Cold source disaster-causing object intelligent identification and early warning method and device integrating AI algorithm

Abstract

The application relates to the technical field of cold source disaster-causing object detection, and discloses a cold source disaster-causing object intelligent identification and early warning method and device integrated with an AI algorithm. The cold source disaster-causing object intelligent identification and early warning method and device integrated with the AI algorithm effectively reduce misidentification, improve the accuracy of cold source disaster-causing object identification, and convert collected data into an input format suitable for an AI algorithm according to a preset format and a preprocessing mode, so that the AI algorithm can efficiently process data, laying a foundation for subsequent accurate analysis and prediction. Moreover, a deep learning model is trained by using the preprocessed data, and a prediction result is output based on a set confidence threshold, so that the prediction has certain reliability and scientificity. Finally, the confidence in the prediction result can be used to dynamically adjust an alarm threshold, so that over-sensitive alarm is avoided, the rationality and effectiveness of early warning are improved, and unnecessary alarm interference is reduced.

Description

TECHNICAL FIELD

[0001] The application relates to the technical field of cold source disaster-causing object detection, in particular to a cold source disaster-causing object intelligent identification and early warning method and device integrated with an AI algorithm. BACKGROUND

[0002] The cold source disaster-causing object intelligent identification and early warning method and device integrated with an AI algorithm is a technical means for realizing real-time monitoring and automatic early warning by combining artificial intelligence technology, which is specially used for identifying potential dangerous objects (such as pipeline leakage and cold source equipment failure) in a low-temperature environment. The device uses a deep learning algorithm to process data streams from various sensors and quickly analyze and identify possible cold source disaster-causing conditions, thereby providing a scientific basis for timely preventive measures. However, this technology currently faces several problems that need to be solved. First, how to accurately regulate the characteristic parameters of cold source disaster-causing objects in different temperature intervals is one of the key challenges. Since changes in environmental temperature may lead to misidentification, the accuracy of judgment is affected. Second, the optimization of data input format and preprocessing method for the AI algorithm is crucial. Data noise can significantly affect the prediction accuracy of the model, and an effective noise reduction mechanism needs to be found. Finally, it is also necessary to adjust the confidence threshold of the deep learning model. Setting too low may cause too many false positive alarms, while setting too high may miss important warning signals. Therefore, finding the appropriate confidence threshold to balance the problem of excessive sensitive alarm is an important consideration. SUMMARY

[0003] The cold source disaster-causing object intelligent identification and early warning method and device integrated with an AI algorithm provided by the embodiments of the present disclosure at least partially solve the problems existing in the prior art.

[0004] The cold source disaster-causing object intelligent identification and early warning method integrated with an AI algorithm comprises the following steps:

[0005] S101. Dynamically regulating the characteristic parameters of cold source disaster-causing objects based on different temperature intervals to reduce misidentification.

[0006] S102. Collecting data after the characteristic parameter regulation, and converting the data into an input format suitable for the AI algorithm according to a preset format and preprocessing method.

[0007] S103. Training a deep learning model using the preprocessed data and outputting a prediction result based on a set confidence threshold.

[0008] S104. Dynamically adjusting the alarm threshold based on the confidence in the prediction result to prevent excessive sensitive alarm.

[0009] The dynamic regulation of the characteristic parameters of cold source disaster-causing objects based on different temperature intervals comprises the following steps:

[0010] Divide the ambient temperature into several different intervals, and define the corresponding characteristic parameter adjustment strategy for each interval;

[0011] Collect the actual temperature distribution of each temperature zone, and determine the temperature threshold Tk(where k represents the kth temperature zone);

[0012] Determine the new characteristic parameter P=Pbase+ΔP in each temperature zone according to the following formula, where ΔP is the specific adjustment value of the temperature zone;

[0013] If the ambient temperature T∈[Tk-1,Tk) and the error E<Tmax, apply the above characteristic parameter P to provide data to the model; where Pbase is the base parameter, Tk-1 and Tk represent the temperature zone range limits, and E represents the difference between the actual measurement and the preset parameter.

