4D millimeter wave ice detection method and platform based on fusion algorithm

Abstract

The present application provides a 4D millimeter wave ice detection method and platform based on a fusion algorithm, which relates to the field of ice detection technology, including: deploying a 4D millimeter radar, presetting trigger rules, and transmitting frequency modulated continuous waves for detection; performing spatial three-dimensional point cloud splicing, determining a four-dimensional signal point cloud, determining an algorithm cascade structure, and training a signal decision model for signal processing and feature extraction; determining a four-dimensional feature point cloud and identifying the distribution of over-limit point clouds, generating ice detection results and risk positioning columns, and performing visual warning and risk plan management on the terminal display interface. This application can solve the technical problem of low ice detection accuracy in the prior art due to limited recognition capabilities of ice type and thickness. By deploying 4D millimeter radar and multi-frequency detection technology, using multiple algorithm processing layers, and implementing over-limit point cloud distribution determination and visual warning, the accuracy of ice detection is improved.

Description

Technical Field

[0001] The present application relates to the technical field of ice detection, and in particular to a 4D millimeter wave ice detection method and platform based on a fusion algorithm. Background Art

[0002] Icing detection is a very important technology in the fields of power systems, communication lines, etc. It is mainly used to detect icing on transmission lines, cables and other facilities. Icing can cause the weight of transmission lines to increase, the structure to deform, and even break. Therefore, accurate detection of icing is crucial to ensure the safe operation of transmission lines. At present, existing icing detection methods include visual inspection, weighing, ultrasonic, infrared thermal imaging, vibration, radar, etc. The radar method uses electromagnetic waves in the radar band to identify icing by analyzing the characteristics of reflected waves, but this method is difficult to accurately distinguish different types of icing and accurately measure the thickness of icing, resulting in inaccurate assessment of icing conditions, which affects subsequent response measures.

[0003] In summary, the prior art has a technical problem in that the accuracy of ice detection is low due to the limited ability to identify the type and thickness of ice. Summary of the invention

[0004] The purpose of this application is to provide a 4D millimeter wave ice detection method and platform based on a fusion algorithm to solve the technical problem in the prior art that the accuracy of ice detection is low due to the limited ability to identify the type and thickness of ice.

[0005] In view of the above problems, the present application provides a 4D millimeter wave ice detection method and platform based on a fusion algorithm.

[0006] In the first aspect, the present application provides a 4D millimeter wave icing detection method based on a fusion algorithm, and the 4D millimeter wave icing detection method based on a fusion algorithm is implemented by a 4D millimeter wave icing detection platform based on a fusion algorithm, wherein the 4D millimeter wave icing detection method based on a fusion algorithm includes: deploying a 4D millimeter radar in a pre-detection area of ​​a transmission line, and setting a preset trigger rule, wherein the preset trigger rule includes a periodic trigger and an active trigger based on a preset environmental variable; generating a detection instruction based on the preset trigger rule, controlling the 4D millimeter radar to transmit a frequency modulated continuous wave, detecting the transmission line and receiving the echo signal, performing spatial three-dimensional point cloud splicing, and determining a four-dimensional signal Point cloud, the detection mode includes multi-target detection and target multi-frequency detection; based on the signal defects and detection feature requirements, the algorithm cascade structure is determined, and the signal decision model is supervised and trained. The signal decision model includes multiple algorithm processing layers, and the algorithm processing layer can execute empty wheels; the four-dimensional signal point cloud is read, and signal processing and feature extraction are performed in combination with the signal decision model to determine the four-dimensional feature point cloud and identify the distribution of the out-of-limit point cloud. The out-of-limit judgment includes feature vector out-of-limit and feature trend out-of-limit; based on the four-dimensional feature point cloud, the ice detection result is generated, and based on the out-of-limit point cloud distribution, a risk positioning column is generated; the risk positioning column is visually warned on the terminal display interface and the risk plan is managed.

[0007] In the second aspect, the present application also provides a 4D millimeter wave ice detection platform based on a fusion algorithm, which is used to execute the 4D millimeter wave ice detection method based on the fusion algorithm as described in the first aspect, wherein the 4D millimeter wave ice detection platform based on the fusion algorithm includes: a trigger rule setting module, the trigger rule setting module is used to deploy a 4D millimeter radar in the pre-detection area of ​​the transmission line, and set a preset trigger rule, the preset trigger rule includes a periodic trigger and an active trigger based on a preset environmental variable; a detection control module, the detection control module is used to generate a detection instruction based on the preset trigger rule, control the 4D millimeter radar to transmit a frequency modulated continuous wave, detect the transmission line and receive the echo signal, perform spatial three-dimensional point cloud splicing, determine the four-dimensional signal point cloud, the detection mode includes multi-target detection and target multi-frequency detection; signal decision A signal decision model training module, the signal decision model training module is used to determine the algorithm cascade structure based on signal defects and detection feature requirements, and supervise the training of the signal decision model. The signal decision model includes multiple algorithm processing layers, and the algorithm processing layer can execute empty wheels; a feature analysis module, the feature analysis module is used to read the four-dimensional signal point cloud, perform signal processing and feature extraction in combination with the signal decision model, determine the four-dimensional feature point cloud and identify the distribution of the out-of-limit point cloud, and the out-of-limit judgment includes feature vector out-of-limit and feature trend out-of-limit; an icing analysis module, the icing analysis module is used to generate icing detection results based on the four-dimensional feature point cloud, and generate a risk positioning column based on the out-of-limit point cloud distribution; an early warning and control module, the early warning and control module is used to perform terminal display interface visual early warning and risk plan control on the risk positioning column.

[0008] One or more technical solutions provided in this application have at least the following technical effects or advantages:

[0009] The 4D millimeter radar is deployed in the pre-detection area of ​​the transmission line, and preset trigger rules are set, wherein the preset trigger rules include periodic triggering and active triggering based on preset environmental variables; a detection instruction based on the preset trigger rules is generated, the 4D millimeter radar is controlled to transmit a frequency modulated continuous wave, the transmission line is detected and the echo signal is received, and spatial three-dimensional point cloud splicing is performed to determine the four-dimensional signal point cloud, and the detection mode includes multi-target detection and target multi-frequency detection; the algorithm cascade structure is determined based on the signal defects and detection feature requirements, and the signal decision model is supervised and trained, and the signal decision model includes multiple algorithm processing layers, and the algorithm processing layer can execute idle rounds; the four-dimensional signal point cloud is read, and signal processing and feature extraction are performed in combination with the signal decision model, and the four-dimensional feature point cloud is determined and the distribution of the over-limit point cloud is identified, and the over-limit judgment includes feature vector over-limit and feature trend over-limit; based on the four-dimensional feature point cloud, an ice detection result is generated, and a risk positioning column is generated based on the over-limit point cloud distribution; the risk positioning column is visually warned on the terminal display interface and the risk plan is controlled. In other words, by deploying 4D millimeter radar and multi-frequency detection technology, adopting multiple algorithm processing layers, and implementing out-of-limit point cloud distribution judgment and visual early warning, the accuracy of ice detection has been improved.

