Intersection signal control system zero sample fault detection method, device and equipment

Abstract

The invention relates to the technical field of information, in particular to an intersection signal control system zero sample fault detection method, device and equipment and a medium, and the method comprises the steps: constructing an intersection signal control state sensing system, and obtaining intersection signal control to-be-detected fault data; the method comprises the following steps: performing feature extraction on known fault data by utilizing a neural network model to obtain a feature vector, performing mapping processing on vocabularies in a fault text by utilizing a word vector model to obtain a semantic vector, and constructing a semantic space based on the feature vector and the semantic vector; and judging the fault type of the signal control to-be-detected fault data of the intersection by using the semantic space, and outputting the fault type and probability. A new thought is provided for the detection and classification of unknown faults in a semantic space construction and multi-modal data fusion mode. In combination with deep learning and generalized zero sample learning methods, accurate detection and classification of known faults and unknown faults are realized.

Description

Technical Field

[0001] The present application relates to the field of information technology, and in particular to a zero-sample fault detection method, device, equipment, and computer storage medium for an intersection signal control system. Background Art

[0002] With the rapid advancement of intelligent transportation, fault detection in intelligent transportation systems has also gained increasing attention. Real-time information in intelligent transportation systems is often transmitted as data, making fault detection in traffic control systems particularly important. Effective intelligent fault detection classification methods and systems can promptly diagnose problems such as traffic signal failures, thereby preventing vehicle collisions caused by abnormal traffic lights, reducing traffic accidents, and ensuring pedestrian safety. Failures in traffic lights, signals, or lines at intersections can easily lead to traffic congestion. Intelligent traffic fault detection systems can promptly detect signal and line failures, enabling traffic management departments to quickly repair and restore normal traffic signal timing, ensuring smooth passage for vehicles and pedestrians and reducing the formation and spread of traffic congestion. Intelligent traffic fault diagnosis also helps traffic management departments better understand the operating status of transportation facilities and rationally allocate resources for maintenance and upgrades. Intelligent traffic fault detection can accurately locate the fault location and cause, saving significant manpower, material resources, and time costs and improving maintenance efficiency. An efficient and intelligent transportation system is a key hallmark of urban modernization. Intersections are crucial hubs in transportation routes. The continuous improvement and advancement of intersection fault detection technology will help promote the development of the entire intelligent transportation system.

[0003] However, in practice, intersection signal control systems generate a large amount of real-time data. Intersection signal control systems' fault detection utilizes this massive amount of data for analysis and judgment. Processing this large amount of data requires the use of artificial intelligence algorithms. Existing deep learning fault diagnosis methods are primarily supervised learning methods. Models typically require a large amount of labeled data for training. Insufficient data or low-quality data may lead to degraded model performance, resulting in overfitting or underfitting. When new fault types that have not appeared in the training data emerge, the model may not be able to accurately identify and diagnose them. Re-collection and labeling of relevant data are necessary, and the model must be retrained and optimized. This can result in delayed responses to new faults.

[0004] In summary, how to design an efficient and accurate intersection signal control system fault detection method is an urgent problem that needs to be solved. Summary of the Invention

[0005] The present application aims to solve one of the technical problems in the related art at least to a certain extent.

[0006] To this end, the first purpose of this application is to propose a zero-sample fault detection method for an intersection signal control system to solve the problem that existing technical means cannot accurately identify and diagnose the fault type of the intersection signal control system.

[0007] The second object of this application is to provide a device.

[0008] The third objective of this application is to provide an electronic device.

[0009] The fourth object of this application is to provide a computer-readable storage medium.

[0010] To achieve the above objectives, the first embodiment of the present application proposes a zero-sample fault detection method for an intersection signal control system, comprising:

[0011] Build an intersection signal control status perception system to obtain intersection signal control fault data to be detected;

[0012] Using a neural network model to extract features from known fault data to obtain feature vectors, using a word vector model to map the vocabulary in the fault text to obtain semantic vectors, and constructing a semantic space based on the feature vectors and the semantic vectors;

[0013] The semantic space is used to determine the fault type of the intersection signal control fault data to be detected, and the fault category and probability are output.

[0014] Preferably, the method of extracting features from known fault data using a neural network model to obtain a feature vector, mapping vocabulary in the fault text using a word vector model to obtain a semantic vector, and constructing a semantic space based on the feature vector and the semantic vector includes:

[0015] The known fault data is input into the neural network model for training, feature extraction and classification are performed, and the features are integrated and reduced in dimension through the neural network model to generate a complete feature vector;

[0016] Use the word vector model to map the words in the fault text into vectors, and combine the vectors of each word in the text to generate a semantic vector;

[0017] The feature vector and the semantic vector are normalized to construct a semantic space.