[0014] When dynamically updating the characteristic parameters of the cold source disaster-causing substances based on the specific conditions of different temperature zones, the following steps are further included:

[0015] Determine the initial parameter baseline value;

[0016] Calculate the variance S of the data collected in each temperature zone;

[0017] Based on the trend of variance S in each interval, adjust the parameter adjustment amplitude A(S)=α*sqrt(S) of the corresponding interval;

[0018] If the current variance increases significantly (e.g., S>δ), increase A, otherwise keep it unchanged, where α represents the gain coefficient and δ represents the decision threshold.

[0019] When dynamically calibrating the characteristic parameters based on the data collected in each temperature zone, the following operations should also be included:

[0020] Periodically record and summarize the temperature data over a period of time;

[0021] Calculate the maximum and minimum temperature range D=Tmax(k)Tmin(k) in each interval over this period of time;

[0022] Analyze the maximum error Emax caused by external factors in each interval;

[0023] If the range exceeds a certain limit and Emax exceeds the set threshold, then according to C=max(D / Emax,γ), the monitoring intensity of these abnormal areas is strengthened, that is, the sensitivity of the characteristic parameter of the section is increased, C refers to the control factor, and γ refers to the preset minimum ratio.

[0024] After adjusting the monitoring focus based on the temperature range evaluation results, the process is further refined as follows:

[0025] Periodic review and correction mechanism is implemented for each temperature zone;

[0026] Data bias accumulated from long-term operation is considered comprehensively;

[0027] Specific solutions are designed for places where certain problems occur frequently;

[0028] When it is found that the misidentification rate exceeds r or the error of three consecutive samples is higher than b in a certain interval, the emergency plan processing mode is started for the temperature zone, r is the tolerance error ratio, and b represents the upper limit of the allowable deviation range.

[0029] The process of optimizing the alarm sensitivity based on misidentification frequency statistics particularly includes:

[0030] The probability of false alarms Fp and the proportion of missed alarms Fneg in each temperature range are counted;

[0031] All potential dangerous conditions and their possible impact severity levels are checked;

[0032] The likelihood of potential disaster events is predicted;

[0033] The warning quality of different intervals is comprehensively evaluated by Q=(Fp / w1)+(Fneg / w2), and stricter standards are adopted for areas with higher Q, w1 and w2 are weight factors used to adjust the different proportions of the contributions of the two.

[0034] In terms of filtering the impact of data noise, in addition to the previous methods, the following improvements are included:

[0035] Adaptive filtering technology is applied to process the raw data output by the sensor;

[0036] A special denoising function is designed to remove random noise components;

[0037] More reliable preprocessing procedures are established using historical experience and knowledge;

[0038] If the signal-to-noise ratio SNR < smin (smin represents the minimum required signal-to-noise ratio), the collection scheme should be adjusted to improve data quality, SNR is a quantitative indicator of the relative intensity contrast between signal and noise.

[0039] More detailed specifications are made on how to effectively reduce unnecessary complexity in model input:

[0040] A feature selection framework is established to select a subset of dimensions that best represent the characteristics of the target;

[0041] The proportion of useless information is reduced to improve training efficiency;

[0042] Dimensionality reduction using sparse coding, preserving core attributes intact;

[0043] When the remaining feature dimension d is greater than a given threshold Nt and the cross-validation accuracy ACC decreases, appropriately reduce the value of d until the optimal performance level is met.

[0044] To prevent the model from relying too much on certain key but unstable factors during prediction, causing instability and uncertainty, the following strategies are proposed:

[0045] A stability evaluation module is constructed to monitor fluctuations during each iteration;

[0046] A comprehensive scan of all feature vectors is conducted to find weak links;

[0047] Strengthen the correlation strength between important nodes to ensure network robustness;

[0048] If any parameter Xn at any time changes more than ∈ or the direction of change is the same for multiple frames in a row, trigger the safety check program to confirm whether to continue learning and updating along this path, where ∈ represents the change amplitude limit standard, which reflects the tolerance level for system stability.