[0010] The above description is only an overview of the technical solution of the present application. In order to more clearly understand the technical means of the present application, it can be implemented according to the contents of the specification, and in order to make the above and other purposes, features and advantages of the present application more obvious and easy to understand, the specific implementation methods of the present application are specifically cited below. It should be understood that the content described in this section is not intended to identify the key or important features of the embodiments of the present application, nor is it intended to limit the scope of the present application. Other features of the present application will become easy to understand through the following description. BRIEF DESCRIPTION OF THE DRAWINGS

[0011] In order to more clearly illustrate the technical solutions in the present application or the prior art, the following briefly introduces the drawings required for use in the embodiments or the description of the prior art. Obviously, the drawings in the following description are only exemplary, and for ordinary technicians in this field, other drawings can be obtained based on the provided drawings without paying any creative work.

[0012] Figure 1 This is a flow chart of a 4D millimeter wave ice detection method based on a fusion algorithm in this application;

[0013] Figure 2 This is a structural diagram of the 4D millimeter wave ice detection platform based on the fusion algorithm of this application.

[0014] Explanation of the reference numerals: trigger rule setting module 11 , detection control module 12 , signal decision model training module 13 , feature analysis module 14 , ice analysis module 15 , early warning and management module 16 . DETAILED DESCRIPTION

[0015] This application solves the technical problem of low accuracy of ice detection in the prior art due to limited ability to identify ice type and thickness by providing a 4D millimeter wave ice detection method and platform based on fusion algorithm. By deploying 4D millimeter radar and multi-frequency detection technology, adopting multi-algorithm processing layer, implementing over-limit point cloud distribution judgment and visual early warning, the accuracy of ice detection is improved.

[0016] Below, the technical solutions in the present application will be clearly and completely described with reference to the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of the present application, rather than all of the embodiments of the present application. It should be understood that the present application is not limited to the example embodiments described herein. Based on the embodiments of the present application, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present application. It should also be noted that, for the convenience of description, only the parts related to the present application are shown in the accompanying drawings, rather than all of them.

[0017] For example, please refer to the attached Figure 1 The present application provides a 4D millimeter wave icing detection method based on a fusion algorithm, wherein the 4D millimeter wave icing detection method based on a fusion algorithm is applied to a 4D millimeter wave icing detection platform based on a fusion algorithm, and the 4D millimeter wave icing detection method based on a fusion algorithm specifically includes the following steps:

[0018] Step 1: Deploy a 4D millimeter radar in the pre-detection area of ​​the transmission line and set a preset trigger rule, which includes a periodic trigger and an active trigger based on preset environmental variables.

[0019] Specifically, multiple 4D millimeter wave radars are deployed in the pre-detection area of ​​the transmission line to provide high-precision detection tools for ice detection and realize comprehensive and real-time monitoring of the ice condition of the transmission line. 4D millimeter radar is a high-precision detection device that can add the time dimension on the basis of three-dimensional space to provide more comprehensive detection data. Trigger rules are formulated, including periodic triggering and active triggering based on environmental variables, to ensure the timeliness and adaptability of detection. The setting of preset trigger rules can enable the radar to automatically start detection under specific conditions, improving the pertinence and efficiency of detection. Periodic triggering refers to regular ice detection of transmission lines at set time intervals to ensure that the ice condition of transmission lines is continuously paid attention to and ice abnormalities are discovered in time. Active triggering based on preset environmental variables means that when the environmental variables (such as temperature, humidity, wind speed, etc.) reach the set threshold, the radar automatically starts detection, and in the case of increased ice risk, detection is carried out in advance to prevent ice accidents. By deploying 4D millimeter radars and setting preset trigger rules, comprehensive and real-time monitoring of the ice condition of transmission lines can be achieved, and the safe operation level of transmission lines can be improved.

[0020] Step 2: Generate detection instructions based on preset trigger rules, control the 4D millimeter radar to transmit frequency modulated continuous waves, detect the transmission lines and receive echo signals, perform spatial three-dimensional point cloud stitching, and determine the four-dimensional signal point cloud. The detection mode includes multi-target detection and target multi-frequency detection.

[0021] Specifically, according to the pre-set trigger rules, the corresponding detection instructions are automatically generated to control the 4D millimeter radar to detect at a specific time or environmental conditions. The detection instruction is a command used to guide the radar to detect, including when to start detection, the frequency of detection, the duration of detection, etc. After the instruction is generated, the 4D millimeter radar is controlled to transmit frequency modulated continuous waves to accurately detect the transmission line. Frequency modulated continuous wave (FMCW) is a radar signal transmission technology that detects the position and speed of the target object by continuously changing the transmission frequency. The FMCW signal emitted by the radar will be reflected when it encounters the transmission line and ice. The radar receives these echo signals and processes the received signals, including filtering, amplification, mixing, etc., to extract useful information. By analyzing the frequency changes of the echo signals, the distance, speed and angle information of the target can be calculated, thereby obtaining the ice coverage on the transmission line. Using the detection data of multiple radars, the spatial three-dimensional point cloud is spliced ​​to create a three-dimensional model of the transmission line. Combined with the time dimension, the three-dimensional point cloud is expanded into a four-dimensional signal point cloud to observe and analyze the changes of ice coverage over time. The four-dimensional signal point cloud contains the position of the target object in space, time information, and changes over time, which is very important for ice detection. The detection mode includes multi-target detection and target multi-frequency detection. Multi-target detection allows the system to detect multiple target objects at the same time, that is, different parts or different points on the transmission line; while target multi-frequency detection is for specific high-risk targets or targets in special locations. It can be detected at different frequencies to obtain more comprehensive information. The four-dimensional signal point cloud combined with the time dimension not only provides spatial information, but also records the changes of ice over time, which helps to analyze the development trend of ice and predict future status.

[0022] Step 3: Determine the algorithm cascade structure based on signal defects and detection feature requirements, and supervise the training of the signal decision model, wherein the signal decision model includes multiple algorithm processing layers, and the algorithm processing layer can execute an empty round.

