the event relationship network, wherein each component item in the event relationship network represents an event, each connection line in the event relationship network represents an event relationship, the first event belongs to the event represented by the first component item in the event relationship network, and the first local relationship network includes the first component item, components within a preset path length range obtained based on the first component item, and connection lines between the components; According to the first local relationship network, obtaining a first relationship embedding array based on a first feature extraction component in a construction monitoring neural network; According to the environmental sensor data set, a sensor embedding array is obtained based on a second feature extraction component in the construction monitoring neural network; generating an integrated embedded array based on the first relational embedded array and the sensor embedded array; According to the integrated embedded array, a construction monitoring result corresponding to the environmental sensing data set is obtained based on a decision component in the construction monitoring neural network.

2. The method according to claim 1, characterized in that: After obtaining the environmental sensor data set to be analyzed, the method further includes: Performing event detection on the environmental sensing data set to obtain an event detection result; When the event detection result indicates that the environmental sensor data set includes an event, comparing the event in the environmental sensor data set with the event in the event relationship network to obtain an event comparison result; When the event comparison result indicates that the event in the environmental sensor data set corresponds to the event in the event relationship network, the corresponding event is used as the first event.

3. The method according to claim 2, characterized in that When the event detection result indicates that the environmental sensor data set includes an event, the event in the environmental sensor data set is compared with the event in the event relationship network to obtain an event comparison result, including: When the event detection result indicates that the environmental sensing data set includes multiple events, obtaining a conditional probability corresponding to each of the multiple events; Taking the event corresponding to the maximum conditional probability among the multiple events as the event in the environmental sensing data set; The events in the environmental sensing data set are compared with the events in the event relationship network to obtain the event comparison results.

4. The method according to claim 1, characterized in that: When the environmental sensing data set includes the first event, obtaining a first local relationship network in the event relationship network includes: taking the component item corresponding to the first event in the event relationship network as the first component item; Based on the event relationship network, a set of components involved with the first component is obtained, wherein the path length between each component in the set of components and the first component is not greater than the preset path length; generating the first local relationship network based on the first component item, the component item set, and the connection lines between the component items; or; taking the component item corresponding to the first event in the event relationship network as the first component item; Based on the event relationship network, a set of components involved with the first component is obtained, wherein the path length between each component in the set of components and the first component is not greater than the preset path length; When the environmental sensing data set includes a first event relationship, based on the event relationship network, obtaining a first local component item set having the first event relationship with the first component item from the component item set; generating the first local relationship network based on the first component item, the first local component item set, and the connection lines between the components; or; taking the component item corresponding to the first event in the event relationship network as the first component item; Based on the event relationship network, a set of components involved with the first component is obtained, wherein the path length between each component in the set of components and the first component is not greater than the preset path length; When the environmental sensing data set includes a first event relationship and a second event relationship, based on the event relationship network, obtaining, from the component item set, a second partial component item set that has one or all of the first event relationship or the second event relationship with the first component item; The first local relationship network is generated based on the first component item, the second local component item set, and the connection lines between the components.

5. The method according to claim 1, characterized in that The step of generating an integrated embedded array based on the first relational embedded array and the sensor embedded array comprises: Performing a product of the sensor embedding array and a first weight array to obtain a first product result, wherein the dimension of the sensor embedding array is i; Flipping a result of multiplying the first relation embedding array and the second weight array to obtain a flipped result, wherein the dimension of the first relation embedding array is s; Performing a product of the first product result and the flip result to obtain a second product result; Adding the second product result to the bias array to obtain the integrated embedded array; The first weight array and the second weight array are obtained by extracting a third-order tensor, the dimension of the third-order tensor is i×s×t, the dimension of the first weight array is i×1×t, the dimension of the second weight array is 1×s×t, i is the dimension of the sensor embedding array, s is the dimension of the first relation embedding array, and t is the dimension of the integrated embedding array; Or: generating an integrated embedded array based on the first relational embedded array and the sensor embedded array, comprising: The first relational embedding array and the sensor embedding array are combined to obtain the integrated embedding array; wherein the dimension of the sensor embedding array is i, the dimension of the first relational embedding array is s, and the dimension of the integrated embedding array is i+s, i≥2, s≥2; Or: generating an integrated embedded array based on the first relational embedded array and the sensor embedded array, comprising: The Hadamard product of the first relational embedding array and the sensor embedding array is obtained to obtain the integrated embedding array; the dimensions of the first relational embedding array, the sensor embedding array and the integrated embedding array are equal.

6. The method according to claim 1, characterized in that The method further comprises: When the environmental sensing data set further includes a second event, obtaining a second local relationship network in the event relationship network, wherein the second event belongs to an event represented by a second component item in the event relationship network, and the second local relationship network includes the second component item, components within the preset path length range obtained based on the second component item, and connecting lines between the components; According to the second local relationship network, obtaining a second relationship embedding array based on the first feature extraction component in the construction monitoring neural network; Generating an integrated embedded array based on the first relational embedded array and the sensor embedded array includes: The integrated embedded array is generated based on the first relational embedded array, the second relational embedded array and the sensor embedded array.