[0049] On the basis of continuously optimizing confidence scores, the adaptability of the system is further enhanced, which is embodied in the following specific aspects:

[0050] Multiple rounds of experimental tests are conducted to obtain optimal performance parameter configurations under various environments;

[0051] Continuous collection of actual use feedback to guide the direction of next iteration optimization;

[0052] Customized personalized service options based on user needs to enhance user experience;

[0053] For each prediction, if the confidence Conf is lower than the critical threshold θ, automatically reduce the urgency level L of the next alarm until it returns to normal after multiple high-accuracy verifications.

[0054] The cold source disaster-causing object intelligent identification and early warning device integrated with AI algorithm, which is executed by the cold source disaster-causing object intelligent identification and early warning method integrated with AI algorithm, the system comprises:

[0055] The control module is responsible for managing and coordinating various data collection devices, dynamically adjusting the working parameters of the collection devices according to different environmental conditions, monitoring requirements, etc., such as the shooting frequency of underwater cameras, the detection range and frequency of sonar devices, etc., to ensure that high-quality, comprehensive and relevant data for cold source disaster-causing object identification and early warning are collected.

[0056] Conversion module: convert some non-numeric data such as text description, category label, etc. into numerical data suitable for AI algorithm processing, for example, convert the type name of cold source disaster into corresponding digital code, or convert the color information in the image into numerical vector, so that the AI model can calculate and learn these data.

[0057] Learning module: automatically extract key features from a large number of cold source disaster related data collected from the data acquisition module, for example, for image data captured by underwater camera, the learning module can learn the shape, texture, color and other visual features of cold source disaster through convolutional neural network technology; for sonar data, it can learn the reflection intensity, echo time and other acoustic features of the target object, so as to convert these complex raw data into more valuable feature table for identification and early warning.

[0058] In summary, the beneficial technical effects of the present application are:

[0059] 1、The integrated AI algorithm cold source disaster intelligent identification and early warning method and device, by dynamically adjusting and controlling the cold source disaster characteristic parameters based on different temperature intervals, effectively reduces the misidentification, improves the accuracy of cold source disaster identification.

[0060] 2、The integrated AI algorithm cold source disaster intelligent identification and early warning method and device, the collected data is converted into an input format suitable for AI algorithm according to the preset format and preprocessing mode, ensuring that the AI algorithm can efficiently process data, laying a foundation for subsequent accurate analysis and prediction.

[0061] 3、The integrated AI algorithm cold source disaster intelligent identification and early warning method and device, the data after preprocessing is used to train a deep learning model, and the prediction result is output based on the set confidence threshold, so that the prediction has certain reliability and scientific nature.

[0062] 4、The integrated AI algorithm cold source disaster intelligent identification and early warning method and device, the alarm threshold is dynamically adjusted according to the confidence in the prediction result, avoiding excessive sensitive alarm, improving the rationality and effectiveness of the early warning, and reducing unnecessary alarm interference. BRIEF DESCRIPTION OF DRAWINGS

[0063] Figure 1 is the flow chart of the integrated AI algorithm cold source disaster intelligent identification and early warning method of the present application.

[0064] Figure 2 is the flow chart of the integrated AI algorithm cold source disaster intelligent identification and early warning system of the present application. DETAILED DESCRIPTION

[0065] Embodiments of the present application are described below in detail, examples of which are shown in the accompanying drawings.

[0066] In the description of the present specification, the description of the terms "certain embodiments", "one embodiment", "some embodiments", "illustrative embodiments", "example", "specific example", or "some examples" means that the specific features, structures, materials or characteristics described in connection with the described embodiments or examples are included in at least one embodiment or example of the present application. In the present specification, the illustrative description of the above terms does not necessarily refer to the same embodiment or example. Also, the specific features, structures, materials or characteristics described can be combined in any one or more embodiments or examples in a suitable manner.

