Abnormality detection method and device

Abstract

The embodiment of the invention provides an anomaly detection method and device, and the method comprises the steps: responding to a storage event of a target object in a spectrum scanning region, calling a spectrum scanning device to scan the target object, and obtaining the spectrum scanning information of the target object; label identification information of the target article is obtained, and the label identification information is obtained based on content identification of an article label of the target article; and inputting the spectrum scanning information and the label identification information into an anomaly detection model to obtain an anomaly detection result of the target object. According to the method, firstly, detection automation is realized through response to a storage event, and efficiency is improved; secondly, fusing the spectral information and the label information, and constructing an objective data set; finally, the accuracy and efficiency of detection are further improved by means of intelligent judgment of the anomaly detection model.

Description

Technical Field

[0001] The embodiments in this specification relate to the field of anomaly detection technology, and in particular to an anomaly detection method and apparatus. Background Technology

[0002] The safe production, storage, and transportation of hazardous chemicals are of paramount importance in both industrial and public safety sectors. Because hazardous chemicals typically possess dangerous characteristics such as flammability, explosiveness, corrosiveness, toxicity, or radioactivity, leaks, misuse, or mixed storage due to management negligence can easily trigger major safety accidents, causing incalculable casualties, property damage, and environmental pollution. Therefore, accurate identification and status monitoring of hazardous chemicals are crucial prerequisites for achieving safe management throughout their entire lifecycle.

[0003] Currently, the detection of anomalies in hazardous chemicals mainly relies on manual judgment. For example, operators use their personal experience to determine whether the label of the hazardous chemical matches the actual product. However, this method is not only inefficient but also prone to misjudgment due to insufficient professional knowledge or negligence, resulting in low accuracy and efficiency in anomaly detection. Therefore, there is an urgent need for an anomaly detection method that can improve the accuracy and efficiency of hazardous chemical anomaly detection. Summary of the Invention

[0004] In view of this, embodiments of this specification provide an anomaly detection method and a method for detecting anomalies in hazardous chemicals. One or more embodiments of this specification also relate to an anomaly detection device and a hazardous chemical anomaly detection device, a computing device, a computer-readable storage medium, and a computer program product, to address the technical deficiencies existing in the prior art.

[0005] According to a first aspect of the embodiments of this specification, an anomaly detection method is provided, comprising: In response to an event that the target item is stored in the spectral scanning area, a spectral scanning device is invoked to scan the target item and obtain the spectral scanning information of the target item; Obtain the tag identification information of the target item, wherein the tag identification information is obtained based on the content recognition of the item tag of the target item; The spectral scanning information and the tag identification information are input into the anomaly detection model to obtain the anomaly detection result of the target item.

[0006] According to a second aspect of the embodiments of this specification, a method for detecting abnormalities in hazardous chemicals is provided, comprising: The scanning module is configured to, in response to an event that a hazardous chemical is stored in the spectral scanning region, invoke a spectral scanning device to scan the hazardous chemical and obtain the spectral scanning information of the hazardous chemical, wherein the hazardous chemical includes at least one of solid hazardous chemicals, liquid hazardous chemicals and gaseous hazardous chemicals; The acquisition module is configured to acquire the label identification information of the hazardous chemical, wherein the label identification information is obtained based on the content recognition of the hazardous chemical label of the hazardous chemical; The detection module is configured to input the spectral scanning information and the label identification information into the anomaly detection model to obtain the anomaly detection results of the hazardous chemical.

[0007] According to a third aspect of the embodiments of this specification, an anomaly detection device is provided, comprising: The scanning module is configured to, in response to an event that the target item is stored in the spectral scanning area, invoke a spectral scanning device to scan the target item and obtain the spectral scanning information of the target item. The acquisition module is configured to acquire the tag identification information of the target item, wherein the tag identification information is obtained based on the content recognition of the item tag of the target item; The detection module is configured to input the spectral scanning information and the tag identification information into the anomaly detection model to obtain the anomaly detection result of the target item.

[0008] According to a fourth aspect of the embodiments of this specification, a hazardous chemical anomaly detection device is provided, comprising: The scanning module is configured to, in response to an event that a hazardous chemical is stored in the spectral scanning region, invoke a spectral scanning device to scan the hazardous chemical and obtain the spectral scanning information of the hazardous chemical, wherein the hazardous chemical includes at least one of solid hazardous chemicals, liquid hazardous chemicals and gaseous hazardous chemicals; The acquisition module is configured to acquire the label identification information of the hazardous chemical, wherein the label identification information is obtained based on the content recognition of the hazardous chemical label of the hazardous chemical; The detection module is configured to input the spectral scanning information and the label identification information into the anomaly detection model to obtain the anomaly detection results of the hazardous chemical.

[0009] According to a fifth aspect of the embodiments of this specification, a computing device is provided, comprising: Memory and processor; The memory is used to store computer-executable instructions, and the processor is used to execute the computer-executable instructions, which, when executed by the processor, implement the steps of the above-described anomaly detection method.

[0010] According to a sixth aspect of the embodiments of this specification, a computer-readable storage medium is provided that stores computer-executable instructions, which, when executed by a processor, implement the steps of the above-described anomaly detection method.

[0011] According to a seventh aspect of the embodiments of this specification, a computer program product is provided, including a computer program / instructions that, when executed by a processor, implement the steps of the above-described anomaly detection method.

[0012] One embodiment of this specification implements an anomaly detection method, including: responding to a storage event of a target item in a spectral scanning area, calling a spectral scanning device to scan the target item and obtaining spectral scanning information of the target item; obtaining tag identification information of the target item, wherein the tag identification information is obtained based on content recognition of the item tag of the target item; inputting the spectral scanning information and the tag identification information into an anomaly detection model to obtain an anomaly detection result of the target item.

[0013] In one embodiment of this specification, the detection process is first automatically triggered by "responding to storage events," which greatly reduces the reliance on manual operation and improves detection efficiency. Secondly, by using a spectral scanning device, molecular spectral information (such as chemical composition) inherent in the item and not visible to the naked eye can be obtained. At the same time, label text is read through automatic recognition technology, constructing two types of objective datasets that are not affected by subjective factors. Finally, these two types of information are input into the anomaly detection model for intelligent comparison and analysis. By leveraging the model's inherent objectivity, ability to handle complex patterns, and intelligent judgment, the accuracy and efficiency of anomaly detection results are further improved. Attached Figure Description

[0014] Figure 1 This is a flowchart of an anomaly detection method provided in one embodiment of this specification; Figure 2 This is a flowchart of a method for detecting abnormalities in hazardous chemicals provided in one embodiment of this specification; Figure 3 This is a flowchart illustrating the processing procedure of an anomaly detection method provided in one embodiment of this specification. Figure 4 This is a schematic diagram of the structure of an anomaly detection device provided in one embodiment of this specification; Figure 5 This is a schematic diagram of the structure of a hazardous chemical anomaly detection device provided in one embodiment of this specification; Figure 6 This is a structural block diagram of a computing device provided in one embodiment of this specification. Detailed Implementation

[0015] Many specific details are set forth in the following description to provide a full understanding of this specification. However, this specification can be implemented in many other ways than those described herein, and those skilled in the art can make similar extensions without departing from the spirit of this specification. Therefore, this specification is not limited to the specific implementations disclosed below.