[0018] Preferably, the determining the fault type of the intersection signal control fault data to be detected by using the semantic space includes:

[0019] Feature extraction is performed on the intersection signal control fault data to be detected, and the extracted feature vector is input into the semantic space. A distance value is obtained by a similarity distance calculation clustering method. If the distance value meets a first preset value, it is classified as a known fault; if the distance value meets a second preset value, it is classified as an unknown fault.

[0020] Preferably, if the distance value meets the first preset value, the classification as a known fault includes: inputting the intersection signal control fault data to be detected classified as a known fault into the neural network model for verification and comparison.

[0021] Preferably, if the distance value meets the second preset value, classifying it as an unknown fault includes:

[0022] An unknown fault classification space is constructed based on a multimodal data fusion method. The intersection signal control fault data to be detected that is classified as an unknown fault is input into the unknown fault space for distance clustering calculation to obtain the unknown fault classification result.

[0023] Preferably, the method of using the semantic space to determine the fault type of the intersection signal control fault data to be detected and outputting the fault category and probability includes:

[0024] Using a function to convert the score of the output layer of the trained neural network model into a probability distribution, and obtain the classification result and probability of the known fault;

[0025] A function is used to convert the similarity into a probability output to obtain the classification result and probability of the unknown fault.

[0026] Preferably, the construction of the intersection signal control status perception system and the acquisition of intersection signal control fault data to be detected include: constructing an intersection signal control status perception system based on an intersection signal control perception detector, a data transmission device and a software monitoring system, and using the intersection signal control status perception system to acquire intersection signal control fault data to be detected.

[0027] To achieve the above-mentioned purpose, the second embodiment of the present application proposes a zero-sample fault detection device for an intersection signal control system, comprising:

[0028] System construction module, builds the intersection signal control status perception system, and obtains the intersection signal control fault data to be detected;

[0029] A semantic space construction module uses a neural network model to extract features from known fault data to obtain feature vectors, uses a word vector model to map the vocabulary in the fault text to obtain semantic vectors, and constructs a semantic space based on the feature vectors and the semantic vectors;

[0030] The fault category classification module uses the semantic space to determine the fault type of the intersection signal control fault data to be detected, and outputs the fault category and probability.

[0031] To achieve the above-mentioned purpose, a third embodiment of the present application provides an electronic device, comprising: a processor, and a memory communicatively connected to the processor;

[0032] The memory stores computer-executable instructions;

[0033] The processor executes the computer-executable instructions stored in the memory to implement any of the above methods.

[0034] To achieve the above-mentioned purpose, the fourth embodiment of the present application proposes a computer-readable storage medium, including computer-executable instructions stored in the computer-readable storage medium, and the computer-executable instructions are used to implement any of the methods described above when executed by a processor.

[0035] The present application provides a zero-sample fault detection method for an intersection signal control system, which can effectively distinguish and classify known faults and unknown faults without the need for additional labeled data. By constructing an intersection signal control state perception system and by building a semantic space and multimodal data fusion method, a new approach is provided for the detection and classification of unknown faults. Combining deep learning with generalized zero-sample learning methods, accurate detection and classification of known and unknown faults are achieved. This method can effectively improve the fault detection capability of intelligent traffic signal control systems and provide technical support for rapid response and efficient maintenance for traffic management departments.

[0036] Additional aspects and advantages of the present application will be given in part in the description below, and in part will become apparent from the description below, or will be learned through practice of the present application. BRIEF DESCRIPTION OF THE DRAWINGS

[0037] The above and / or additional aspects and advantages of the present application will become apparent and easily understood from the following description of the embodiments in conjunction with the accompanying drawings, in which:

[0038] Figure 1 This is a flowchart of a first specific embodiment of a zero-sample fault detection method for an intersection signal control system provided by the present invention;

[0039] Figure 2 This is a flow chart of the zero-sample fault detection method for intersection signal control system;

[0040] Figure 3 This is a schematic diagram of the intersection signal control state perception system;

[0041] Figure 4 To construct a schematic diagram of the semantic space of known faults;

[0042] Figure 5 This is a schematic diagram of the method for distinguishing known faults from unknown faults;

[0043] Figure 6 Schematic diagram for classification of unknown faults;

[0044] Figure 7 To construct an unknown fault prediction diagram in the unknown fault semantic space;

[0045] Figure 8 This is a structural block diagram of a zero-sample fault detection device for an intersection signal control system provided by an embodiment of the present invention. DETAILED DESCRIPTION

[0046] The core of the present invention is to provide a zero-sample fault detection method, device, electronic equipment and medium for intersection signal control systems. By constructing an intersection signal control status perception system and combining multimodal data with semantic vectors, accurate discrimination and classification of known and unknown faults can be achieved.

[0047] In order to enable those skilled in the art to better understand the present invention, the present invention is further described in detail below with reference to the accompanying drawings and specific embodiments. Obviously, the embodiments described are only some embodiments of the present invention, not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative work are within the scope of protection of the present invention.