[0067] Referring to the accompanying drawings Figure 1 The integrated AI algorithm cold source disaster-causing object intelligent identification and early warning method of the present application is described, which includes dynamically regulating the characteristic parameters of cold source disaster-causing objects in different temperature intervals to reduce the possibility of misidentification. By adjusting the characteristic parameters, the data changes under different environmental conditions can be effectively adapted. At the same time, the data is collected and subjected to strict formatting and preprocessing steps before being converted into a format suitable for input into a deep learning model. Based on this, the trained model outputs a prediction result based on a pre-set confidence threshold. Finally, according to the confidence in the obtained prediction result, dynamic adjustment of the alarm threshold is implemented to avoid excessive sensitivity of the alarm.

[0068] Based on different temperature intervals, the system first dynamically regulates the characteristic parameters of cold source disaster-causing objects. This means that in different working environments (such as high-temperature areas, normal temperatures, and low-temperature areas), the environmental state is monitored by sensors and other devices, and then the key parameters that are prone to misinterpretation or false identification under specific conditions are optimized, such as object shape, material thermal properties, and surface temperature fluctuations. These adjustments ensure that high-accuracy data is still obtained in complex and variable environments for input into the next processing module.

[0069] In the specific operation process, adjustment rules in multiple sub-dimensions can be set according to different categories of targets, and an intelligent judgment mechanism can be formed by combining historical experience data accumulation to update the weights of each factor in real time. For example, in a chemical production workshop, there are various chemical medicine tanks stored in different heat preservation rooms, and the environment they are in can differ greatly from indoor to outdoor extreme weather. In view of this scene characteristic, the adjustment formula for the associated attributes of each warehouse type can be built into the software system, and when actual detection is performed, the related physical indicators can be automatically corrected in response to environmental changes to make the subsequent link processing more accurate and efficient.

[0070] After the initial parameter adjustment is completed, the device will reorganize and classify all data streams from the field monitoring station according to a unified planning format and perform necessary signal cleaning methods to meet the information requirements of downstream algorithms, i.e. convert the original collected data into a form that meets the specifications of a specific artificial intelligence technology by predefining the format and method, such as removing redundant parts and only retaining key elements that help identify and analyze, or using mathematical transformations to eliminate certain periodic disturbance components to ensure consistency and stability between data samples, which directly improves the quality of the final output, reduces computational burden, and enhances model robustness and generalization ability.

[0071] The clean and standardized data set processed by the above method can be used to train advanced statistical classifiers such as deep neural networks with self-learning ability to gradually master the ability to distinguish between normal situations and other potential hazards. During the entire iterative evolution period, new batches of labeled samples need to be continuously supplemented to the training set to expand coverage and improve feature extraction logic until the desired performance level is reached before entering the formal use stage. In this stage, a large number of trial experiments will be involved to evaluate model precision and recall performance to find the best configuration point. Special attention should be paid to prevent excessive or insufficient bias or complete mismatch.

[0072] The complete product version after strict debugging and verification is put into operation according to the previously determined credible score threshold as the trigger condition to issue risk warning information to the user end platform to inform relevant personnel to take emergency refuge measures to contain and prevent the spread of loss and deterioration of the situation. However, relying solely on fixed limits for judgment will cause frequent false alarms and disturb daily management affairs, so a flexible response strategy is introduced to establish a dynamic decision-making model by learning the statistical laws of past cumulative cases, i.e. with the system accumulating enough long-term stable operation statistical data to support a series of probability estimation functions to guide the current situation where a single prediction score falls within a more reasonable and scientific range to adjust the alarm frequency. This not only protects the system's sensitive touch, but also balances the cost-benefit relationship, making the entire early warning mechanism more practical and closer to the needs of application scenarios.

[0073] In one embodiment, it is assumed that the innovative technology is started to carry out comprehensive diagnostic inspection activities before the ice blockage precursor is found in the subway station air conditioning chilled water pipeline. In the initial stage, the engineer team first collects the microclimate state and the concentration information of the substances carried by each node position in the cold water supply chain in the station area, and calculates the expected problem characteristic curve by means of professional knowledge background knowledge, and then formulates a set of detailed and specific adjustment standards by comparing theoretical literature materials. Then, the professional measuring instruments are used to scan the readings along the pipeline path in turn, and the visual table is generated to facilitate further comparison and calibration. Subsequently, the data cleaning process is carried out according to the pre-written number to remove abnormal burrs and maintain the overall smoothness to facilitate the efficient operation of the model training and recognition process. After several rounds of careful polishing and optimization, a highly reliable and accurate report is finally obtained to help the operation department to make early preparations for various preventive measures to ensure the safety of citizens and smooth travel environment without interference and threat.