[0023] Specifically, first, analyze the signal defects and detection feature requirements to understand the specific problems that need to be solved in ice detection, such as distinguishing different types of ice and identifying changes in ice thickness. Select appropriate algorithms according to the requirements, such as convolutional neural networks (CNN) for image recognition, long short-term memory networks (LSTM) for time series data analysis, and support vector machines (SVM) for classification. Cascade the selected algorithms in a certain order to form a processing chain. Each algorithm layer solves a specific problem and then passes the output to the next layer. The algorithm cascade structure is a structure that connects multiple algorithms in sequence. Each algorithm processing layer is responsible for different signal processing tasks, which can improve the accuracy and efficiency of signal processing because each algorithm processing layer can focus on specific signal processing tasks, thereby improving the performance of the entire system. Next, collect a large amount of ice detection data, including radar echo signals, ice thickness and type and other label information. Pre-train each algorithm layer, combine the pre-trained algorithm layers, and use supervised learning methods to jointly train the entire model.

[0024] The signal decision model is a machine learning-based model that can output the decision result of the signal based on the input signal data. This model contains multiple algorithm processing layers, each of which is responsible for different signal processing tasks. The first layer of algorithms focuses on extracting useful features from the original signal, such as edge detection, texture analysis, etc.; the subsequent algorithm layers will enhance or transform the extracted features to better adapt to classification or regression tasks; the final algorithm layer makes the final decision based on the features processed by the previous layers, such as ice classification or thickness estimation. In order to improve the accuracy and efficiency of the model, supervised training of this model is required to ensure that each processing layer can perform its tasks correctly. The algorithm processing layer can perform idle rounds, which means that each algorithm processing layer can perform idle rounds when necessary to improve the robustness and stability of the model. Idle rounds are a temporary suspension of the training of a processing layer during the model training process to avoid overfitting or other problems. It can improve the generalization ability of the model and enable it to maintain good performance in different environments and conditions. Through multiple algorithm processing layers, more complex features can be extracted from the original signal. Each algorithm layer focuses on different feature extraction and pattern recognition tasks, so that the model can understand the ice data more comprehensively.

[0025] Step 4: Read the four-dimensional signal point cloud, perform signal processing and feature extraction in combination with the signal decision model, determine the four-dimensional feature point cloud and identify the distribution of the out-of-limit point cloud. The out-of-limit judgment includes feature vector out-of-limit and feature trend out-of-limit.

[0026] Specifically, the four-dimensional signal point cloud is read, which contains the position of the target object detected by the radar in space, time information, and changes over time. Four-dimensional signal point cloud data, which usually contains spatial coordinates and timestamps, forms a four-dimensional data set. The pre-processed data is further processed using signal processing technology, such as using Fourier transform to analyze the signal frequency components, or using wavelet transform to analyze the time-frequency characteristics of the signal. Combined with the signal decision model, the features in the data are extracted, including but not limited to: geometric features, physical features, and timing features. The data is input into the multi-algorithm processing layer in the signal decision model. Each algorithm layer performs a specific task. For example, the first layer focuses on basic feature extraction, such as edge detection; the second layer performs feature enhancement, such as using principal component analysis (PCA) to reduce dimensionality; and subsequent layers perform more complex pattern recognition and classification tasks. By combining the signal decision model, the four-dimensional signal point cloud is deeply analyzed and processed to extract key feature information.

[0027] Determine the four-dimensional feature point cloud, which refers to the point cloud data containing the position of the target object in space, time information, and changes over time. By extracting features, the key information of the point cloud data is obtained to determine whether the transmission line is covered with ice. The over-limit judgment includes feature vector over-limit and feature trend over-limit. Feature vector over-limit means that a feature in the point cloud data exceeds the preset threshold, indicating that the ice state is abnormal; and feature trend over-limit means that the change trend of a feature in the point cloud data over time exceeds the preset threshold, indicating that the ice state is rapidly deteriorating. Extracting key features from the four-dimensional signal point cloud and judging whether there is an over-limit point cloud distribution based on these features can more comprehensively evaluate the ice state and predict possible risks.

[0028] Step 5: Generate ice detection results based on the four-dimensional feature point cloud, and generate a risk location column based on the distribution of the out-of-limit point cloud.

[0029] Specifically, by using the extracted four-dimensional feature point cloud, by extracting and analyzing these features, it is determined whether there is icing on the transmission line and the severity of the icing. For example, key parameters such as ice thickness and distribution range can be determined to generate icing detection results. The results reflect the icing status on the transmission line, including information such as ice thickness, distribution, and type. The over-limit point cloud distribution refers to all point cloud data that exceeds the preset threshold. By integrating these over-limit point cloud data, a risk positioning column is generated, showing the specific location, risk level, possible impact, etc. of the icing risk. This column provides decision makers with detailed information on the icing risk, helping to understand which areas need special attention and measures. By generating a risk positioning column, accurate icing detection results and risk positioning columns can be provided to help operators promptly identify potential risks and take corresponding measures to ensure the safe operation of the transmission line.

[0030] Step 6: Perform visual warning and risk plan management on the terminal display interface for the risk positioning column.

[0031] Specifically, the information in the risk location column is visualized through the terminal display interface, so that the operator can intuitively see the location, risk level and possible impact of the risk point. Visual warnings can be displayed through various graphics and charts, and color coding can be used to represent areas with different risk levels, or highlighting can be used to highlight risk areas, such as heat maps of risk points, bar charts of risk levels, etc., so that operators can quickly understand and identify risks. Through the visual warning of the terminal display interface, decision makers and maintenance personnel can quickly identify and understand the location and severity of icing risks, which helps to make decisions quickly and take appropriate measures to ensure the safe operation of transmission lines. Traverse the risk location column, determine the risk plan and identify multiple concurrent risk points. Based on the spatial location distribution and the urgency of the risk, make collision avoidance decisions for multiple concurrent risk points. Combined with the control constraint characteristics, the phase avoidance control of the execution of the risk plan is carried out. Combined with the risk plan control, according to the information in the risk location column, formulate corresponding risk plans, including preventive measures, emergency response strategies, maintenance plans, etc., to effectively respond to and mitigate the possible impact of icing risks. Visual early warning and risk plan management on the terminal display interface can provide intuitive risk warning and effective risk plan management, helping operators to promptly identify and respond to potential risks, thereby improving the safe operation level of transmission lines.