[0048] Please refer to Figure 1 , Figure 1 This is a flowchart of a first specific embodiment of a zero-sample fault detection method for an intersection signal control system provided by the present invention; the specific operating steps are as follows:

[0049] Step S101: constructing an intersection signal control state perception system to obtain intersection signal control fault data to be detected;

[0050] An intersection signal control state perception system is constructed based on an intersection signal control perception detector, a data transmission device, and a software monitoring system, and the intersection signal control state perception system is used to obtain intersection signal control fault data to be detected.

[0051] Intersection signal control perception detectors include but are not limited to: video detectors, geomagnetic detectors, current detectors, voltage detectors, photoelectric detectors and other detection equipment.

[0052] Data transmission equipment: Wired transmission equipment includes but is not limited to copper cable transmission equipment, optical fiber, and power line communication (PLC) equipment; wireless transmission equipment includes but is not limited to a transmitter board containing a wireless transmitter module for generating and transmitting traffic light control signals, a receiver board for receiving the transmitted traffic light control signals, and a single-chip microcomputer for controlling the traffic light display.

[0053] The software system includes but is not limited to: database software system (relational database, non-relational database); data analysis and processing software; user interface (UI) software (which facilitates traffic management personnel to view the traffic status and signal light status of the intersection through the browser, and operate signal light timing adjustment and other functions); system monitoring and management software.

[0054] Step S102: Using a neural network model to extract features from known fault data to obtain feature vectors, using a word vector model to map the vocabulary in the fault text to obtain semantic vectors, and constructing a semantic space based on the feature vectors and semantic vectors;

[0055] The known fault data is input into the neural network model for training, feature extraction and classification are performed, and the features are integrated and reduced in dimension through the neural network model to generate a complete feature vector;

[0056] Use the word vector model to map the words in the fault text into vectors, and combine the vectors of each word in the text to generate a semantic vector;

[0057] The feature vector and the semantic vector are normalized to construct a semantic space.

[0058] In one embodiment, known fault data generated by the system is fed into a neural network for training to extract and classify features. The training networks used for feature extraction and classification include, but are not limited to, convolutional neural networks (CNNs), recurrent neural networks (RNNs) and their variants, and the discriminator in a generative adversarial network (GAN).

[0059] The extracted feature vectors are normalized and concatenated with the fault semantic vectors to construct a semantic space for fault detection testing. Fusion of feature vectors and fault semantic vectors includes but is not limited to simple sequential concatenation, weighted concatenation, multi-layer perceptron (MLP), covariance fusion, and semantic embedding fusion.

[0060] Step S103: using the semantic space to determine the fault type of the intersection signal control fault data to be detected, and outputting the fault category and probability.

[0061] The features of the intersection signal control fault data to be detected are extracted, and the extracted feature vectors are input into the semantic space. The distance value is obtained by the similarity distance calculation clustering method. If the distance value meets the first preset value, it is classified as a known fault. If the distance value meets the second preset value, it is classified as an unknown fault.

[0062] The intersection signal control fault data to be detected, which is classified as a known fault, is input into the neural network model for verification and comparison. The score of the output layer of the trained neural network model is converted into a probability distribution using a function to obtain the classification result and probability of the known fault.

[0063] An unknown fault classification space is constructed based on the multimodal data fusion method. The intersection signal control fault data classified as unknown faults are input into the unknown fault space for distance clustering calculation to obtain the unknown fault classification results.

[0064] Use a function to convert similarity into probability output to obtain the classification results and probability of unknown faults.

[0065] In one embodiment, real-time fault data to be detected undergoes feature extraction to obtain feature vectors, which are then entered into a semantic space. A similarity distance calculation clustering method is used to classify known faults into unknown faults based on the calculated distance values. This similarity distance calculation utilizes the DTW (Dynamic Time Warping) method. The DTW (Dynamic Time Warping) method is highly adaptable to nonlinear changes in time series, can handle time series of varying lengths, and exhibits strong noise immunity.

[0066] Construct an unknown fault classification space. The feature vectors of the unknown fault classification space are constructed using multimodal data fusion. The multimodal data sources for unknown fault prediction include, but are not limited to, the system's own detector data (voltage, current, image, and frequency), equipment operation history data, environmental data, and technical text data such as equipment manuals, operating instructions, and design drawings.

[0067] The data identified as unknown faults enter the unknown fault classification space, and the distance clustering calculation is performed on the data features in the unknown fault classification space to classify the unknown faults.

[0068] in,

[0069] Output fault category: Fault categories are broadly divided into: signal light fault, signal machine fault, and line fault. Each broad fault category can be further divided into several subcategories based on the results.

[0070] Traffic light group failures can be divided into: bulb failure; internal circuit failure of the light group; incorrect signal light color display; physical damage to the lamp; signal light power failure.