[0074] The integrated AI algorithm cold source disaster-causing object intelligent identification and early warning method and device of the application comprises:

[0075] Firstly, the overall steps involve data collection, feature parameter dynamic regulation, data preprocessing, deep learning model training, confidence threshold setting and alarm control of cold source disaster-causing objects. Specifically, the system will comprehensively and real-time monitor the cold source disaster-causing objects (such as leakage gas or ice blockage phenomenon) in different temperature intervals through various sensors, and adjust the feature parameters according to the specific conditions of the environment to reduce the possibility of misidentification.

[0076] In order to solve the problem of how to regulate the feature parameters of cold source disaster-causing objects in different temperature intervals to solve the problem of misidentification, the system will automatically select the most suitable feature regulation mode according to the different temperature environments during actual operation, that is, the system is built-in with a series of feature regulation parameter combination schemes corresponding to various temperature conditions. These schemes can effectively reduce the misidentification rate caused by the uncertainty of temperature changes, for example, for the problem of solidification of special materials in low temperature environment, the system can enhance the effective identification of target substances under this condition by increasing the specific wave band spectrum or improving the detection sensitivity of certain physical and chemical properties, while filtering out the temperature-related fluctuations that are not disaster-causing.

[0077] Next, regarding how to regulate the data input format and preprocessing method of the AI algorithm to solve the problem of data noise affecting the accuracy of the model, the application designs a special data conversion and preprocessing logic to ensure that the original monitoring data is converted into a standardized and denoised model input form. This process consists of two steps: first, normalize the collected feature parameters - this step ensures that all types of data are compared on the same scale, eliminating interference caused by dimensional differences; second, deeply process the possible data noise problem, introduce adaptive filter and feature enhancement algorithm to remove or weaken the influence of abnormal mutation value. In addition, according to the characteristics of the cold source scene, a redundant verification link is specially set to cross-compare the consistency and reliability of multiple measurement results, so as to further optimize the model training quality and ensure the robustness of the prediction results.

[0078] Finally, around how to regulate the confidence threshold of the deep learning model to solve the problem of over-sensitive alarm, the application proposes a method of dynamically setting the alert limit based on real-time feedback mechanism. Traditional fixed confidence is too low to cause frequent false alarms and disturb people, and too high quality standard may miss the real crisis moment. Our solution combines actual environmental variable factors, past case learning experience and current system performance evaluation three comprehensive indicators to realize the accurate configuration of the best alarm sensitivity. In this way, under normal operation, a relatively strict threshold can be set to avoid unnecessary warnings; when potential high risk is detected, the alarm threshold is quickly and flexibly adjusted down to start the emergency response program in time. In addition, a self-checking and review mechanism is established to regularly review and analyze false alarm situations and fine-tune parameters, forming a benign self-evolution closed-loop system, so as to ensure that the equipment maintains high and stable operation efficiency for a long time, and effectively meets the user's safety protection needs.

[0079] In one embodiment, an integrated AI algorithm cold source disaster-causing object intelligent identification and early warning device is also disclosed. The evaluation system is executed by the above-mentioned integrated AI algorithm cold source disaster-causing object intelligent identification and early warning method, as shown in Figure 2 The system includes:

[0080] The regulation module is responsible for managing and coordinating various data collection devices, dynamically adjusting the working parameters of the collection devices such as the shooting frequency of underwater cameras, the detection range and frequency of sonar devices, etc. according to different environmental conditions, monitoring requirements, etc., to ensure that high-quality, comprehensive and relevant data for cold source disaster-causing object identification and early warning are collected.