[0032] Further, step 2 of this application includes:

[0033] Based on the coverage of the first coded radar, multiple detection targets based on a preset step interval are determined, and a first detection scheme is determined. The continuous timestamp is used as the emission constraint, and the first coded radar is any one of the 4D millimeter radars; the second detection scheme is determined by multi-frequency detection of special detection targets, and the special detection targets are high-risk detection targets or special position detection targets; the first detection scheme and the second detection scheme are fitted to determine the detection scheme of the first coded radar, and the first coded radar is controlled to perform frequency modulated continuous wave transmission and echo reception.

[0034] Specifically, according to the range that any specific radar in the 4D millimeter radar can effectively detect, multiple detection targets are determined. These targets are located in the radar coverage area and are selected according to the preset step interval. They can be different parts or different points on the transmission line. The preset step interval means that when determining the detection target, the time point is selected according to the preset distance interval. A detection plan is formulated for each detection target, taking into account the distribution of the detection target and the performance parameters of the radar. With continuous timestamps as emission constraints, detection will be performed at a series of continuous time points to ensure the continuity and stability of the radar signal. For special detection targets, such as high-risk detection targets or special position detection targets, multi-frequency detection is used to determine the second detection plan. Multi-frequency detection means detecting the same target at different frequencies to obtain more comprehensive information. High-risk detection targets are areas with a higher risk of icing due to environmental conditions, structural characteristics or other factors, while special position detection targets are due to their importance or inaccessibility to the operation of the transmission line.

[0035] The first detection scheme is fitted with the second detection scheme to determine a comprehensive detection scheme, which involves adjusting the detection parameters, detection frequency, etc. to ensure that the two schemes can be effectively combined. The purpose of fitting is to ensure that the two detection schemes can work in coordination without interfering with each other, and can provide the most comprehensive ice coverage information. The fitted detection scheme is then used to determine the specific detection scheme of the first coding radar. According to this scheme, the first coding radar is controlled to transmit frequency modulated continuous waves and receive echo signals. Frequency modulated continuous wave transmission is a radar signal transmission method that detects the position and speed of the target object by continuously changing the frequency. Echo reception is used to receive the signal reflected from the target object by the signal transmitted by the radar. This detection method can provide more accurate target position and speed information, thereby improving the accuracy of detection. Through this complex detection scheme, the 4D millimeter radar can detect ice coverage more accurately, quickly identify high-risk areas, and provide detailed information on the ice coverage of the transmission line, including the location, thickness and changes over time of ice coverage, which helps to take preventive measures in time to prevent line failures and safety risks caused by ice coverage.

[0036] Furthermore, the present application also includes the following steps:

[0037] Determine N detection coordinate systems of the 4D millimeter radar, which have a one-to-one correspondence with the 4D millimeter radar; determine a main coordinate system covering the pre-detection area, and obtain the spatial conversion relationship between the N detection coordinate systems and the main coordinate system; determine N four-dimensional local point clouds based on the echo signal, and perform spatial three-dimensional conversion pre-point cloud stitching based on the main coordinate system in combination with the spatial conversion relationship to determine the four-dimensional signal point cloud.

[0038] Specifically, for each radar in the 4D millimeter radar, a detection coordinate system is determined. This coordinate system is one-to-one with each radar and is used to describe the location of the spatial points detected by the radar. The coordinate system defines the position and direction of the radar in space, as well as the coverage of each detection unit. The 4D millimeter wave radar system is able to provide more accurate distance and speed information by adding the time dimension (t) on the basis of the traditional three spatial dimensions (x, y, z). This high-precision detection capability is critical for ice detection, as ice may cause changes in the geometry and electrical characteristics of the transmission line. Determine the main coordinate system covering the pre-detection area, which describes the standard coordinate system of the entire pre-detection area, usually corresponding to the geographic coordinate system or the coordinate system of a specific area. Obtaining the spatial transformation relationship between each detection coordinate system and the main coordinate system involves calculating the transformation relationship between the coordinates of each detection unit of the radar and the main coordinate system. These transformation relationships include translation and rotation, which are used to transform the point cloud data in the detection coordinate system into the main coordinate system.

[0039] Each radar determines a local point cloud based on the received echo signal. This point cloud contains the position information and time information of the spatial points within the radar detection range, that is, the coordinates in four dimensions. When determining the local point cloud, the time axis remains unchanged, that is, each point in the point cloud contains time information, indicating the state of the point at different times. Through these point cloud data, the ice condition on the transmission line can be detected more accurately. Spatial three-dimensional conversion pre-point cloud splicing is a key technology in the ice detection system. It involves converting point cloud data from different locations into a unified coordinate system and splicing them into a continuous three-dimensional model. The local point cloud of each radar is converted to the main coordinate system using the spatial conversion relationship between the detection coordinate system and the main coordinate system. The converted local point cloud is spliced ​​in the main coordinate system to form a complete four-dimensional signal point cloud. This point cloud contains not only spatial information, but also time information, which can be used to analyze the changes of ice over time. By determining a detection coordinate system for each 4D millimeter radar and establishing a spatial conversion relationship with the main coordinate system, the local point cloud detected by each radar can be accurately converted into a unified reference frame, improving the spatial positioning accuracy of the ice coverage of the transmission line. The deployment of 4D millimeter radars and the use of multiple detection coordinate systems ensure full coverage of the pre-detection area, and every part of the transmission line can be effectively monitored, reducing monitoring blind spots.

[0040] Further, step 4 of this application includes:

[0041] Traverse the multiple algorithm processing layers, set data screening levels and configure them at the processing layer ports; transmit the four-dimensional signal point cloud to the first algorithm processing layer, and determine the preprocessing part and the idle part based on the data screening level; the first algorithm processing layer performs the signal point cloud processing of the preprocessing part to determine the first processing result; fuse the first processing result and the idle part, execute the iterative processing flow of the algorithm processing layer, and integrate and determine the four-dimensional feature point cloud.

[0042] Specifically, each algorithm processing layer in the signal decision model is traversed, and each processing layer has its specific algorithm and processing requirements, such as filtering, feature extraction, pattern recognition, etc. By traversing these processing layers, the data type and features required by each processing layer can be identified. A data screening level is set at the entrance of each algorithm processing layer. The data screening level refers to a specific threshold or condition set based on the characteristics, quality, importance or other standards of the data during the data processing process, which is used to filter out data that meets the requirements of the processing layer and determine which data needs to be processed at this layer and which data can be skipped (i.e., an empty round is executed). These levels can be executed for the part of the point cloud that has the algorithm processing requirements of this layer, and the data screening level is configured at the input port of the algorithm processing layer to ensure that only the necessary data is passed to the next processing layer. Since not all data are necessary for processing, the screening level can avoid passing unnecessary data to the subsequent processing layer, thereby saving processing resources and improving processing speed. At the same time, the screening level can also help the system eliminate noise and interference, improve the purity of the signal, and thus improve the accuracy and reliability of ice detection.