[0071] Traffic light failures can be divided into: hardware failures; software failures; and failures caused by external interference.

[0072] Line faults can be divided into: open circuit faults; short circuit faults; line interference faults.

[0073] Output fault classification probability:

[0074] Known faults are extracted and classified using a trained neural network. A function is used at the network's output layer to convert the scores into a probability distribution, directly obtaining the classification results and probabilities. Output layer probability distribution functions include, but are not limited to, the softmax function.

[0075] After the unknown fault is classified by the system, a function is used to convert the similarity into a probability output. The probability generation function is a Gaussian kernel function.

[0076] The fault information is fed back to the bus-type intersection signal control system to report the fault for repair, and the intersection is automatically adjusted and controlled.

[0077] This embodiment provides a zero-sample fault detection method for an intersection signal control system, which can effectively distinguish and classify known faults and unknown faults without the need for additional labeled data. By constructing an intersection signal control state perception system and building a semantic space and multimodal data fusion approach, a new approach is provided for the detection and classification of unknown faults. Combining deep learning with generalized zero-sample learning methods, accurate detection and classification of known and unknown faults is achieved. This method can effectively improve the fault detection capability of intelligent traffic signal control systems and provide technical support for rapid response and efficient maintenance for traffic management departments.

[0078] Based on the above embodiment, this embodiment describes a zero-sample fault detection method for an intersection signal control system. Figure 2 As shown, the details are as follows:

[0079] (1) Constructing an intersection signal control status perception system:

[0080] like Figure 3 As shown, the intersection signal control perception detector includes video detectors, geomagnetic detectors, current detectors, voltage detectors, photoelectric detectors and other detection equipment. Video detectors are installed in monitoring cameras and other equipment, and the signal light group itself is equipped with current detectors, voltage detectors, and photoelectric detectors. Current detectors and voltage detectors are connected at the connection between the equipment (monitoring equipment, signal light group, signal machine). In addition to real-time detection of the equipment itself, these detectors can also detect the entire area near the equipment through changes in current and voltage. The equipment is connected by wires, and information is transmitted wirelessly. The geomagnetic detector can also measure factors such as the strength of the geomagnetic field in the environment.

[0081] The system's hardware and software are interconnected. The software system configuration includes a database system (relational or non-relational); data analysis and processing software; user interface (UI) software; and system monitoring and management software. Real-time data received by the hardware is wirelessly transmitted to the database system. Relational databases enable preliminary analysis of fault factors, while non-relational databases support multi-source data input and, through the use of data processing software, enable in-depth mining of potential information. The user interface provides managers with a visual interface, facilitating subsequent systematic management and overall planning.

[0082] After completing the combination of the hardware system and the software system, the system can be used to generate data for training and verification.

[0083] (2) Known fault detection network training:

[0084] Convolutional neural networks are used to extract features and classify test samples.

[0085] Normalize and standardize the data to ensure that the data value range is within the appropriate range to facilitate neural network learning. The data can also be cleaned to remove noise data and outliers.

[0086] After data processing, it is labeled to clearly identify the major fault category (such as signal light group, signal machine fault, line fault, etc.) or normal state corresponding to each data sample. Then, each major fault category is divided into subcategories.

[0087] Traffic light group failures can be divided into: bulb failure; internal circuit failure of the light group; incorrect signal light color display; physical damage to the lamp; signal light power failure.

[0088] Traffic light failures can be divided into: hardware failures; software failures; and failures caused by external interference.

[0089] Line faults can be divided into: open circuit faults; short circuit faults; line interference faults.

[0090] The training samples are input into the convolutional neural network for feature extraction, and the output is converted into probability distribution of each category.

[0091] In one embodiment, the known fault detection network training includes:

[0092] Fault data is fed into a neural network for training, feature extraction, and classification. The data is fed into a convolutional neural network and passes through the input layer, convolution layer, activation function layer, pooling layer, fully connected layer, and output layer. The pooling layer integrates and reduces the features, and the fully connected layer forms a complete feature vector. The fully connected layer determines the image category based on the previously extracted image features, and then outputs the predicted value.

[0093] Concatenate the feature vector with the semantic vector of the fault semantic description to construct a semantic space: Use a word embedding model (such as Word2Vec or BERT) to map words in the text into vector representations. Then, using a method such as average pooling, combine the vectors of each word in the text to form a semantic vector that represents the fault semantic description. Vector concatenation can be done directly or with weighted concatenation. The semantic vector and feature vector are concatenated to form a semantic space.

[0094] Eigenvector where f i represents the value of the i-th feature, and n is the number of features.

[0095] Fault semantic description semantic vector where s j represents the value of the jth semantic dimension, and m is the dimension of the semantic vector.

[0096] Concatenate these two vectors to construct a new vector in the semantic space To ensure comparability of features and semantics across different dimensions, vectors must undergo some preprocessing or transformation, such as normalization. The concatenation formula after normalization is as follows.