[0081] Conversion module: convert some non-numeric data, such as text description, category label, etc., into numerical data suitable for AI algorithm processing, for example, convert the type name of cold source disaster-causing object into corresponding digital code, or convert the color information in the image into numerical vector, so that the AI model can calculate and learn these data.

[0082] Learning module: automatically extract key features from a large amount of cold source disaster-causing object related data collected from the data acquisition module, for example, for image data taken by underwater camera, the learning module can learn the shape, texture, color and other visual features of cold source disaster-causing object through convolutional neural network technology; for sonar data, it can learn the reflection intensity, echo time and other acoustic features of the target object, so as to convert these complex raw data into more valuable feature tables for identification and early warning.

[0083] The various processes and actions described above, such as a method, can be performed by a CPU. For example, in some embodiments, a method can be implemented as a computer software program tangibly embodied in a machine-readable medium, such as a memory. In some embodiments, part or all of the computer program can be loaded and / or installed on a computing device via a ROM and / or a communication unit. When the computer program is loaded into the RAM and executed by the CPU, the above-described method or action based on multi-sensor fusion intelligent cold source disaster-causing object monitoring can be performed.

[0084] The present disclosure relates to methods, computing devices, computer-readable storage media, and / or computer program products. The computer program product can include computer-readable program instructions for performing various aspects of the present disclosure.

[0085] Computer readable storage media can be tangible storage media which can retain and store instructions for use by an instruction execution device. Computer readable storage media can be, for example, but is not limited to, an electronic storage device, a magnetic storage device, an optical storage device, an electromagnetic storage device, a semiconductor storage device, or any suitable combination of the foregoing. More specific examples (a non-exhaustive list) of computer readable storage media include the following: a portable computer diskette, a hard disk, a random access memory (RAM), a read-only memory (ROM), an erasable programmable read-only memory (EPROM or Flash memory), a static random access memory (SRAM), a portable compact disc read-only memory (CD-ROM), a digital versatile disk (DVD), a memory stick, a floppy disk, a mechanically encoded device such as punch-cards or raised structures in a groove having instructions recorded thereon, and any suitable combination of the foregoing. A computer readable storage medium, as used herein, is not to be construed as being transitory signals per se, such as radio waves or other freely propagating electromagnetic waves, electromagnetic waves propagating through a waveguide or other transmission media (e.g., light pulses passing through a fiber-optic cable), or electrical signals transmitted through a wire.

[0086] Computer readable program instructions described herein can be downloaded to respective computing / processing devices from a computer readable storage medium or to an external computer or external storage device via a network, for example, the Internet, a local area network, a wide area network and / or a wireless network. The network can comprise copper transmission cables, optical transmission fibers, wireless transmission, routers, firewalls, switches, gateway computers and / or edge servers. A network adapter card or network interface in each computing / processing device receives computer readable program instructions from the network and forwards the computer readable program instructions for storage in a computer readable storage medium within the respective computing / processing device.

[0087] Computer readable program instructions for carrying out operations of the present disclosure can be assembly instructions, instruction-set-architecture (ISA) instructions, machine instructions, machine dependent instructions, microcode, firmware instructions, state-setting data, or source code or object code written in any combination of one or more programming languages, including an object oriented programming language such as Smalltalk, C++ or the like, and conventional procedural programming languages, such as the "C" programming language or similar programming languages. The computer readable program instructions can execute entirely on the user's computer, partly on the user's computer, as a stand-alone software package, partly on the user's computer and partly on a remote computer or entirely on the remote computer or server. In the latter scenario, the remote computer can be connected to the user's computer through any type of network, including a local area network (LAN) or a wide area network (WAN), or the connection can be made to an external computer (for example, through the Internet using an Internet Service Provider). In some embodiments, electronic circuitry including, for example, programmable logic circuitry, field-programmable gate array (FPGA), or programmable logic array (PLA) can execute the computer readable program instructions by utilizing state information of the computer readable program instructions to personalize the electronic circuitry, in order to perform aspects of the present disclosure.