[0016] The terminology used in one or more embodiments of this specification is for the purpose of describing particular embodiments only and is not intended to be limiting of the one or more embodiments of this specification. The singular forms “a” and “the” as used in one or more embodiments of this specification and the appended claims are also intended to include the plural forms, unless the context clearly indicates otherwise. It should also be understood that the term “and / or” as used in one or more embodiments of this specification refers to and includes any or all possible combinations of one or more associated listed items.

[0017] It should be understood that although the terms first, second, etc., may be used to describe various information in one or more embodiments of this specification, such information should not be limited to these terms. These terms are only used to distinguish information of the same type from one another. For example, first may also be referred to as second without departing from the scope of one or more embodiments of this specification, and similarly, second may also be referred to as first. Depending on the context, the word "if" as used herein may be interpreted as "when," "when," or "in response to a determination."

[0018] Furthermore, it should be noted that the user information (including but not limited to user device information, user personal information, etc.) and data (including but not limited to data used for analysis, stored data, displayed data, etc.) involved in one or more embodiments of this specification are all information and data authorized by the user or fully authorized by all parties. Moreover, the collection, use and processing of related data must comply with the relevant laws, regulations and standards of the relevant countries and regions, and corresponding operation entry points are provided for users to choose to authorize or refuse.

[0019] First, the terms and concepts involved in one or more embodiments of this application will be explained.

[0020] JSON (JavaScript Object Notation) is a lightweight, text-based data-interchange format that uses a human-readable key-value pair structure to represent structured data.

[0021] Radio Frequency Identification (RFID) is an automatic identification technology that uses radio waves for non-contact, two-way communication. Its basic principle is to emit high-frequency radio frequency signals to activate and read the information stored in electronic tags, thereby achieving the identification of target objects and data exchange. Depending on the power supply method, electronic tags can be divided into three types: passive, active, and battery-assisted passive. RFID technology has advantages such as no line-of-sight alignment required, batch reading capability, strong environmental adaptability, and long service life, and is widely used in logistics, warehousing, transportation, security, retail, and smart lockers.

[0022] Convolutional Neural Network (CNN): A deep learning model inspired by biological visual systems, specifically designed to process data with a grid-like topology, such as images, videos (image sequences), and speech signals.

[0023] OCR (Optical Character Recognition) is a technology that can convert text in an image (whether printed or handwritten) into editable and searchable text data.

[0024] This specification provides an anomaly detection method and a hazardous chemical anomaly detection method. It also relates to an anomaly detection device and a hazardous chemical anomaly detection device, a computing device, a computer-readable storage medium, and a computer program product, which will be described in detail in the following embodiments.

[0025] See Figure 1 , Figure 1 A flowchart of an anomaly detection method according to an embodiment of this specification is shown, specifically including the following steps 102-106.

[0026] Step 102: In response to the event of the target item being stored in the spectral scanning area, call the spectral scanning device to scan the target item and obtain the spectral scanning information of the target item.

[0027] The target item refers to the item to be subjected to anomaly detection. For example, the hazardous chemicals in a hazardous chemical storage cabinet.

[0028] The spectral scanning area is a specially designed physical zone equipped with sensors or triggers that automatically initiate the scanning process when an object is placed in this area. This area is typically calibrated to ensure the accuracy and consistency of the scan.

[0029] A spectral scanning device is an instrument used to collect spectral data of an object, such as a near-infrared spectrometer, a Raman spectrometer, or an infrared imaging device. It obtains the spectral characteristics of an object by emitting light of a specific wavelength and analyzing the reflected or transmitted light.

[0030] Spectral scanning information refers to the property information of an item determined by feature identification after scanning spectral data. It includes property information directly obtained from spectral analysis (such as item type: "electronic product" / "food"), material composition (such as "plastic" / "metal"), and may also include information that needs to be inferred and verified (such as hazard level).

[0031] For example, when the target item is a hazardous chemical, spectral scanning information refers to the structured hazardous chemical attribute information generated by processing the spectral scan of the hazardous chemical using a chemical feature recognition algorithm. It includes hazardous chemical attribute information directly obtained from spectral analysis (such as chemical identification, main components, and concentration), and may also include information that requires reasoning and verification (such as hazard classification and safety recommendations).

[0032] Example: Event Trigger: An operator places a bottle of chemical requiring verification (the target item) in a specially designed spectral scanning area (e.g., a sample stage). The weight sensor in the spectral scanning area detects the item being placed, triggering a system event.

[0033] Scan Invocation: The system (such as a software platform or control unit) automatically activates a spectral scanning device (e.g., a Raman spectrometer). This device emits a laser towards the sample and collects its unique "spectral fingerprint."

[0034] Information processing and generation: The built-in chemical feature recognition algorithm compares and analyzes the raw spectral data with a known hazardous chemical spectral library.

[0035] Output: The system generates the final spectral scan information—this is not a spectral curve, but rather the directly interpreted, semantic hazardous material attribute information.

[0036] In one possible implementation of this embodiment, a spectral scanning device is invoked to scan the target item and obtain its spectral scanning information. This includes: the system sending a "scanning and recognition command" to the spectral scanning device; the spectral scanning device completing the scan locally, generating a spectral feature map, and immediately calling its built-in algorithms and database for analysis and recognition; and the device directly returning the final generated, structured spectral scanning information to the system. The system no longer needs to perform complex analysis and can directly use the result.

[0037] In another possible implementation of this embodiment, a spectral scanning device is invoked to scan the target item to obtain the spectral scanning information of the target item, including: Send a spectral scanning command for the target item to the spectral scanning device, and receive the spectral feature map of the target item fed back by the spectral scanning device in response to the spectral scanning command; Analyze the spectral feature map to obtain the spectral scanning information of the target item.

[0038] The spectral scan command is a control command sent by the system (such as a computer or controller) to the spectral scanning device to initiate the scanning process of the target item. The command may include scanning parameters (such as scan range, resolution, or duration).

[0039] Spectral feature map: The raw data results returned by the spectral scanning device after responding to the command, usually represented in the form of a graph or digital matrix, showing the spectral characteristics of the target item at different wavelengths (such as absorption peaks, reflectance, or emission spectrum). These characteristics are related to the chemical composition or physical state of the item.

[0040] Specifically, the system sends a scanning command for the target item to the spectral scanning device. The command triggers the device to perform a non-contact or contact scan of the item.

[0041] The spectral scanning device performs a scan, captures the spectral data of the target item, and generates a spectral feature map as a response. This feature map is returned to the system via a communication interface (such as a network or Bluetooth).