[0043] The four-dimensional signal point cloud data is transmitted to the first algorithm processing layer of the signal decision model. According to the data screening level, the data is divided into two parts: one part is the signal point cloud that needs to be preprocessed, and the other part is the signal point cloud that does not need to be processed in this layer (empty part). The corresponding processing operations such as filtering, denoising, coordinate transformation, feature extraction, etc. are performed on the signal point cloud that needs to be preprocessed to improve the quality and accuracy of the data. Filtering can remove noise and outliers in the point cloud data, denoising can eliminate random noise in the point cloud data, coordinate transformation can convert the point cloud data into a unified coordinate system, and feature extraction is to extract useful information for subsequent analysis and decision-making from the processed signal. Through preprocessing, the first algorithm processing layer determines the first processing result, including the processed point cloud data, feature extraction results, etc. The first processing result and the data of the empty part are fused, including merging, marking or reorganizing the data. The fused data is passed to the next algorithm processing layer to continue iterative processing. Each subsequent processing layer will decide whether to process the incoming data according to its data screening level. Through iterative processing, the processing results of each algorithm layer are gradually integrated, and finally a four-dimensional feature point cloud containing rich features is determined. By optimizing the data processing process, the efficiency and pertinence of ice detection are achieved.

[0044] Furthermore, the present application also includes the following steps:

[0045] Taking the echo frequency change as the identification target, the four-dimensional signal point cloud is traversed to extract the Doppler frequency shift feature and determine the relative motion feature. The relative motion feature is the transient change feature of the ice layer. There is a linear relationship between the frequency shift vector and the echo frequency change. The relative motion feature is added to the point cloud feature class.

[0046] Specifically, the four-dimensional feature point cloud refers to the point cloud data with the time dimension added to the three-dimensional space coordinate system. These point clouds contain detection echo signals at different time and space positions. Each point cloud corresponds to a detection echo signal at a time and space position, so multiple feature types can be extracted from these point clouds, such as geometric features, physical features, and time series features. The multi-point cloud feature class refers to a set of point cloud data with similar features in the point cloud data, which can be classified based on the position, shape, texture and other features of the point cloud. When extracting features, the frequency change of the echo signal is used as the recognition target, because the frequency change can reflect the information of the tiny vibration or deformation of the surface of the object. The four-dimensional signal point cloud is traversed, and the Doppler shift feature is extracted for each point cloud. The Doppler shift refers to the change in the frequency of the echo signal caused by the relative motion of the target object. By analyzing the frequency change of the echo signal, the relative motion characteristics of the target object are determined. The relative motion characteristics refer to the motion state of the target object relative to the radar, such as speed, direction, etc. In ice detection, the growth and shedding of ice may be accompanied by tiny vibrations or changes, which can be detected by millimeter wave radar. Based on this change, the dynamic change analysis of ice layer is carried out to more accurately describe the state and change trend of ice layer. The relative motion characteristics can reflect the transient changes of ice layer during the growth or shedding process, that is, the motion state of ice layer at a certain moment.

[0047] There is a linear relationship between the frequency shift vector and the echo frequency change, which means that in the Doppler effect, there is a simple mathematical relationship between the speed of an object relative to the radar and the change in the echo frequency received by the radar. When the signal emitted by the radar encounters a moving target, the frequency of the signal reflected by the target will change. This frequency change is related to the speed of the target, which is specifically manifested as: when the target is far away from the radar, the frequency of the reflected signal decreases; when the target is close to the radar, the frequency of the reflected signal increases. The frequency shift vector refers to this frequency change, which is proportional to the speed of the target relative to the radar. In practical applications, the speed of the target can be calculated by measuring the frequency shift vector. Adding relative motion features to the point cloud feature class can more accurately describe the characteristics and properties of point cloud data. By analyzing the relationship between the frequency shift vector and the echo frequency change, the dynamic changes of the ice layer, including the growth rate and shedding rate of the ice layer, can be identified, which helps to timely detect changes in the ice layer, so as to take corresponding measures to ensure the safe operation of the transmission line.

[0048] Furthermore, the present application also includes the following steps:

[0049] The historical detection data of the pre-detection area is read, risk feature mining is performed, and a risk feature library is constructed, wherein the risk feature satisfies a preset frequency, and the risk feature library includes a risk type-feature vector threshold / feature trend threshold; the four-dimensional feature point cloud is traversed, feature matching and over-limit judgment are performed based on the risk feature library, and risk type identification is performed, the distribution of the over-limit point cloud is integrated and determined, and the four-dimensional feature point cloud is identified.

[0050] Specifically, ice detection data of the pre-detection area over a period of time in the past is collected and read, including ice detection results conducted in the same area at different time points in the past, including information such as the thickness, distribution, and shape of the detected ice layer. Risk characteristics refer to characteristics related to ice risk, such as the thickness, shape, and distribution of the ice layer. These risk characteristics are identified and extracted by analyzing historical data. The risk feature library is a database that stores risk characteristics, which contains risk types and corresponding feature vector thresholds or feature trend thresholds. Risk types refer to different states or degrees of ice, such as light ice, heavy ice, etc. The feature vector threshold refers to the threshold for judging the ice state, and the feature trend threshold refers to the threshold for judging the development trend of ice. For example, some data of the pre-detection area in January of a certain year in the past are as follows: January 5: temperature -2°C, humidity 60%, wind speed 6m / s, echo intensity 10dB, ice thickness 2mm, risk level is slight; January 8: temperature -3°C, humidity 68%, wind speed 4m / s, echo intensity 14dB, ice thickness 5mm, risk level is slight; January 10: temperature -4°C, humidity 70%, wind speed 3m / s, echo intensity 18dB, ice thickness 8mm, risk level is moderate; January 15: temperature -3°C, humidity 68%, wind speed 4m / s, echo intensity 14dB, ice thickness 5mm , the risk level is mild; January 18: temperature -5℃, humidity 72%, wind speed 2m / s, echo intensity 20dB, ice thickness 12mm, the risk level is moderate; January 20: temperature -6℃, humidity 83%, wind speed 1m / s, echo intensity 29dB, ice thickness 22mm, the risk level is severe; January 25: temperature -7℃, humidity 78%, wind speed 1m / s, echo intensity 33dB, ice thickness 16mm, the risk level is severe; January 30: temperature -4℃, humidity 76%, wind speed 3m / s, echo intensity 17dB, ice thickness 9mm, the risk level is moderate. Based on these data, a risk feature library can be constructed as follows: Mild risk: temperature threshold > -3℃, humidity threshold < 70%, wind speed threshold > 4m / s, echo intensity < 15dB, ice thickness threshold < 5mm; Moderate risk: temperature threshold -5℃ to -3℃, humidity threshold 70% to 80%, wind speed threshold 2m / s to 4m / s, echo intensity 15dB to 25dB, ice thickness threshold 5mm to 15mm; Severe risk: temperature threshold < -5℃, humidity threshold > 80%, wind speed threshold < 2m / s, echo intensity > 25dB, ice thickness threshold > 15mm. Note: The lower the temperature, the higher the ice risk; the higher the humidity, the higher the ice risk; the lower the wind speed, the higher the ice risk; the greater the echo intensity, the more severe the ice; the greater the ice thickness, the more severe the ice.