[0088] These computer readable program instructions can be provided to a processor of a general purpose computer, special purpose computer, or other programmable data processing apparatus to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing apparatus, create means for implementing the functions / acts specified in the flowchart and / or block diagram block or blocks. These computer readable program instructions can also be stored in a computer readable storage medium that can include a non-transitory computer readable storage medium that can direct a computer, a programmable data processing apparatus, and / or other devices to function in a particular manner, such that the computer readable storage medium having instructions stored therein comprises an article of manufacture including a manufacture of one or more aspects of a multi-sensor fusion based intelligent cold source hazard monitoring method or actions.

[0089] The computer readable program instructions can also be loaded onto a computer, other programmable data processing apparatus, or other device to cause a series of operational steps to be performed on the computer, other programmable apparatus or other device to produce a computer implemented process such that the instructions which execute on the computer, other programmable data processing apparatus, or other device implement a multi-sensor fusion based intelligent cold source hazard monitoring method or actions.

[0090] Embodiments of the present disclosure have been described above, the above description is exemplary, not exhaustive, and is not limited to the disclosed embodiments. Many modifications and changes are obvious to those of ordinary skill in the art without departing from the scope and spirit of the described embodiments. The selection of terms used herein is intended to best explain the principles of the embodiments, practical application, or technical improvement in the market, or to enable other ordinary skilled persons in the art to understand the embodiments disclosed herein.

[0091] Based on the same inventive concept, the embodiments of the present application provide an intelligent terminal, comprising a memory and a processor, the memory stores a computer program capable of being loaded and executed by the processor.

[0092] Although the embodiments of the present application have been shown and described above, it should be understood that the above-described embodiments are exemplary and should not be construed as limiting the present application, and those of ordinary skill in the art can make changes, modifications, replacements and variations to the above-described embodiments within the scope of the present application.

Claims

1. A cold source disaster-causing substance intelligent recognition and early warning method integrated with an AI algorithm, characterized in that, Comprise: S101, dynamically regulate the characteristic parameters of the cold source disaster-causing object based on different temperature intervals to reduce misidentification; The dynamic regulation specifically includes: dividing the environmental temperature into several different intervals, and defining the corresponding characteristic parameter adjustment strategy for each interval; collecting the actual temperature distribution of each temperature zone to determine the temperature threshold Tk, wherein k represents the kth temperature zone; determining the new characteristic parameter according to the formula P=Pbase+ΔP in each temperature zone, wherein Pbase is the base parameter, and ΔP is the specific adjustment value of the temperature zone; calculate the variance S of the data collected in each temperature zone, based on the change trend of the variance S in each interval, adjust the parameter adjustment amplitude A=α*sqrt(S) of the corresponding interval, if the current variance S>δ, wherein δ is the judgment threshold, then increase A, otherwise keep unchanged, α represents the gain coefficient; if the environmental temperature T∈[Tk−1,Tk) and the error E<Tmax, wherein E represents the difference between the actual measurement and the preset parameter, and Tmax is the maximum allowed error, then apply the above characteristic parameter P to provide data to the model; S102, collect the data after the characteristic parameter regulation, and convert it into an input format suitable for AI algorithm according to the preset format and preprocessing method; the preprocessing method includes applying adaptive filtering technology to process sensor output raw data, designing special denoising functions to remove random noise components, and using historical experience knowledge to establish a cold source disaster-causing object preprocessing process; if the signal-to-noise ratio SNR<smin, wherein smin represents the minimum required signal-to-noise ratio, then adjust the collection scheme to improve data quality; S103, train the deep learning model using the preprocessed data, and output the prediction result based on the set confidence threshold; S104, dynamically adjust the alarm threshold based on the confidence in the prediction result to prevent over-sensitive alarm.

2. The integrated AI algorithm cold source disaster-causing object intelligent identification and early warning method according to claim 1, characterized in that, When dynamically calibrating the characteristic parameters based on the data collected in each temperature zone, the following operations should also be included: Periodically record and summarize the temperature data over a period of time; Calculate the maximum and minimum temperature range D=Tmax(k)-Tmin(k) in each interval within this period of time; Analyze the maximum error Emax caused by external factors in each interval; If the range exceeds a certain limit value and Emax exceeds the set threshold, then according to C=max(D / Emax,γ), wherein C refers to the control factor and γ refers to the preset minimum ratio, the monitoring intensity of these abnormal areas is strengthened, that is, the sensitivity of the characteristic parameters in this interval is improved.