[0097] Calculate eigenvectors Length of mold

[0098] Calculate the semantic vector of fault semantic description Length of mold

[0099] Get the normalized eigenvector Normalized fault semantic description semantic vector Splicing

[0100] (3) Identify known faults or unknown faults:

[0101] Acquisition of semantic vectors: Using word embedding technology: Using the Word2Vec word embedding model, we train the corpus of various fault-related texts in (2). These models can map each word into a low-dimensional vector space, and can directly obtain the semantic vectors of fault-related words.

[0102] Semantic space construction: Figure 4As shown, the semantic vectors and the feature vectors of the training samples are weighted and concatenated according to the accuracy of the description. The concatenated vectors are then linearly transformed to construct a semantic space. In this semantic space, each dimension represents a comprehensive semantic feature, encompassing information from both the feature vector and the semantic vector. Let's assume the feature vector is x, the semantic vector is y, and the weights are w1 and w2, respectively (where w1 + w2 = 1). The concatenated vector z can be expressed as z = w1x + w2y.

[0103] After the test samples are input into the semantic space, they are judged and the clustering method is calculated using the similarity distance, such as Figure 5 As shown, faults are divided into known faults and unknown faults. If the distance between the test sample and the vector in the space is very far, it is an unknown fault and enters (4) for further classification prediction. If there is a vector close to the vector in the space, it can be judged as a known fault and the known fault enters the neural network in (2) for classification.

[0104] In one embodiment, determining whether a known fault or an unknown fault (DTW method) includes:

[0105] Determining whether a fault is known or unknown: The test fault input semantic space is used to perform distance calculation using a clustering algorithm. Similarity calculations are performed using DTW dynamic time warping, and clustering is performed using the K-means method. If the distance calculated by the clustering algorithm is close to that of known fault vectors, the fault is considered to be of the same type. If the distance is far from all vectors in the space, the fault is considered to be unknown.

[0106] Known faults enter step S2 for classification.

[0107] Unknown faults enter the S4 unknown fault space for further classification.

[0108] DTW (Dynamic Time Warping) is a method for measuring the similarity between two time series. It is used to calculate the similarity distance between two time series. Time series are not completely aligned on the time axis. DTW "bends" or "warps" the time series on the time axis to find the best match between the two time series, minimizing the distance between them and measuring similarity. The calculation formula is as follows:

[0109] Calculate the distance matrix: For two vectors X={x1,x2,…x m} and Y={y1,y2,…y n}, calculate the distance between their elements and form an m×n matrix D, where D(i,j) represents x i and y j The distance between them is usually expressed as Euclidean distance

[0110] Initialize the cumulative distance matrix: create an m×n matrix M and initialize M(1,1)=D(1,1). Initialize the first row and first column of M at the same time, M(i,1)=M(i-1,1)+D(i,1)(i>1); M(1,j)=M(1,j-1)+D(1,j)(j>1)

[0111] Dynamic programming calculates the cumulative distance: Use the dynamic programming method to fill in the other elements of M M(i,j) = M(i,j) + min{M(i-1,j),M(i,j-1),M(i-1,j-1)}, where i>1 and j>1. This step considers the cumulative distance from three possible directions (up, left, upper left) to the current position, and selects the smallest one plus the distance of the current position.

[0112] Backtrack to find the optimal path: Start from M(m,n) and backtrack. According to the source of the cumulative distance (i.e., M(i,j) is the minimum value obtained by M(i-1,j), M(i,j-1), and M(i-1,j-1)), the optimal path is determined. This path reflects the best alignment between the two time series.

[0113] Obtaining the regularized distance: The final value of M(m,n) is the DTW distance between the two time series. The smaller the value, the more similar the two vectors are.

[0114] The K-Means clustering method repeatedly calculates the distance between the sample point and the cluster center, assigns the sample point to the cluster to which the nearest cluster center belongs, and updates the cluster center until the cluster center no longer changes or reaches the preset number of iterations or other stopping conditions.

[0115] (4) Unknown fault classification

[0116] The data identified as unknown faults in (3) are classified again using the clustering similarity distance calculation method, such as Figure 6 As shown, unknown faults can be classified.

[0117] Multimodal semantic spaces can provide rich semantic information, integrating multiple types of data. These different modalities of information complement each other in the semantic space, providing more comprehensive semantic clues for understanding faults. Semantic information from different modalities can also help fill knowledge gaps caused by a lack of training samples in zero-shot scenarios.

[0118] In one embodiment, calculating the distance unknown fault classification includes:

[0119] The data feature vectors identified as unknown faults in step S3 are first subjected to the same DTW similarity distance calculation (the principle is the same as above), and the unknown faults can be divided into different categories.