[0051] Traverse the four-dimensional feature point cloud, which contains the position of the target object detected by the radar in space, time information, and changes over time. Match each point in the four-dimensional feature point cloud with the features in the risk feature library to determine whether each point in the point cloud data meets the definition of a certain feature. Over-limit determination involves checking whether each feature in the point cloud data exceeds a preset threshold. These thresholds are set based on understanding and experience of normal conditions, and they define the safe range of feature values. If the measured value of a feature exceeds this threshold, the feature is judged to be over-limit, which means there is an icing risk. For example, if there is a feature in the risk feature library called "ice thickness" and its threshold is set to 10 mm, and the ice thickness of a certain point detected by the system is 12 mm, then this point is judged to be over-limit, and the system will mark this point as an area with icing risk. If a feature in the point cloud data exceeds the preset threshold, it will be identified as a specific risk type. For example, if the ice thickness exceeds a certain threshold, the system will mark it as a severe icing risk.

[0052] All feature point clouds that are judged to be out of limit are integrated to determine the distribution of the out-of-limit point clouds, that is, the spatial distribution of the out-of-limit features. In the four-dimensional feature point cloud, the out-of-limit features are marked for subsequent processing and analysis. The four-dimensional feature point cloud of the pre-detection area is read in real time, and feature matching and out-of-limit judgment are performed based on the risk feature library, realizing the automation of ice detection.

[0053] Furthermore, step six of this application includes:

[0054] Traverse the risk location column, determine the risk plan and identify multiple concurrent risk points; based on the spatial location distribution and the risk urgency, make collision avoidance decisions for the multiple concurrent risk points and determine the management and control constraint characteristics; combine the management and control constraint characteristics to perform phase avoidance control of the execution of the risk plan.

[0055] Specifically, the risk location column is a list containing information about icing risk areas, detailing the location, range, risk type, and other information of the over-limit point cloud distribution. By traversing this list, each risk point can be analyzed in detail to determine its characteristics and possible impacts. According to the characteristics and possible impacts of each risk point, such as risk type, risk level, urgency, etc., corresponding risk plans are formulated. These plans include preventive measures, emergency response strategies, maintenance plans, etc., in order to effectively respond to and mitigate the possible impacts of icing risks. In some cases, icing detection will find multiple simultaneous and interrelated risk points, and these multiple concurrent risk points need to be managed and responded to at the same time. By identifying these risk points, relevant personnel can have a more comprehensive understanding of the icing situation and formulate more effective risk management strategies.

[0056] Based on the relative position and distribution of risk points in space and the urgency of the risk, collision avoidance decisions are made for multiple concurrent risk points. The urgency of the risk is usually related to factors such as the severity of ice cover, the scope of impact, and the potential consequences. Each risk point is evaluated based on these factors and its urgency is determined. Collision avoidance decision means that the system conducts a comprehensive analysis and decision based on the relative position and distribution of risk points in space, as well as the urgency of each risk point. Through collision avoidance decision, the interaction between risk points and the potential collision risk can be evaluated, and it can be determined how to avoid these risks.

[0057] Based on the collision avoidance decision, the control constraint characteristics are determined, including resource constraints, time constraints, priority constraints, etc. The control constraint characteristics refer to the restrictions or rules set to ensure that the ice risk is effectively managed and controlled, including specific operating procedures, safety distance requirements, maintenance plans, etc. By determining these constraint characteristics, it can be ensured that potential collisions and conflicts can be effectively avoided when dealing with multiple concurrent risk points.

[0058] Combined with the characteristics of control constraints, the execution of risk plans is controlled in a phased avoidance manner, that is, the execution of different risk plans is coordinated and arranged in time. Phased avoidance control refers to determining the processing priority and schedule based on factors such as resources, equipment, and operation limitations when dealing with multiple simultaneous and interrelated risk points. For example, if there are multiple risk points that need to be responded to at the same time, but the concurrent plan response cannot be met due to constraints such as equipment and resources, avoidance analysis is performed to determine the processing priority. By conducting risk plan control, multiple concurrent risk points can be effectively managed to ensure that risk plans can be executed in an orderly and effective manner under limited resources, thereby minimizing the impact of risks on the safe operation of transmission lines.

[0059] In summary, the 4D millimeter wave ice detection method based on the fusion algorithm provided in this application has the following technical effects:

[0060] The 4D millimeter radar is deployed in the pre-detection area of ​​the transmission line, and preset trigger rules are set, wherein the preset trigger rules include periodic triggering and active triggering based on preset environmental variables; a detection instruction based on the preset trigger rules is generated, the 4D millimeter radar is controlled to transmit a frequency modulated continuous wave, the transmission line is detected and the echo signal is received, and spatial three-dimensional point cloud splicing is performed to determine the four-dimensional signal point cloud, and the detection mode includes multi-target detection and target multi-frequency detection; the algorithm cascade structure is determined based on the signal defects and detection feature requirements, and the signal decision model is supervised and trained, and the signal decision model includes multiple algorithm processing layers, and the algorithm processing layer can execute idle rounds; the four-dimensional signal point cloud is read, and signal processing and feature extraction are performed in combination with the signal decision model, and the four-dimensional feature point cloud is determined and the distribution of the over-limit point cloud is identified, and the over-limit judgment includes feature vector over-limit and feature trend over-limit; based on the four-dimensional feature point cloud, an ice detection result is generated, and a risk positioning column is generated based on the over-limit point cloud distribution; the risk positioning column is visually warned on the terminal display interface and the risk plan is controlled. In other words, by deploying 4D millimeter radar and multi-frequency detection technology, adopting multiple algorithm processing layers, and implementing out-of-limit point cloud distribution judgment and visual early warning, the accuracy of ice detection has been improved.