[0042] After receiving the spectral feature map, the system calls the parsing module (such as a feature recognition algorithm or analysis software) to process the feature map. The parsing process may include noise filtering, peak detection, spectral library matching, or model prediction.

[0043] The analysis results are converted into structured spectral scan information, including direct and inferred information. For example, for hazardous chemicals, the analysis may identify the type and chemical composition of the hazardous chemical and infer its hazard based on a rule base.

[0044] Example: The process of obtaining spectral scanning information is explained using the example of a hazardous chemical (such as a bottle of unknown liquid).

[0045] The system sends a spectral scanning command to the spectral scanning device, specifying the scanning mode as "infrared spectral scanning" and the scanning time as 5 seconds.

[0046] The spectral scanning device scans the unknown liquid, obtains infrared spectral data, and generates a spectral feature map. The device then returns this feature map to the system.

[0047] The system analysis module processes the spectral feature map: it uses chemical feature recognition algorithms (such as principal component analysis or neural networks) to compare it with a standard spectral database and identifies that the feature matches ethanol (alcohol).

[0048] After parsing, the system generates spectral scan information, including: Direct information: chemical label (“ethanol”), main component (“C2H5OH”), concentration (“95%”).

[0049] Inference information: Based on the chemical knowledge base, the hazard classification ("flammable liquid") and safety advice ("keep away from fire and store in a cool, ventilated place") are inferred.

[0050] Finally, the spectral scan information is output in a structured format (such as JSON) for subsequent use (such as safety assessment or label generation). The entire process is efficient and reliable, and suitable for hazardous chemical management scenarios.

[0051] In the embodiments described in this specification, the core analytical algorithm and a large feature database are deployed on a central server with stronger computing power. This allows the use of more complex and accurate machine learning models or professional analysis software, avoiding recognition errors that may occur due to limited computing resources at the scanning device end. The spectral scanning device itself does not need to integrate a powerful computing chip and a large database; it only needs to focus on high-precision data acquisition and transmission. This significantly reduces the hardware cost, power consumption, and size of individual terminal devices, which is conducive to the large-scale deployment of this technology. Furthermore, when it is necessary to support new item types or improve recognition capabilities, only the algorithm model needs to be updated or the database expanded on the central server. There is no need to upgrade the firmware of all spectral scanning devices distributed in the field one by one, which greatly reduces maintenance costs and workload, enabling the system to quickly adapt to new needs.

[0052] In one possible implementation of this embodiment, the spectral feature map is analyzed to obtain the spectral scanning information of the target item, including: identifying and quantifying key feature parameters, such as the position (wavenumber / wavelength), intensity, full width at half maximum (FWHM), and peak shape, from the original spectral feature map using a signal processing algorithm; comparing the extracted feature parameter set with a standard substance spectral database, wherein each record in the database contains standard feature peak information of a substance; calculating the similarity (e.g., cosine similarity, Euclidean distance, etc.) between the spectrum to be measured and each record in the database, and returning one or more candidate substances and their matching scores; based on the matching results (e.g., top 1 candidate substances and their confidence scores), combined with an independent attribute knowledge base (e.g., a hazardous materials database, a materials database), and through a preset "if-then" rule (e.g., IF the matching substance is "acetone" THEN its classification is "flammable liquid, irritating"), reasoning and assembling complete "spectral scanning information".

[0053] In another possible implementation of this embodiment, analyzing the spectral feature map to obtain the spectral scanning information of the target item includes: By inputting the spectral feature map into the spectral analysis model, the spectral scanning information of the target item can be obtained.

[0054] Among them, the spectral analysis model is usually an algorithmic model trained based on machine learning (especially deep learning, such as convolutional neural networks). Its essence is a complex, parameterized mathematical function that encapsulates the "knowledge" and "experience" learned from massive amounts of spectral data.

[0055] Specifically, the spectral feature map (which may be preprocessed into a standard format) is input into the spectral analysis model that has already been loaded into memory.

[0056] The multi-layered neural network inside the model begins to work, extracting features layer by layer (from simple edges and peaks to complex spectral shape patterns). Finally, the model's output layer generates a structured JSON object, which is the spectral scan information.

[0057] The spectral scan information directly output by the model may look like the following: json { "Recognition Result": { Item Name: Acetone Confidence level: 0.987 }, "Classification": { Flammable liquids, Irritating Hazard Statement: "Highly flammable liquid and vapor, causes severe eye irritation." }, "Physicochemical property prediction": { Physical state: Liquid "Main functional groups": ["Carbonyl (C=O)", "Methyl (CH3)"] }, Safety Handling Recommendations: Store in a cool, well-ventilated place, away from fire. Use explosion-proof equipment. In the embodiments described in this specification, an end-to-end one-time conversion from the original spectral image to semantic information (spectral scan information) is achieved. This eliminates the need for multiple steps such as feature extraction, database matching, and attribute library queries, simplifying the processing flow and improving the efficiency of anomaly detection.

[0058] Before deploying the spectral analysis model, it needs to be trained using a large, labeled dataset of "spectral feature maps - item attributes". For example, it can be trained with tens of thousands of known spectra of "acetone", "ethanol", and "sulfuric acid" and their corresponding attributes, so that it learns the mapping relationship from spectral features to item attributes.

[0059] In one possible implementation of this embodiment, before inputting the spectral feature map into the spectral analysis model to obtain the spectral scanning information of the target item, the following steps are also included: Obtain a sample set, which includes multiple sample spectral feature maps and sample spectral scan information corresponding to the multiple sample spectral feature maps; The initial analytical model is invoked to analyze the spectral feature maps of multiple samples, thereby obtaining the predicted spectral scanning information corresponding to the spectral feature maps of multiple samples respectively; Based on the sample spectral scanning information and the predicted spectral scanning information, the initial analytical model is trained to obtain the spectral analytical model.

[0060] The sample set is a labeled dataset used to train the model.

[0061] The sample set consists of many "sample pairs," each of which includes: Sample spectral feature map: Raw spectral data of a known item.

[0062] Sample spectral scanning information: The manually labeled, structured attribute information corresponding to this item.

[0063] Example location: Sample spectral scan information: json { "Chemical identifier": {"Chemical name": "Acetone"}, "Classification": {"Flammable Liquids", "Irritating"}, "Physical state": "Liquid"} The initial parsing model is a machine learning model that has not been trained or has only been initialized (such as a neural network with random initial weights).

[0064] The predicted spectral scan information is the output of the initial analytical model after parsing the spectral feature maps of the input samples during training. In the early stages of training, this prediction result is usually completely wrong or extremely inaccurate because the model has not yet learned the correct mapping relationship.

[0065] The spectral analysis model is the final, usable version obtained after training the initial analysis model. At this point, the model's internal parameters have been optimized to accurately map the input spectral feature map to the correct spectral scan information.