[0120] Multimodal data fusion constructs an unknown fault classification space: Based on current and voltage changes, multimodal splicing is performed in combination with image and frequency features to obtain a joint feature vector. This joint feature vector is then combined with the semantic vector to form an unknown fault classification space. By inputting the fault, the unknown fault type can be predicted.

[0121] Perform classification operations on unknown faults.

[0122] The execution formula of the multimodal data fusion and splicing principle is:

[0123] Current characteristic vector: I=(i1,i2,…i m )

[0124] Voltage characteristic vector: V=(v1,v2,…v n )

[0125] Image feature vector: I M =(im1,im2,…im p )

[0126] Frequency eigenvector: F=(f1,f2,…f q )

[0127] Semantic vector: S=(s1,s2,…s r )

[0128] Multimodal splicing to obtain joint feature vector: The joint feature vector J is obtained by splicing the current, voltage, image and frequency feature vectors in sequence.

[0129] Construct unknown fault classification space: Concatenate the joint feature vector J and the semantic vector S to form the unknown fault classification space vector U.

[0130] like Figure 7 Through experiments, the current and voltage data, image features and frequency features of the equipment status monitor under fault conditions are combined respectively, and the joint feature vectors of different combinations are obtained by splicing them according to the semantic vector to form the semantic space of unknown faults.

[0131] The unknown fault is input into the unknown fault semantic space and the distance calculation method of clustering is used. When the distance between the unknown fault vector and the vector in a certain category of unknown fault classification space is closest, the specific fault type of the unknown fault can be predicted.

[0132] (5) Output fault category and probability

[0133] The system can output detailed information for both classified known faults and unknown faults, including fault category and probability.

[0134] In one embodiment, outputting the fault category and probability includes:

[0135] The system confirms the location of the fault: the fault information is sent to the system software to detect the fault information, fault type and probability.

[0136] Identify the major fault category (such as signal light group, signal machine fault, line fault, etc.) or normal state corresponding to each data sample. Then divide each major fault category into subcategories.

[0137] Traffic light failures can be categorized as: bulb failure (completely off, abnormal brightness, flickering); internal circuit failure (short circuit, open circuit); incorrect signal light color display (color cast, mixed colors); physical damage to the lamp (cracked lampshade, deformed lamp body); signal light power failure (unstable power supply, damaged power module).

[0138] Traffic light failures can be categorized as: hardware failures (chassis and internal component damage, interface failures, power failures, heat dissipation problems); software failures (program crashes or freezes, control logic errors, communication software failures); and failures caused by external interference (electromagnetic interference, environmental interference).

[0139] Line faults can be divided into: open circuit faults (line aging and breakage, external force damage, loose connection points); short circuit faults (insulation layer damage, environmental impact); line interference faults (electromagnetic interference, signal attenuation).

[0140] Use a function at the output layer of the network to output the known fault probability: Take the softmax function as an example. The softmax function converts a real number vector into a probability distribution.

[0141] After the unknown fault is classified, a function is used to convert the similarity into probability. The Gaussian kernel function can be used:

[0142] Information feedback and fault reporting: After receiving the fault information, maintenance personnel will confirm the fault and report it. The repair operation can be based on the fault type and severity. The system software can directly control the rapid recovery of the hardware and perform manual repairs for serious faults.

[0143] In summary, the present invention's "Zero-Sample Fault Detection Method and System for Intersection Signal Control Systems" builds an intersection signal control status perception system, collects data from multiple detectors in real time, and uploads it to a database for preprocessing. Through network training and semantic space similarity calculation, it can distinguish between known and unknown faults. Similarity calculation is also applied to unknown faults to derive a classification. Multimodal information can also be used to predict the type of unknown fault. Finally, the fault classification and probability are output.

[0144] Please refer to Figure 8 , Figure 8 This is a structural block diagram of a zero-sample fault detection device for an intersection signal control system provided by an embodiment of the present invention; the specific device may include:

[0145] System construction module 100, constructing an intersection signal control state perception system and acquiring intersection signal control fault data to be detected;

[0146] The semantic space construction module 200 uses a neural network model to extract features from known fault data to obtain feature vectors, uses a word vector model to map the vocabulary in the fault text to obtain semantic vectors, and constructs a semantic space based on the feature vectors and the semantic vectors;

[0147] The fault category classification module 300 uses the semantic space to determine the fault type of the intersection signal control fault data to be detected, and outputs the fault category and probability.

[0148] A zero-sample fault detection device for an intersection signal control system in this embodiment is used to implement the aforementioned zero-sample fault detection method for an intersection signal control system. Therefore, the specific implementation method of a zero-sample fault detection device for an intersection signal control system can be seen in the embodiment part of the zero-sample fault detection method for an intersection signal control system in the previous text. For example, the system construction module 100, the semantic space construction module 200, and the fault category classification module 300 are respectively used to implement steps S101, S102, and S103 in the aforementioned zero-sample fault detection method for an intersection signal control system. Therefore, its specific implementation method can refer to the description of the corresponding embodiments of each part and will not be repeated here.