[0061] Embodiment 2, based on the same inventive concept as the 4D millimeter wave ice detection method based on the fusion algorithm in the previous embodiment, the present application also provides a 4D millimeter wave ice detection platform based on the fusion algorithm, please refer to the attached Figure 2 The 4D millimeter wave ice detection platform based on fusion algorithm includes:

[0062] The trigger rule setting module 11 is used to deploy 4D millimeter radar in the pre-detection area of ​​the transmission line and set preset trigger rules, wherein the preset trigger rules include periodic triggering and active triggering based on preset environmental variables.

[0063] The detection control module 12 is used to generate detection instructions based on preset trigger rules, control the 4D millimeter radar to transmit frequency modulated continuous waves, detect the transmission lines and receive echo signals, perform spatial three-dimensional point cloud stitching, and determine the four-dimensional signal point cloud. The detection mode includes multi-target detection and target multi-frequency detection.

[0064] The signal decision model training module 13 is used to determine the algorithm cascade structure based on signal defects and detection feature requirements, and supervise the training of the signal decision model. The signal decision model includes multiple algorithm processing layers, and the algorithm processing layers can execute empty rounds.

[0065] The feature analysis module 14 is used to read the four-dimensional signal point cloud, perform signal processing and feature extraction in combination with the signal decision model, determine the four-dimensional feature point cloud and identify the distribution of the out-of-limit point cloud. The out-of-limit judgment includes feature vector out-of-limit and feature trend out-of-limit.

[0066] The ice analysis module 15 is used to generate an ice detection result based on the four-dimensional feature point cloud and generate a risk positioning column based on the distribution of the over-limit point cloud.

[0067] The early warning and control module 16 is used to perform visual early warning on the terminal display interface and risk plan control on the risk positioning column.

[0068] Furthermore, the detection control module 12 in the 4D millimeter wave ice detection platform based on the fusion algorithm is also used for:

[0069] Based on the coverage of the first coded radar, multiple detection targets based on a preset step interval are determined, and a first detection scheme is determined. The continuous timestamp is used as the emission constraint, and the first coded radar is any one of the 4D millimeter radars; the second detection scheme is determined by multi-frequency detection of special detection targets, and the special detection targets are high-risk detection targets or special position detection targets; the first detection scheme and the second detection scheme are fitted to determine the detection scheme of the first coded radar, and the first coded radar is controlled to perform frequency modulated continuous wave transmission and echo reception.

[0070] Furthermore, the detection control module 12 in the 4D millimeter wave ice detection platform based on the fusion algorithm is also used for:

[0071] Determine N detection coordinate systems of the 4D millimeter radar, which have a one-to-one correspondence with the 4D millimeter radar; determine a main coordinate system covering the pre-detection area, and obtain the spatial conversion relationship between the N detection coordinate systems and the main coordinate system; determine N four-dimensional local point clouds based on the echo signal, and perform spatial three-dimensional conversion pre-point cloud stitching based on the main coordinate system in combination with the spatial conversion relationship to determine the four-dimensional signal point cloud.

[0072] Furthermore, the feature analysis module 14 in the 4D millimeter wave ice detection platform based on the fusion algorithm is also used for:

[0073] Traverse the multiple algorithm processing layers, set data screening levels and configure them at the processing layer ports; transmit the four-dimensional signal point cloud to the first algorithm processing layer, and determine the preprocessing part and the idle part based on the data screening level; the first algorithm processing layer performs the signal point cloud processing of the preprocessing part to determine the first processing result; fuse the first processing result and the idle part, execute the iterative processing flow of the algorithm processing layer, and integrate and determine the four-dimensional feature point cloud.

[0074] Furthermore, the feature analysis module 14 in the 4D millimeter wave ice detection platform based on the fusion algorithm is also used for:

[0075] Taking the echo frequency change as the identification target, the four-dimensional signal point cloud is traversed to extract the Doppler frequency shift feature and determine the relative motion feature. The relative motion feature is the transient change feature of the ice layer. There is a linear relationship between the frequency shift vector and the echo frequency change. The relative motion feature is added to the point cloud feature class.

[0076] Furthermore, the feature analysis module 14 in the 4D millimeter wave ice detection platform based on the fusion algorithm is also used for:

[0077] The historical detection data of the pre-detection area is read, risk feature mining is performed, and a risk feature library is constructed, wherein the risk feature satisfies a preset frequency, and the risk feature library includes a risk type-feature vector threshold / feature trend threshold; the four-dimensional feature point cloud is traversed, feature matching and over-limit judgment are performed based on the risk feature library, and risk type identification is performed, the distribution of the over-limit point cloud is integrated and determined, and the four-dimensional feature point cloud is identified.

[0078] Furthermore, the early warning and control module 16 in the 4D millimeter wave ice detection platform based on the fusion algorithm is also used for:

[0079] Traverse the risk location column, determine the risk plan and identify multiple concurrent risk points; based on the spatial location distribution and the risk urgency, make collision avoidance decisions for the multiple concurrent risk points and determine the management and control constraint characteristics; combine the management and control constraint characteristics to perform phase avoidance control of the execution of the risk plan.

[0080] The various embodiments in this specification are described in a progressive manner, and each embodiment focuses on the differences from other embodiments. Figure 1The 4D millimeter wave icing detection method based on fusion algorithm and the specific examples in the first embodiment are also applicable to the 4D millimeter wave icing detection platform based on fusion algorithm in this embodiment. Through the above detailed description of the 4D millimeter wave icing detection method based on fusion algorithm, those skilled in the art can clearly know the 4D millimeter wave icing detection platform based on fusion algorithm in this embodiment, so for the sake of brevity of the specification, it will not be described in detail here. For the platform disclosed in the embodiment, since it corresponds to the method disclosed in the embodiment, the description is relatively simple, and the relevant parts can be referred to the method part description.

[0081] The above description of the disclosed embodiments enables those skilled in the art to implement or use the present application. Various modifications to these embodiments will be apparent to those skilled in the art, and the general principles defined herein may be implemented in other embodiments without departing from the spirit or scope of the present application. Therefore, the present application will not be limited to the embodiments shown herein, but will conform to the widest scope consistent with the principles and novel features disclosed herein.