[0066] Specifically, the model calculates the output (predicted spectral scan information) from the input (spectral feature map) based on the current parameters. The model's predicted spectral scan information is compared with the actual sample spectral scan information, and the difference between the two (i.e., the "error amount") is calculated using a "loss function". Based on the calculated "error amount", millions or even billions of parameters within the model are adjusted in reverse, so that when the model encounters the same or similar "spectral feature map" again, it can provide an output that more closely approximates the sample spectral scan information.

[0067] Example: Train a Raman spectral analysis model that can identify common hazardous chemicals.

[0068] Obtaining the sample set: Researchers used Raman spectrometry to scan thousands of pure samples of known hazardous chemicals, such as acetone, ethanol, sulfuric acid, and benzene.

[0069] For each sample scanned, two data points are saved as a sample pair: Data A (Sample Spectral Characteristic Map): The raw spectral data file of this chemical.

[0070] Data B (Sample Spectral Scan Information): A JSON file containing the standard properties of the chemical, manually compiled by experts based on the safety data sheet.

[0071] Collecting thousands of such "sample pairs" forms a sample set for training.

[0072] Choose a neural network architecture (such as a convolutional neural network) and create an initial analytical model. At this point, its parameters are randomized.

[0073] Model training: Loop begins: Take a batch (e.g., 32) of sample spectral feature maps from the sample set and input them into the initial model.

[0074] The model outputs 32 predicted spectral scans. During the first loop, these predictions may be completely garbled, for example, predicting "acetone" as "water".

[0075] Loss calculation: The system compares these 32 flawed prediction results with the spectral scan information of 32 real samples (standard answers) and calculates a huge "total error" (loss value).

[0076] Optimization: Using the backpropagation algorithm, fine-tune all the parameters of the model based on the "total error" so that it can perform slightly better next time.

[0077] End of loop: Repeat the above process tens of thousands or even millions of times, traversing the entire sample set multiple times.

[0078] After multiple rounds of training, the loss value has dropped to a very low level. This means that the model can now see a spectrum of acetone and output its correct property information very accurately.

[0079] At this point, the well-trained and high-performing model is saved as the final spectral analysis model.

[0080] In the embodiments described in this specification, through learning from massive amounts of samples, the model can become "immune" to various interferences (such as noise and baseline drift), and its recognition accuracy and stability can surpass traditional methods based on simple rules. Furthermore, the model can be iteratively optimized by injecting new sample data, enabling it to continuously adapt to new chemicals and new recognition needs.

[0081] Step 104: Obtain the tag identification information of the target item, wherein the tag identification information is obtained based on the content recognition of the item tag of the target item.

[0082] Among them, item tags are labels attached to target items to identify them. They can be physical tags (such as stickers or tags) or electronic tags (such as RFID tags).

[0083] Tag identification information refers to the structured data obtained by identifying the physical carrier of an item's tag. This information is not the original image or signal of the tag, but rather the semantic result after parsing. For example, for a barcode, it is the information obtained after decoding; for a text tag, it is the key fields (such as item name and production date) extracted according to predefined rules after OCR recognition.

[0084] This step is independent of the preceding spectral scanning step and provides item information from an alternative source (i.e., information from the tag). This step can be performed alone, or in parallel or sequentially with other steps (such as spectral scanning).

[0085] In one possible implementation of this embodiment, item information is obtained directly and quickly by scanning machine-readable codes (such as barcodes, QR codes, or RFID) on the item label.

[0086] Specifically, specialized scanning equipment (such as laser scanners or RFID readers) is used to read the codes on the tags. These codes themselves do not contain detailed item attributes, but rather a unique "ID" (identification number). The system sends this unique ID to a local or cloud-based item information database for querying. The database finds the pre-entered, complete item information associated with that ID and returns it to the system, forming structured tag identification information.

[0087] In another possible implementation of this embodiment, obtaining the tag identification information of the target item includes: Use an image acquisition device to capture images of the item's label. Text recognition is performed based on the image of the item label to obtain the label recognition information of the target item.

[0088] Image acquisition equipment is hardware used to capture visual information from item tags. It is not limited to ordinary cameras but can take various forms depending on the application scenario. For example, handheld smart terminals: used by warehouse managers for flexible inventory checks; high-definition surveillance cameras: installed in storage cabinets or at entrances for remote monitoring and identification.

[0089] Item label images are digital image files generated by image acquisition devices, containing the complete visual content of the label. Image enhancement (e.g., increasing contrast) can be applied to the item label image before text recognition.

[0090] Specifically, the process involves locating all regions in the image where text appears. Image pixels within each text region are converted into corresponding text characters. The identified text is then analyzed, and key fields such as "item name" and "production date" are extracted based on a pre-defined template or natural language processing technology, and assigned semantic meaning.

[0091] Example: An image acquisition device (a camera) fixed above a hazardous materials storage cabinet captures a clear, distortion-free image of the item label and uploads it to the server. The OCR recognition service on the server receives this label image, identifies all the text blocks in the image, and converts them into the original text. This might result in the following garbled image: text Product Name: Acetone Hazard statement: Highly flammable liquids and vapors... Production batch number: 20240521A The system extracts key fields from the above text based on the preset "hazardous chemical label parsing template".

[0092] Finally, the system generates and outputs structured label recognition information: json { "Source Identification": "Image Recognition" Basic Information: { Product Name: Acetone Specifications: Analytical Reagent }, "Production Information": { Production batch number: 20240521A }... Production Date: 2024-05-21 }, "Hazard Information": { "Classification": ["Flammable Liquids", "Irritating"], Hazard Statement: ["Highly flammable liquid and vapor", "Causes severe eye irritation"] }, "Identification confidence level": 95.5 In the embodiments described in this specification, the entire process requires no human contact or touch of the object, making it particularly suitable for operations in toxic, harmful, or sterile environments. This achieves fully automated information collection, significantly improving work efficiency and safety. Furthermore, modern text recognition technology boasts an extremely high recognition rate for printed text (>99%). Combined with specialized image preprocessing (such as skew correction and noise reduction), it effectively overcomes challenges in real-world scenarios such as label wrinkles, uneven lighting, and partial soiling, ensuring the accuracy of the results.

[0093] Step 106: Input the spectral scanning information and tag identification information into the anomaly detection model to obtain the anomaly detection results of the target item.

[0094] This step inputs two independent information sources (spectral scanning information and label recognition information) into a pre-trained anomaly detection model. The model analyzes whether the two types of information are consistent and whether there is an abnormal pattern, and finally outputs an anomaly detection result.

[0095] Spectral scanning information: The item's attribute information (such as chemical name, classification, etc.) is obtained by scanning the target item with a spectral scanning device and then analyzing it.

[0096] Label recognition information: Item attribute information (such as product name, production batch number, etc.) obtained by recognizing the label attached to the target item (such as OCR text recognition or code scanning).

[0097] Anomaly detection model: A model capable of identifying anomalies based on input multimodal information (in this case, spectral scanning information and label recognition information). Anomalies may include, but are not limited to: inconsistent information (e.g., spectral recognition identifies it as acetone, but the label shows it as ethanol), missing label information, and abnormal spectral information (e.g., not conforming to the characteristics of known substances).