[0149] In order to implement the above embodiments, the present application also proposes an electronic device, comprising: a processor, and a memory communicatively connected to the processor; the memory stores computer-executable instructions; the processor executes the computer-executable instructions stored in the memory to implement the method provided by the above embodiments.

[0150] In order to implement the above embodiments, the present application also proposes a computer-readable storage medium, in which computer-executable instructions are stored. When the computer-executable instructions are executed by a processor, they are used to implement the methods provided by the above embodiments.

[0151] In order to implement the above embodiments, the present application also proposes a computer program product, including a computer program, which implements the methods provided by the above embodiments when executed by a processor.

[0152] The collection, storage, use, processing, transmission, provision and disclosure of user personal information involved in this application are in compliance with relevant laws and regulations and do not violate public order and good morals.

[0153] It is important to note that personal information collected from users should be used for legitimate and reasonable purposes and should not be shared or sold beyond these legitimate uses. Furthermore, such collection / sharing should be conducted only after receiving the user's informed consent, including but not limited to notifying the user to read the user agreement / user notice and sign an agreement / authorization that includes the relevant user information before using the feature. Furthermore, any necessary steps must be taken to safeguard and secure access to such personal information and ensure that others with access to personal information comply with its privacy policy and procedures.

[0154] This application contemplates providing implementations that allow users to selectively block the use or access of personal information data. Specifically, this disclosure contemplates providing hardware and / or software to prevent or block access to such personal information data. Risks can be minimized by limiting data collection and deleting data once it is no longer needed. Furthermore, where applicable, such personal information can be de-identified to protect user privacy.

[0155] In the descriptions of the foregoing embodiments, the reference terms "one embodiment", "some embodiments", "example", "specific example", or "some examples" mean that the specific features, structures, materials or characteristics described in conjunction with the embodiment or example are included in at least one embodiment or example of the present application. In this specification, the schematic expressions of the above terms do not necessarily refer to the same embodiment or example. Moreover, the specific features, structures, materials or characteristics described may be combined in any one or more embodiments or examples in a suitable manner. In addition, those skilled in the art may combine and combine the different embodiments or examples described in this specification and the features of the different embodiments or examples, unless they are mutually inconsistent.

[0156] Furthermore, the terms "first" and "second" are used for descriptive purposes only and should not be construed as indicating or implying relative importance or implicitly specifying the number of the technical features being referred to. Thus, a feature defined as "first" or "second" may explicitly or implicitly include at least one of such features. Throughout the description of this application, "plurality" means at least two, for example, two, three, etc., unless otherwise specifically defined.

[0157] Any process or method description in a flowchart or otherwise described herein may be understood to represent a module, segment or portion of code comprising one or more executable instructions for implementing the steps of a custom logical function or process, and the scope of the preferred embodiments of the present application includes alternative implementations in which functions may be performed out of the order shown or discussed, including performing functions in a substantially simultaneous manner or in the reverse order depending on the functions involved, which should be understood by those skilled in the art to which the embodiments of the present application belong.

[0158] The logic and / or steps represented in the flowcharts or otherwise described herein, for example, can be considered as a sequenced list of executable instructions for implementing the logical functions, and can be embodied in any computer-readable medium for use by, or in conjunction with, an instruction execution system, apparatus, or device (e.g., a computer-based system, a system including a processor, or other system that can fetch and execute instructions from an instruction execution system, apparatus, or device). For purposes of this specification, a "computer-readable medium" can be any device that can contain, store, communicate, propagate, or transport a program for use by, or in conjunction with, an instruction execution system, apparatus, or device. More specific examples (a non-exhaustive list) of computer-readable media include the following: an electrical connection with one or more wires (electronic devices), a portable computer disk cartridge (magnetic device), random access memory (RAM), read-only memory (ROM), erasable and programmable read-only memory (EPROM or flash memory), fiber optic devices, and a portable compact disc read-only memory (CDROM). Furthermore, the computer-readable medium may even be paper or other suitable medium on which the program is printed, since the program may be obtained electronically, for example, by optically scanning the paper or other medium and then editing, interpreting or processing it in another suitable manner if necessary, and then storing it in a computer memory.

[0159] It should be understood that various parts of the present application can be implemented using hardware, software, firmware, or a combination thereof. In the above embodiments, multiple steps or methods can be implemented using software or firmware stored in a memory and executed by a suitable instruction execution system. For example, if implemented using hardware, as in another embodiment, any one of the following technologies known in the art or a combination thereof can be used to implement: a discrete logic circuit having a logic gate circuit for implementing a logic function on a data signal, an application-specific integrated circuit having a suitable combination of logic gate circuits, a programmable gate array (PGA), a field programmable gate array (FPGA), etc.