[0082] Obviously, those skilled in the art can make various changes and modifications to the present application without departing from the spirit and scope of the present application. Thus, if these modifications and variations of the present application belong to the scope of the present application and its equivalent technology, the present application is also intended to include these modifications and variations.

Claims

1. A 4D millimeter wave ice detection method based on a fusion algorithm is characterized by: include: Deploy 4D millimeter radar in the pre-detection area of ​​the power transmission line and set preset trigger rules, wherein the preset trigger rules include periodic triggering and active triggering based on preset environmental variables; Generate detection instructions based on preset trigger rules, control the 4D millimeter radar to transmit frequency modulated continuous waves, detect the transmission line and receive echo signals, perform spatial three-dimensional point cloud stitching, and determine the four-dimensional signal point cloud. The detection mode includes multi-target detection and target multi-frequency detection; Based on the signal defects and detection feature requirements, the algorithm cascade structure is determined, and the signal decision model is supervised and trained. The signal decision model includes multiple algorithm processing layers, and the algorithm processing layer can execute an empty round; Read the four-dimensional signal point cloud, perform signal processing and feature extraction in combination with the signal decision model, determine the four-dimensional feature point cloud and identify the distribution of the over-limit point cloud, and the over-limit judgment includes feature vector over-limit and feature trend over-limit; Based on the four-dimensional feature point cloud, an ice detection result is generated, and based on the distribution of the over-limit point cloud, a risk location column is generated; Conduct visual early warning and risk plan management on the terminal display interface for the risk positioning column; The controlling the 4D millimeter radar to transmit a frequency modulated continuous wave comprises: Determine a first detection scheme based on a first coded radar coverage, a plurality of detection targets based on a preset step interval, and a first detection scheme with continuous timestamps as emission constraints, wherein the first coded radar is any one of the 4D millimeter radars; Determine a second detection scheme by multi-frequency detection of a special detection target, wherein the special detection target is a high-risk detection target or a special position detection target; Fitting the first detection scheme and the second detection scheme to determine the detection scheme of the first coding radar, and controlling the first coding radar to perform frequency modulated continuous wave transmission and echo reception; Receive echo signals, perform spatial 3D point cloud stitching, and determine 4D signal point cloud, including: Determine N detection coordinate systems of the 4D millimeter radar, where the detection coordinate systems correspond one to one with the 4D millimeter radar; Determine a main coordinate system covering the pre-detection area, and obtain a spatial transformation relationship between the N detection coordinate systems and the main coordinate system; Determine N four-dimensional local point clouds based on the echo signal, and perform spatial three-dimensional conversion pre-point cloud splicing based on the principal coordinate system in combination with the spatial conversion relationship to determine the four-dimensional signal point cloud; The determining of the four-dimensional feature point cloud comprises: Traversing the multi-algorithm processing layer, setting data screening levels and configuring them at the processing layer ports; Transmitting the four-dimensional signal point cloud to the first algorithm processing layer, and determining a preprocessing part and an idle part based on the data screening level; The first algorithm processing layer performs the signal point cloud processing of the preprocessing part to determine a first processing result; The first processing result and the idle wheel part are integrated, and the iterative processing flow of the algorithm processing layer is executed to integrate and determine the four-dimensional feature point cloud.

2. The 4D millimeter wave ice detection method based on fusion algorithm according to claim 1, characterized in that: Each 4D feature point cloud contains multiple point cloud feature classes for feature extraction, including: Taking the echo frequency change as the identification target, traversing the four-dimensional signal point cloud, extracting the Doppler frequency shift feature, and determining the relative motion feature, the relative motion feature is the transient change feature of the ice layer, and there is a linear relationship between the frequency shift vector and the echo frequency change; The relative motion feature is added to the point cloud feature class.

3. The 4D millimeter wave ice detection method based on fusion algorithm according to claim 1, characterized in that: The identification of the out-of-limit point cloud distribution includes: Read the historical detection data of the pre-detection area, perform risk feature mining, and build a risk feature library, wherein the risk feature meets the preset frequency, and the risk feature library includes risk type-feature vector threshold / feature trend threshold; The four-dimensional feature point cloud is traversed, feature matching and over-limit determination are performed based on the risk feature library, and risk type identification is performed, the distribution of the over-limit point cloud is integrated and determined, and the four-dimensional feature point cloud is identified.

4. The 4D millimeter wave ice detection method based on fusion algorithm according to claim 1, characterized in that: Conduct risk plan management and control on the risk positioning column, including: Traverse the risk location list, determine the risk plan and identify multiple concurrent risk points; Based on the spatial location distribution and risk urgency, collision avoidance decisions are made for the multiple concurrent risk points to determine the control constraint characteristics; In combination with the control constraint characteristics, temporal avoidance control is performed during the execution of the risk plan.

5. The 4D millimeter wave ice detection platform based on fusion algorithm is characterized by: The steps for implementing the 4D millimeter wave icing detection method based on the fusion algorithm as described in any one of claims 1 to 4, the 4D millimeter wave icing detection platform based on the fusion algorithm comprises: A trigger rule setting module, which is used to deploy a 4D millimeter radar in the pre-detection area of ​​the power transmission line and set a preset trigger rule, wherein the preset trigger rule includes a periodic trigger and an active trigger based on a preset environmental variable; A detection control module, which is used to generate detection instructions based on preset trigger rules, control the 4D millimeter radar to transmit frequency modulated continuous waves, detect the transmission line and receive echo signals, perform spatial three-dimensional point cloud stitching, and determine the four-dimensional signal point cloud. The detection mode includes multi-target detection and target multi-frequency detection; A signal decision model training module, wherein the signal decision model training module is used to determine the algorithm cascade structure based on signal defects and detection feature requirements, and supervise the training of the signal decision model, wherein the signal decision model includes multiple algorithm processing layers, and the algorithm processing layer can execute an empty round; A feature analysis module, which is used to read the four-dimensional signal point cloud, perform signal processing and feature extraction in combination with the signal decision model, determine the four-dimensional feature point cloud and identify the distribution of the over-limit point cloud, and the over-limit judgment includes feature vector over-limit and feature trend over-limit; An icing analysis module, the icing analysis module is used to generate an icing detection result based on the four-dimensional feature point cloud, and generate a risk positioning column based on the distribution of the over-limit point cloud; The early warning and control module is used to perform visual early warning and risk plan control on the risk positioning column on the terminal display interface.

Images