[0098] Anomaly detection results: The model output is usually structured data, which can be a yes / no anomaly, or more detailed results such as anomaly type, anomaly level, confidence level, etc.

[0099] In one possible implementation of this embodiment, the structured spectral scanning information and tag identification information are converted into a unified vector representation; the comprehensive similarity score of the two types of information is calculated in the embedding space; and based on the similarity score and their respective confidence levels, a final anomaly judgment is obtained through a fusion algorithm.

[0100] In another possible implementation of this embodiment, the spectral scanning information corresponds to at least one feature dimension; By inputting spectral scanning information and tag identification information into the anomaly detection model, the anomaly detection results for the target item are obtained, including: The spectral scanning information and label recognition information are input into the anomaly detection model to obtain at least one sub-detection result, wherein the sub-detection result corresponds one-to-one with the feature dimension; Based on at least one sub-detection result, determine the abnormal detection result of the target item.

[0101] Among them, feature dimension refers to a certain attribute of an item, such as chemical name, classification, physical state, etc. Each feature dimension is a perspective for anomaly detection.

[0102] Sub-detection results: The results of anomaly detection for each feature dimension, which typically include consistency or anomalies in that dimension.

[0103] Anomaly detection results: An overall anomaly assessment of the target item is made by combining the sub-detection results of all feature dimensions.

[0104] Specifically, multiple feature dimensions corresponding to the spectral scanning information and label recognition information are determined. For example, these may include chemical names, classifications, etc. For each feature dimension, the values ​​of that dimension in the spectral scanning information and label recognition information are compared to obtain sub-detection results (e.g., consistent, inconsistent, partially consistent, etc.). Based on the sub-detection results of all feature dimensions, the final anomaly detection result is determined through certain rules (e.g., voting, weighted scoring, etc.) or models (e.g., classifiers).

[0105] Example: Suppose there are two feature dimensions: chemical name and hazard classification.

[0106] The input spectral scan information is as follows: { "Chemical Name": "Acetone", "Classification": "Flammable Liquid"} The tag identification information is as follows: { "Chemical Name": "Ethanol", "Classification": "Flammable Liquid"} Step 1: We set two feature dimensions: chemical name and hazard classification.

[0107] Step 2: Perform sub-detections separately: Chemical name dimension: Spectral scanning information is "acetone", label identification information is "ethanol", inconsistency -> Sub-detection result: abnormal.

[0108] Hazard classification dimension: Both are "flammable liquids", consistent -> Sub-test result: normal.

[0109] Step 3: Based on the sub-detection results and the established rules: if any key dimension (such as chemical name) is abnormal, the entire system is considered abnormal. Therefore, the final anomaly detection result is abnormal.

[0110] In the embodiments described in this specification, multi-dimensional detection allows for precise identification of which attributes exhibit abnormalities, rather than simply determining whether the overall system is abnormal. Furthermore, since each dimension yields a separate result, when an anomaly occurs, it can be clearly identified which attribute is inconsistent, facilitating subsequent processing. Additionally, different weights can be assigned based on the importance of different feature dimensions, leading to a more reasonable assessment of overall anomalies. For example, inconsistencies in chemical names are considered more serious than inconsistencies in physical states.

[0111] In one optional implementation of this embodiment, at least one feature dimension includes item type, and the spectral scanning information includes the confidence score of the target item in at least one item type; the spectral scanning information and tag recognition information are input into the anomaly detection model to obtain at least one sub-detection result, including: An anomaly detection model is used to extract the category of the target item from the label recognition information; The confidence score of the spectral scanning information under the identified item type is determined as the category matching degree, and the sub-detection result is determined based on the category matching degree.

[0112] Among them, the category of the item is identified as the category to which the item, as claimed on the label, belongs, extracted from the label identification information.

[0113] Category matching degree: Here, the confidence score of the spectral scan information under the identified item category is used directly as the matching degree.

[0114] Specifically, the identifying item type of the target item (i.e., the type claimed on the label) is extracted from the label identification information.

[0115] In the spectral scan information, view the confidence score of the target item under this identified item category (this score indicates the degree to which the spectral scan results support the item belonging to this category).

[0116] This confidence score is used as the species matching score, and then the anomaly is determined based on this matching score (for example, if the matching score is very low, it means that the spectral scan results do not support the species claimed on the label, i.e., there is an anomaly).

[0117] Example: Suppose an item's tag identification information identifies it as "acetone," while the spectral scan information contains confidence scores for multiple categories (such as acetone, ethanol, water, etc.). The confidence score for "acetone" is 0.05 (5%). Therefore, the category match is 0.05. Because this match is very low, we can determine that there is an anomaly in the item category dimension.

[0118] In the embodiments described in this specification, since type is usually one of the most critical characteristics of an item, this detection is able to detect most cases of mislabeling.

[0119] In another optional implementation of this embodiment, a detection strategy based on predefined rules is adopted. By constructing a rule engine containing multiple business rules, logical judgments are made on spectral scanning information and tag recognition information to obtain anomaly detection results.

[0120] Specifically, a set of detection rules is predefined to cover various abnormal scenarios; key features required for rule judgment are extracted from two types of information; the preconditions of each rule are evaluated one by one; and the final anomaly detection result is inferred based on the triggered rules and priorities.

[0121] In one optional implementation of this embodiment, after inputting spectral scanning information and tag identification information into the anomaly detection model to obtain the anomaly detection result of the target item, the method further includes: if the anomaly detection result indicates that the target item is abnormal, obtaining the current environmental information and environmental tag information of the target item; and determining the cause of the anomaly of the target item based on the current environmental information and environmental tag information.

[0122] Current environmental information: refers to the real-time environmental parameters of the target item, such as temperature, humidity, light, and location.

[0123] Environmental label information: refers to the pre-defined environmental requirements that the item should meet under normal conditions, such as storage temperature range, humidity range, and whether it should be protected from light.

[0124] Specifically, when the anomaly detection model outputs an anomaly result, the system further collects the current environmental information of the target item (through sensors and other devices) and preset environmental label information.

[0125] The current environmental information is compared with the environmental label information to analyze whether there are any anomalies caused by environmental factors.

[0126] Based on the comparison results, determine the cause of the anomaly, such as unsuitable environmental conditions leading to product deterioration or incorrect label information.

[0127] Example: Suppose in a hazardous chemical storage scenario, the anomaly detection model detects a discrepancy between the actual spectral scan information (identified as methanol) and the label information (identified as ethanol) of a certain chemical container, classifying it as an anomaly. Subsequently, the system obtains the current environmental information of the chemical container: temperature 35℃, humidity 80%; and the environmental label information: required storage temperature 0-25℃, humidity <70%. By comparison, it is found that the current temperature of 35℃ exceeds the maximum storage temperature of 25℃, and the humidity of 80% exceeds the maximum humidity of 70%. Therefore, the system determines that the anomaly may be due to improper storage conditions leading to chemical deterioration or damage to the label information caused by environmental factors.