[0160] Those skilled in the art will understand that all or part of the steps in the method of the above embodiment can be completed by instructing related hardware through a program, and the program can be stored in a computer-readable storage medium. When the program is executed, it includes one or a combination of the steps of the method embodiment.

[0161] In addition, the functional units in the various embodiments of the present application may be integrated into a processing module, or each unit may exist physically separately, or two or more units may be integrated into a module. The above-mentioned integrated module may be implemented in the form of hardware or in the form of a software functional module. If the integrated module is implemented in the form of a software functional module and sold or used as an independent product, it may also be stored in a computer-readable storage medium.

[0162] The storage medium mentioned above may be a read-only memory, a magnetic disk, or an optical disk, etc. Although the embodiments of the present application have been shown and described above, it is understood that the above embodiments are exemplary and should not be construed as limiting the present application. Persons skilled in the art may make changes, modifications, substitutions, and variations to the above embodiments within the scope of the present application.

Claims

1. A zero-sample fault detection method for an intersection signal control system, characterized in that: include: Build an intersection signal control status perception system to obtain intersection signal control fault data to be detected; Using a neural network model to extract features from known fault data to obtain feature vectors, using a word vector model to map the vocabulary in the fault text to obtain semantic vectors, and constructing a semantic space based on the feature vectors and the semantic vectors; The semantic space is used to determine the fault type of the intersection signal control fault data to be detected, and the fault category and probability are output.

2. The zero-sample fault detection method for intersection signal control system according to claim 1, characterized in that: The method of extracting features from known fault data using a neural network model to obtain a feature vector, mapping vocabulary in the fault text using a word vector model to obtain a semantic vector, and constructing a semantic space based on the feature vector and the semantic vector includes: The known fault data is input into the neural network model for training, feature extraction and classification are performed, and the features are integrated and reduced in dimension through the neural network model to generate a complete feature vector; Use the word vector model to map the words in the fault text into vectors, and combine the vectors of each word in the text to generate a semantic vector; The feature vector and the semantic vector are normalized to construct a semantic space.

3. The zero-sample fault detection method for intersection signal control system according to claim 1, characterized in that: The determining of the fault type of the intersection signal control fault data to be detected by using the semantic space includes: Feature extraction is performed on the intersection signal control fault data to be detected, and the extracted feature vector is input into the semantic space. A distance value is obtained by a similarity distance calculation clustering method. If the distance value meets a first preset value, it is classified as a known fault; if the distance value meets a second preset value, it is classified as an unknown fault.

4. The zero-sample fault detection method for intersection signal control system according to claim 3, characterized in that: If the distance value meets the first preset value, the method classifies the fault as a known fault, and includes: inputting the intersection signal control fault data to be detected that is classified as a known fault into the neural network model for verification and comparison.

5. The zero-sample fault detection method for intersection signal control system according to claim 4, characterized in that: If the distance value meets the second preset value, the fault is classified as an unknown fault, which includes: An unknown fault classification space is constructed based on a multimodal data fusion method. The intersection signal control fault data to be detected that is classified as an unknown fault is input into the unknown fault space for distance clustering calculation to obtain the unknown fault classification result.

6. The zero-sample fault detection method for intersection signal control system according to claim 5, characterized in that: The method of using the semantic space to determine the fault type of the intersection signal control fault data to be detected and outputting the fault category and probability includes: Using a function to convert the score of the output layer of the trained neural network model into a probability distribution, and obtain the classification result and probability of the known fault; A function is used to convert the similarity into a probability output to obtain the classification result and probability of the unknown fault.

7. The zero-sample fault detection method for intersection signal control system according to claim 1, characterized in that: The construction of the intersection signal control state perception system and the acquisition of intersection signal control fault data to be detected include: constructing the intersection signal control state perception system based on the intersection signal control perception detector, data transmission equipment and software monitoring system, and using the intersection signal control state perception system to acquire the intersection signal control fault data to be detected.

8. A zero-sample fault detection device for an intersection signal control system, characterized in that: include: System construction module, builds the intersection signal control status perception system, and obtains the intersection signal control fault data to be detected; A semantic space construction module uses a neural network model to extract features from known fault data to obtain feature vectors, uses a word vector model to map the vocabulary in the fault text to obtain semantic vectors, and constructs a semantic space based on the feature vectors and the semantic vectors; The fault category classification module uses the semantic space to determine the fault type of the intersection signal control fault data to be detected, and outputs the fault category and probability.

9. An electronic device, characterized in that: include: a processor, and a memory communicatively connected to the processor; The memory stores computer-executable instructions; The processor executes the computer-executable instructions stored in the memory to implement the method according to any one of claims 1 to 7.

10. A computer-readable storage medium, characterized in that The computer-readable storage medium stores computer-executable instructions, which are used to implement the method according to any one of claims 1 to 7 when executed by a processor.

Images