3. The cold source disaster-causing object intelligent identification and early warning method integrated with an AI algorithm according to claim 2, characterized in that, After adjusting the monitoring focus based on the temperature range evaluation result, further refinement is made as follows: Periodically review and correct each temperature zone; Consider the data bias accumulated over a long period of time; Design targeted solutions for specific problem-prone areas; When it is found that the misidentification rate of a certain interval exceeds r or the error of three consecutive samples is higher than b, start the emergency plan processing mode for this temperature zone, r is the tolerance error ratio, and b represents the upper limit of the allowable deviation range.

4. The cold source disaster-causing object intelligent identification and early warning method integrated with an AI algorithm according to claim 3, characterized in that, The process of optimizing the alarm sensitivity based on misidentification frequency statistics specifically includes: Statistics the probability of false alarm Fp and the proportion of missed alarm Fneg in each temperature range; Check all potential dangerous conditions and their possible impact severity level; Predict the likelihood of potential disaster events; Comprehensively evaluate the warning quality of different intervals through Q=(Fp / w1)+(Fneg / w2), and take more stringent standards for areas with higher Q, w1 and w2 are weight factors used to adjust the different proportions of their contributions.

5. The cold source disaster-causing object intelligent identification and early warning method integrated with AI algorithms according to claim 4, characterized in that, More detailed specifications are given on how to effectively reduce unnecessary complexity in model input: Establish a feature selection framework to select a subset of dimensions that best represent the characteristics of the target; Reduce the proportion of useless information to improve training efficiency; Use sparse coding to reduce dimensions while preserving core attributes; When the remaining feature dimension d is greater than the given threshold Nt and the cross-validation accuracy ACC decreases, appropriately reduce the value of d until the best performance level is met.

6. The cold source disaster-causing object intelligent identification and early warning method integrated with an AI algorithm according to claim 5, characterized in that, To prevent the model from relying too much on some key but unstable factors during prediction, causing instability and uncertainty, the following strategies are proposed: Build a stability evaluation module to monitor fluctuations during each iteration; Comprehensively scan all feature vectors to find weak links; Strengthen the correlation between important nodes to ensure network robustness; If any parameter Xn at any time varies more than Or the same direction of change in a plurality of frames, trigger security check procedure, to confirm whether to continue along this path learning update, wherein Represent the change limit standard, which reflects the tolerance of the system stability; Further enhance the system's adaptive response based on continuous optimization of confidence scores, which is embodied in the following system: Implement multiple rounds of experimental tests to obtain optimal performance parameter configurations in various environments; Continuously collect actual use feedback to guide the next iteration optimization direction; Customize personalized service options based on user needs to enhance user experience; For each prediction, if the confidence Conf is lower than the critical threshold θ, automatically reduce the urgency level L of the next alarm until it returns to normal state after multiple high-accuracy verifications.

7. A cold source disaster-causing substance intelligent identification and early warning device integrated with an AI algorithm, characterized in that, The device is executed by the integrated AI algorithm cold source disaster-causing object intelligent identification and early warning method of claim 6, and the device comprises: A control module: responsible for managing and coordinating various data collection devices, dynamically adjusting the working parameters of the collection devices according to different environmental conditions and monitoring needs, to ensure that high-quality, comprehensive and cold source disaster-causing object identification and early warning related data are collected; A conversion module: converts some non-numeric data into numeric data suitable for AI algorithm processing, so that the AI model can calculate and learn these data; A learning module: automatically extracts key features from a large amount of cold source disaster-causing object related data collected from the data collection module, and for sonar data, it can learn the reflection intensity and echo time acoustic features of the target object, thereby converting these complex raw data into more valuable feature representations for identification and early warning.

Images