[0128] The embodiments in this specification can identify abnormal situations caused by environmental conditions, providing a basis for environmental improvement and prevention of similar abnormalities.

[0129] In another optional implementation of this embodiment, after inputting spectral scanning information and tag identification information into the anomaly detection model to obtain the anomaly detection result of the target item, the method further includes: issuing an anomaly alarm when the anomaly detection result indicates that the target item is abnormal.

[0130] Among them, abnormal alarms refer to the system issuing warnings to relevant personnel through sound, light, electricity, message notifications, etc., to indicate that abnormal situations have occurred.

[0131] Specifically, when the anomaly detection model outputs an abnormal result, the system automatically triggers an alarm mechanism.

[0132] The alarm mechanism sends alarm information to relevant personnel through various channels (such as the monitoring center screen, SMS, email, etc.) based on the preset alarm strategy.

[0133] Alarm information typically includes the identifier of the abnormal item, the type of abnormality, the level of abnormality, and the time of occurrence, in order to enable a rapid response.

[0134] Example: Again, using a hazardous chemical storage scenario, the system detects an anomaly (inconsistent identification) in a container of chemicals through spectral scanning and label identification, immediately triggering an anomaly alarm. The alarm information is sent in the following way: A red alert was displayed on the large screen in the warehouse monitoring center, and an alarm sounded.

[0135] Send a text message to the warehouse manager: "Warning: An anomaly has been found in the chemicals in area A3. The item was labeled as ethanol, but it has been tested as methanol. Please handle it immediately!" At the same time, detailed alarm information is pushed to the enterprise's security management applications, including item images, detection time, and suggested actions.

[0136] Enable real-time notification of abnormal situations to ensure that relevant personnel are informed and can take action immediately.

[0137] In the embodiments described in this specification, alarms help to facilitate rapid response and handling, prevent the anomaly from escalating, and reduce losses.

[0138] It should be noted that the above-mentioned anomaly analysis scheme and anomaly alarm scheme can coexist.

[0139] In order to achieve the detection of anomalies in hazardous chemicals, the embodiments of this specification also provide a method for detecting anomalies in hazardous chemicals.

[0140] See Figure 2 , Figure 2 A flowchart of a method for detecting abnormalities in hazardous chemicals according to an embodiment of this specification is shown, specifically including the following steps 202-206.

[0141] Step 202: In response to the event of the storage of hazardous chemicals in the spectral scanning area, the spectral scanning equipment is invoked to scan the hazardous chemicals and obtain the spectral scanning information of the hazardous chemicals, wherein the hazardous chemicals include at least one of solid hazardous chemicals, liquid hazardous chemicals and gaseous hazardous chemicals; Step 204: Obtain the label identification information of hazardous chemicals, wherein the label identification information is obtained based on the content recognition of the hazardous chemical label; Step 206: Input the spectral scanning information and label identification information into the anomaly detection model to obtain the anomaly detection results for hazardous chemicals.

[0142] The spectral scanning area refers to the scanning area inside the hazardous chemical storage cabinet.

[0143] It should be noted that the technical solution of this hazardous chemical anomaly detection method is based on the same concept as the technical solution of the aforementioned anomaly detection method. For details not described in detail in the technical solution of the hazardous chemical anomaly detection method, please refer to the description of the technical solution of the aforementioned anomaly detection method.

[0144] The following is in conjunction with the appendix Figure 3 Taking the application of the anomaly detection method provided in this specification in the anomaly detection of hazardous chemicals as an example, the anomaly detection method will be further explained. Among other things, Figure 3The present specification shows a flowchart of an anomaly detection method according to an embodiment, which includes the following steps.

[0145] Step 302: In response to the event of hazardous chemicals being stored in the spectral scanning area, invoke the spectral scanning equipment to scan the hazardous chemicals and obtain the spectral scanning information of the hazardous chemicals.

[0146] Application scenario: In a smart warehouse of a chemical plant, an operator places a barrel of chemical labeled "acetone" into the spectral scanning area of ​​an explosion-proof storage cabinet.

[0147] Specifically, the infrared position sensor confirms that the object is accurately placed on the scanning platform, the system automatically starts the integrated Raman spectrometer, performs non-contact spectral scanning, generates a standardized spectral feature map, and analyzes the structured spectral scanning information based on the spectral feature map.

[0148] Step 304: Obtain the label identification information of hazardous chemicals, wherein the label identification information is obtained based on the content recognition of the hazardous chemical label.

[0149] Specifically, the system uses a built-in industrial camera (5 megapixels) in the storage cabinet to automatically focus and photograph the labels on the hazardous chemical containers, capturing high-definition label images; then, an OCR engine is used to recognize the text content and generate structured label recognition information.

[0150] Step 306: Input the spectral scanning information and label identification information into the anomaly detection model to obtain the anomaly detection results for hazardous chemicals.

[0151] Specifically, a pre-trained hazardous chemical anomaly detection model is loaded, spectral scanning information and label recognition information are input, multi-dimensional consistency analysis is performed, and structured anomaly detection results are output.

[0152] Step 308: If the anomaly detection results indicate that the target item is abnormal, obtain the current environmental information and environmental label information of the target item.

[0153] Specifically, the system calls upon the storage cabinet's environmental sensor network to collect current environmental parameters and extracts environmental requirements from the tag identification information.

[0154] Step 310: Determine the cause of the anomaly of the target item based on the current environmental information and environmental label information.

[0155] Step 312: Issue an anomaly alert.

[0156] Corresponding to the above method embodiments, this specification also provides embodiments of anomaly detection devices. Figure 4 A schematic diagram of an anomaly detection device according to one embodiment of this specification is shown. Figure 4 As shown, the device includes: The scanning module 402 is configured to respond to a storage event of the target item in the spectral scanning area, call the spectral scanning device to scan the target item, and obtain the spectral scanning information of the target item; The acquisition module 404 is configured to acquire the tag identification information of the target item, wherein the tag identification information is obtained based on the content recognition of the item tag of the target item; The detection module 406 is configured to input spectral scanning information and tag identification information into the anomaly detection model to obtain the anomaly detection result of the target item.

[0157] Optionally, the scanning module 402 is further configured to: Send a spectral scanning command for the target item to the spectral scanning device, and receive the spectral feature map of the target item fed back by the spectral scanning device in response to the spectral scanning command; Analyze the spectral feature map to obtain the spectral scanning information of the target item.

[0158] Optionally, the scanning module 402 is further configured to: By inputting the spectral feature map into the spectral analysis model, the spectral scanning information of the target item can be obtained.

[0159] Optionally, the above-mentioned anomaly detection device further includes a training module; the training module is configured to: Obtain a sample set, which includes multiple sample spectral feature maps and sample spectral scan information corresponding to the multiple sample spectral feature maps; The initial analytical model is invoked to analyze the spectral feature maps of multiple samples, thereby obtaining the predicted spectral scanning information corresponding to the spectral feature maps of multiple samples respectively; Based on the sample spectral scanning information and the predicted spectral scanning information, the initial analytical model is trained to obtain the spectral analytical model.