7. The method according to claim 6, characterized in that After obtaining the environmental sensor data set to be analyzed, the method further includes: Performing event detection on the environmental sensing data set to obtain an event detection result; When the event detection result indicates that the environmental sensor data set includes two events, the two events in the environmental sensor data set are respectively compared with the events in the event relationship network to obtain an event comparison result for each event; When an event comparison result shows that an event in the environmental sensor data set corresponds to an event in the event relationship network, the corresponding event is used as the first event; When the comparison result of the remaining event indicates that the remaining event in the environmental sensing data set corresponds to an event in the event relationship network, the corresponding remaining event is used as the second event.

8. The method according to claim 6, characterized in that The step of generating the integrated embedded array based on the first relational embedded array, the second relational embedded array and the sensor embedded array comprises: Obtaining the Hadamard product of the first relation embedding array and a preset parameter array to obtain a first intermediate result; Obtaining a Hadamard product of embedding the second relation into an array of preset parameter arrays to obtain a second intermediate result; Merging the first intermediate result and the second intermediate result to obtain a target relation embedding array; Performing a product of the sensor embedding array and a first weight array to obtain a first product result, wherein the dimension of the sensor embedding array is i; The target relationship embedding array is multiplied by the second weight array and then flipped to obtain a flip result, wherein the dimension of the target relationship embedding array is s; Performing a product of the first product result and the flip result to obtain a second product result; Adding the second product result to the bias array to obtain the integrated embedded array; The first weight array and the second weight array are obtained by extracting a third-order tensor, the dimension of the third-order tensor is i×s×t, the dimension of the first weight array is i×1×t, the dimension of the second weight array is 1×s×t, and the dimension of the integrated embedding array is t; Or: generating the integrated embedded array based on the first relational embedded array, the second relational embedded array and the sensor embedded array comprises: Obtaining the Hadamard product of the first relation embedding array and a preset parameter array to obtain a first intermediate result; Obtaining a Hadamard product of the second relation embedding array and the preset parameter array to obtain a second intermediate result; Merging the first intermediate result with the second intermediate result to obtain a target relation embedding array; Combining the target relationship embedding array and the sensor embedding array to obtain the integrated embedding array; The dimension of the sensor embedding array is i, the dimension of the target relationship embedding array is s, and the dimension of the integrated embedding array is i+s. Or: generating the integrated embedded array based on the first relational embedded array, the second relational embedded array and the sensor embedded array comprises: Obtaining the Hadamard product of the first relation embedding array and a preset parameter array to obtain a first intermediate result; Obtaining a Hadamard product of the second relation embedding array and the preset parameter array to obtain a second intermediate result; Merging the first intermediate result with the second intermediate result to obtain a target relation embedding array; Fusing the target relationship embedding array with the sensor embedding array to obtain the integrated embedding array; Among them, the dimensions of the target relationship embedding array, the sensor embedding array and the integrated embedding array are consistent.

9. The method according to claim 1, characterized in that: The method further comprises: Acquire an environmental sensor data set sample library, wherein each environmental sensor training data set in the environmental sensor data set sample library includes a priori event marker, the priori event marker in each environmental sensor training data set belongs to the event represented by the target component item in the event relationship network, and each environmental sensor training data set carries a corresponding priori construction monitoring result marker; For each of the environmental sensing training data sets, extracting a training local relationship network corresponding to the environmental sensing training data set in the event relationship network, wherein the training local relationship network includes the target component item, the component items within the preset path length range obtained according to the target component item, and the connecting lines between the component items; For each of the environmental sensing training data sets, obtaining a training relation embedding array based on the training local relation network and the first feature extraction component in the construction monitoring neural network; For each of the environmental sensor training data sets, obtaining a training sensor embedding array based on the second feature extraction component in the construction monitoring neural network according to the environmental sensor training data set; For each of the environmental sensor training data sets, generating a training integrated embedding array based on the training relational embedding array and the training sensor embedding array; For each of the environmental sensing training data sets, according to the training integrated embedding array, based on the decision component in the construction monitoring neural network, obtain the construction monitoring prediction result corresponding to the environmental sensing training data set; The construction monitoring neural network is trained according to the construction monitoring prediction results corresponding to each environmental sensing training data set and the errors marked by the priori construction monitoring results.

10. A monitoring system, characterized in that: include: one or more processors; Memory; One or more computer programs; wherein the one or more computer programs are stored in the memory and configured to be executed by the one or more processors, and when the one or more computer programs are executed by the processors, the method according to any one of claims 1 to 9 is implemented.

Images