[0160] Optionally, the spectral scanning information corresponds to at least one feature dimension; the detection module 406 is further configured to: The spectral scanning information and label recognition information are input into the anomaly detection model to obtain at least one sub-detection result, wherein the sub-detection result corresponds one-to-one with the feature dimension; Based on at least one sub-detection result, determine the abnormal detection result of the target item.

[0161] Optionally, at least one feature dimension includes item type, and the spectral scan information includes the confidence score of the target item in each item type; the detection module 406 is further configured to: An anomaly detection model is used to extract the category of the target item from the label recognition information; The confidence score of the spectral scanning information under the identified item type is determined as the category matching degree, and the sub-detection result is determined based on the category matching degree.

[0162] Optionally, module 404 is further configured as follows: Use an image acquisition device to capture images of the item's label. Text recognition is performed based on the image of the item label to obtain the label recognition information of the target item.

[0163] Optionally, the above-mentioned anomaly detection device further includes a post-processing module; the first post-processing module is configured to: When the anomaly detection results indicate that the target item is abnormal, obtain the target item's current environmental information and environmental label information; Based on the current environmental information and environmental label information, determine the cause of the anomaly of the target item.

[0164] Optionally, the above-mentioned anomaly detection device further includes a post-processing module; the second post-processing module is configured to: When the anomaly detection results indicate that the target item is abnormal, an anomaly alarm is issued.

[0165] The above is a schematic scheme of an anomaly detection device according to this embodiment. It should be noted that the technical solution of this anomaly detection device and the technical solution of the above-described anomaly detection method belong to the same concept. For details not described in detail in the technical solution of the anomaly detection device, please refer to the description of the technical solution of the above-described anomaly detection method.

[0166] Corresponding to the above method embodiments, this specification also provides embodiments of hazardous chemical anomaly detection devices. Figure 5 A schematic diagram of a hazardous chemical anomaly detection device according to one embodiment of this specification is shown. Figure 5 As shown, the device includes: The scanning module 502 is configured to respond to an event in which hazardous chemicals are stored in the spectral scanning area, call a spectral scanning device to scan the hazardous chemicals, and obtain spectral scanning information of the hazardous chemicals, wherein the hazardous chemicals include at least one of solid hazardous chemicals, liquid hazardous chemicals and gaseous hazardous chemicals; The acquisition module 504 is configured to acquire the label identification information of hazardous chemicals, wherein the label identification information is obtained based on the content recognition of the hazardous chemical label. The detection module 506 is configured to input spectral scanning information and label identification information into the anomaly detection model to obtain anomaly detection results for hazardous chemicals.

[0167] The above is a schematic scheme of a hazardous chemical anomaly detection device according to this embodiment. It should be noted that the technical solution of this hazardous chemical anomaly detection device and the technical solution of the above-described anomaly detection method belong to the same concept. For details not described in detail in the technical solution of the hazardous chemical anomaly detection device, please refer to the description of the technical solution of the above-described anomaly detection method.

[0168] Figure 6 A structural block diagram of a computing device 600 according to one embodiment of this specification is shown. The components of the computing device 600 include, but are not limited to, a memory 610 and a processor 620. The processor 620 is connected to the memory 610 via a bus 630, and a database 650 is used to store data.

[0169] The computing device 600 also includes an access device 640, which enables the computing device 600 to communicate via one or more networks 660. Examples of these networks include Public Switched Telephone Network (PSTN), Local Area Network (LAN), Wide Area Network (WAN), Personal Area Network (PAN), or combinations of communication networks such as the Internet. The access device 640 may include one or more of any type of wired or wireless network interface (e.g., a network interface card (NIC)), such as an IEEE 802.11 Wireless Local Area Network (WLAN) wireless interface, a Wi-MAX (Worldwide Interoperability for Microwave Access) interface, an Ethernet interface, a Universal Serial Bus (USB) interface, a cellular network interface, a Bluetooth interface, or a Near Field Communication (NFC) interface.

[0170] In one embodiment of this specification, the above-described components of the computing device 600 and Figure 6 Other components, not shown, can also be connected to each other, for example, via a bus. It should be understood that... Figure 6 The block diagram of the computing device shown is for illustrative purposes only and is not intended to limit the scope of this specification. Those skilled in the art can add or replace other components as needed.

[0171] The computing device 600 can be any type of stationary or mobile computing device, including mobile computers or mobile computing devices (e.g., tablet computers, personal digital assistants, laptop computers, notebook computers, netbooks, etc.), mobile phones (e.g., smartphones), wearable computing devices (e.g., smartwatches, smart glasses, etc.) or other types of mobile devices, or stationary computing devices such as desktop computers or personal computers (PCs). The computing device 600 can also be a mobile or stationary server.

[0172] The processor 620 is used to execute the following computer-executable instructions, which, when executed by the processor, implement the steps of the above-described anomaly detection method.

[0173] The above is an illustrative scheme of a computing device according to this embodiment. It should be noted that the technical solution of this computing device and the technical solution of the above-described anomaly detection method belong to the same concept. For details not described in detail in the technical solution of the computing device, please refer to the description of the technical solution of the above-described anomaly detection method.

[0174] An embodiment of this specification also provides a computer-readable storage medium storing computer-executable instructions that, when executed by a processor, implement the steps of the above-described anomaly detection method.

[0175] The above is an illustrative embodiment of a computer-readable storage medium. It should be noted that the technical solution of this storage medium and the technical solution of the above-described anomaly detection method belong to the same concept. Details not described in detail in the technical solution of the storage medium can be found in the description of the technical solution of the above-described anomaly detection method.

[0176] An embodiment of this specification also provides a computer program, wherein when the computer program is executed in a computer, it causes the computer to perform the steps of the above-described anomaly detection method.

[0177] The above is an illustrative example of a computer program according to this embodiment. It should be noted that the technical solution of this computer program and the technical solution of the above-described anomaly detection method belong to the same concept. Details not described in detail in the computer program's technical solution can be found in the description of the technical solution of the above-described anomaly detection method.

[0178] The foregoing has described specific embodiments of this specification. Other embodiments are within the scope of the appended claims. In some cases, the actions or steps recited in the claims may be performed in a different order than that shown in the embodiments and may still achieve the desired result. Furthermore, the processes depicted in the drawings do not necessarily require the specific or sequential order shown to achieve the desired result. In some embodiments, multitasking and parallel processing are possible or may be advantageous.

[0179] The computer instructions include computer program code, which may be in the form of source code, object code, executable file, or certain intermediate forms. The computer-readable medium may include: any entity or device capable of carrying the computer program code, recording media, USB flash drive, portable hard drive, magnetic disk, optical disk, computer memory, read-only memory (ROM), random access memory (RAM), electrical carrier signals, telecommunication signals, and software distribution media, etc. It should be noted that the content included in the computer-readable medium may be appropriately added or removed according to the requirements of patent practice. For example, in some regions, according to patent practice, computer-readable media may not include electrical carrier signals and telecommunication signals.

[0180] It should be noted that, for the sake of simplicity, the foregoing method embodiments are all described as a series of actions. However, those skilled in the art should understand that the embodiments in this specification are not limited to the described order of actions, because according to the embodiments in this specification, some steps can be performed in other orders or simultaneously. Furthermore, those skilled in the art should also understand that the embodiments described in this specification are all preferred embodiments, and the actions and modules involved are not necessarily essential to the embodiments in this specification.

[0181] In the above embodiments, the descriptions of each embodiment have different focuses. For parts not described in detail in a certain embodiment, please refer to the relevant descriptions of other embodiments.

[0182] The preferred embodiments disclosed above are merely illustrative of this specification. The optional embodiments do not exhaustively describe all details, nor do they limit the invention to the specific implementations described. Clearly, many modifications and variations can be made based on the embodiments described herein. These embodiments are selected and specifically described in this specification to better explain the principles and practical applications of the embodiments, thereby enabling those skilled in the art to better understand and utilize this specification. This specification is limited only by the claims and their full scope and equivalents.

Claims

1. An anomaly detection method, characterized in that, include: In response to an event that the target item is stored in the spectral scanning area, a spectral scanning device is invoked to scan the target item and obtain the spectral scanning information of the target item; Obtain the tag identification information of the target item, wherein the tag identification information is obtained based on the content recognition of the item tag of the target item; The spectral scanning information and the tag identification information are input into the anomaly detection model to obtain the anomaly detection result of the target item.

2. The method according to claim 1, characterized in that, The step of calling a spectral scanning device to scan the target item and obtain the spectral scanning information of the target item includes: Send a spectral scanning command for the target item to the spectral scanning device, and receive the spectral feature map of the target item fed back by the spectral scanning device in response to the spectral scanning command; The spectral feature map is analyzed to obtain the spectral scanning information of the target item.

3. The method according to claim 2, characterized in that, The step of analyzing the spectral feature map to obtain the spectral scanning information of the target item includes: The spectral feature map is input into the spectral analysis model to obtain the spectral scanning information of the target item.

4. The method according to claim 3, characterized in that, Before inputting the spectral feature map into the spectral analysis model to obtain the spectral scanning information of the target item, the method further includes: Obtain a sample set, wherein the sample set includes multiple sample spectral feature maps and sample spectral scanning information corresponding to the multiple sample spectral feature maps respectively; The initial analytical model is invoked to analyze the spectral feature maps of the multiple samples, thereby obtaining the predicted spectral scanning information corresponding to the spectral feature maps of the multiple samples respectively; Based on the sample spectral scanning information and the predicted spectral scanning information, the initial analytical model is trained to obtain the spectral analytical model.

5. The method according to claim 1, characterized in that, The spectral scanning information corresponds to at least one feature dimension; The step of inputting the spectral scanning information and the tag identification information into the anomaly detection model to obtain the anomaly detection result of the target item includes: The spectral scanning information and the label recognition information are input into the anomaly detection model to obtain at least one sub-detection result, wherein the sub-detection result corresponds one-to-one with the feature dimension; Based on the at least one sub-detection result, an anomaly detection result for the target item is determined.

6. The method according to claim 5, characterized in that, The at least one feature dimension includes an item category, and the spectral scan information includes the confidence score of the target item under at least one item category; The step of inputting the spectral scanning information and the tag identification information into the anomaly detection model to obtain at least one sub-detection result includes: The anomaly detection model is used to extract the identifying item type of the target item from the label recognition information; The confidence score of the spectral scanning information under the identified item category is determined as the category matching degree, and the sub-detection result is determined based on the category matching degree.

7. The method according to claim 1, characterized in that, The step of obtaining the tag identification information of the target item includes: The image acquisition device is invoked to acquire an image of the item label of the target item; Text recognition is performed based on the item label image to obtain the label recognition information of the target item.

8. The method according to any one of claims 1 to 7, characterized in that, After inputting the spectral scanning information and the tag identification information into the anomaly detection model to obtain the anomaly detection result of the target item, the method further includes: If the anomaly detection result indicates that the target item is abnormal, the current environmental information and environmental tag information of the target item are obtained; Based on the current environmental information and the environmental tag information, determine the cause of the anomaly of the target item.

9. The method according to any one of claims 1 to 7, characterized in that, After inputting the spectral scanning information and the tag identification information into the anomaly detection model to obtain the anomaly detection result of the target item, the method further includes: If the anomaly detection result indicates that the target item is abnormal, an anomaly alarm will be issued.

10. A method for detecting anomalies in hazardous chemicals, characterized in that, include: In response to an event that a hazardous chemical is stored in the spectral scanning region, a spectral scanning device is invoked to scan the hazardous chemical to obtain the spectral scanning information of the hazardous chemical, wherein the hazardous chemical includes at least one of solid hazardous chemicals, liquid hazardous chemicals and gaseous hazardous chemicals; Obtain the label identification information of the hazardous chemical, wherein the label identification information is obtained based on the content recognition of the hazardous chemical label; The spectral scanning information and the label identification information are input into the anomaly detection model to obtain the anomaly detection results of the hazardous chemical.

11. An anomaly detection device, characterized in that, include: The scanning module is configured to, in response to an event that the target item is stored in the spectral scanning area, invoke a spectral scanning device to scan the target item and obtain the spectral scanning information of the target item. The acquisition module is configured to acquire the tag identification information of the target item, wherein the tag identification information is obtained based on the content recognition of the item tag of the target item; The detection module is configured to input the spectral scanning information and the tag identification information into the anomaly detection model to obtain the anomaly detection result of the target item.

12. A hazardous chemical anomaly detection device, characterized in that, include: The scanning module is configured to, in response to an event that a hazardous chemical is stored in the spectral scanning region, invoke a spectral scanning device to scan the hazardous chemical and obtain the spectral scanning information of the hazardous chemical, wherein the hazardous chemical includes at least one of solid hazardous chemicals, liquid hazardous chemicals and gaseous hazardous chemicals; The acquisition module is configured to acquire the label identification information of the hazardous chemical, wherein the label identification information is obtained based on the content recognition of the hazardous chemical label of the hazardous chemical; The detection module is configured to input the spectral scanning information and the label identification information into the anomaly detection model to obtain the anomaly detection results of the hazardous chemical.

13. A computing device, characterized in that, include: Memory and processor; The memory is used to store computer-executable instructions, and the processor is used to execute the computer-executable instructions, which, when executed by the processor, implement the steps of the anomaly detection method according to any one of claims 1-9.

14. A computer-readable storage medium, characterized in that, It stores computer-executable instructions that, when executed by a processor, implement the steps of the anomaly detection method according to any one of claims 1-9.

15. A computer program product, characterized in that, Includes a computer program / instructions that, when executed by a processor, implement the steps of the anomaly detection method according to any one of claims 1-9.

Images