214 results about "Step detection" patented technology

The invention belongs to the field of electric power and the technical field of artificial intelligence, and particularly relates to an electric power field operation safety detection method and system based on a hybrid expert model. According to the method, the visual language model and the YOLO target detection model are combined, and efficient and accurate electric power field operation safety detection is realized under the condition of few samples through a hybrid expert model architecture. In the training stage, through multi-model cooperative training, an expert model set for different scenes and targets is constructed. In the detection stage, a two-stage detection mechanism is adopted, YOLO is firstly used for rapid preliminary detection, then VLM is used for rechecking a low-confidence target, and finally detection results are integrated to ensure the accuracy of detection. The method aims to effectively cope with complex environments and potential safety hazards in electric power field operation, safety detection efficiency and accuracy are remarkably improved, and compared with a single model architecture, the method shows wider application scenes and higher detection quality in an electric power field operation scene.

The invention provides a real-time dynamic harmonic detection method and system for a traction power supply system, and relates to the technical field of harmonic detection, and the detection method comprises the following steps: collecting a voltage signal of the traction power supply system in real time, and dynamically tracking fundamental frequency fluctuation through a phase-locked loop technology; dynamically adjusting the data window length of FFT analysis; dynamically adjusting a frequency spectrum calculation kernel of the FFT; carrying out linear compensation on the detected harmonic phase; taking the preliminarily detected harmonic parameter as an initial value, and iteratively updating the harmonic parameter until convergence by minimizing a signal reconstruction error; dynamically adjusting a sampling interval according to the fundamental frequency fluctuation, and carrying out anti-aliasing processing by adopting a filter with a variable cut-off frequency; and constructing a matrix model containing a fundamental frequency fluctuation parameter, and calculating a real harmonic parameter through a least square method. According to the method, the detection defect of the fixed window FFT during frequency fluctuation can be systematically solved, and high-reliability real-time data support is provided for harmonic resonance suppression of the electrified railway.

The invention relates to a chip surface defect detection method, system and equipment based on an improved YOLOv10 model and a medium. The method comprises the following steps: constructing a hierarchical detection model; inputting the chip surface image into a feature extraction network to obtain an initial feature map; embedding a coordinate attention module into a predetermined node of the feature fusion network, performing orthogonal direction convolutional coding on the initial feature map to generate a space attention weight, and outputting an enhanced feature map through weighted fusion; constructing a central point prediction branch in a middle layer of the feature extraction network, and outputting a defect thermodynamic diagram, a size parameter and category probability distribution in parallel based on the initial feature map; and inputting the enhanced feature map into a detection head, carrying out space alignment processing in combination with the defect thermodynamic diagram to generate a preliminary detection frame, and carrying out redundancy suppression processing to generate a detection result. According to the invention, the coordinate attention module and the central point prediction branch are introduced to carry out collaborative improvement on the YOLOv10 model, and the problem of pain points in chip surface defect identification is solved.

The invention discloses a frequency hopping signal real-time detection method and system based on FPGA + DSP. The system comprises a time frequency analysis module, a data communication module and a frequency hopping signal detection module. The method comprises the following steps: a time frequency analysis module performs analog-to-digital conversion and digital down-conversion on an intermediate frequency signal received from a radio frequency end, and performs short-time Fourier transform on a zero intermediate frequency signal by adopting a four-pipeline processing mode; the data communication module transmits the frequency spectrum information from the FPGA to the DSP through an SRIO protocol; the frequency hopping signal detection module calculates a dynamic threshold of a current frame signal according to the received frequency spectrum information, and completes preliminary detection of the signal according to a threshold value; and carrying out binarization processing and connected domain marking on the detected frequency spectrum, screening non-frequency hopping signals according to time-frequency information of a connected domain, and packaging and uploading parameter information of the detected frequency hopping signals to an upper computer. The method is simple in algorithm, low in hardware resource requirement and high in detection efficiency, and the feasibility of engineering implementation is improved.

The invention relates to the technical field of nondestructive testing, and discloses a crack detection device and method for precise automobile parts. The method comprises the following steps: carrying out ultrasonic multi-dimensional pre-scanning and acoustic impedance abnormal region identification on a to-be-detected automobile part, carrying out influence range calculation and sensitive threshold classification based on abnormal region distribution data, and obtaining stress-acoustic response time sequence data through partition grabbing force distribution and stress loading; three-dimensional stress field reconstruction and stress concentration area identification are realized; performing dual-frequency stress modulation excitation and multi-physics field synchronous detection on the stress concentration area to obtain a multi-physics field response signal, performing multi-parameter fusion calculation and multi-scale wavelet decomposition of acoustic components on the multi-physics field response signal, and performing feature dimension splicing to obtain crack feature classification data; and calculating and generating a crack detection result based on feature similarity matching and extension prediction. The precision and reliability of crack detection of the automobile parts are remarkably improved, and the damage risk in the detection process is avoided.

The invention discloses a multi-view middle school experiment step detection method and system based on differential feature enhancement, and the method comprises the steps: carrying out the preprocessing of a video, extracting original and differential video frame segments, obtaining a visual time sequence feature and a differential motion feature through a feature extractor, inputting the features into an encoder, and learning a space-time dependence relation; and carrying out feature fusion by using a cross attention mechanism, and finally outputting a classification result from the fused multi-view visual time sequence features through a decoder. The method can more accurately capture the change and action characteristics of the experiment steps, improves the precision and robustness of the classification model, can effectively solve the problems that the experiment steps in manual detection consume a large amount of manpower resources, the conventional method does not fully explore the time sequence relation, and the efficiency is low, improves the detection precision, and reduces the detection cost. Therefore, the method has relatively high application value and application prospect.

The invention discloses an electric power engineering technical-economic review method based on deep learning. The method comprises the following steps: S1, automatically capturing project data such as engineering quantity, an equipment list and material cost, and carrying out preliminary detection; s2, analyzing data by using a deep learning model, and analyzing and classifying text information through natural language processing; s3, establishing a charging benchmark rule base, and automatically checking the cost data in the technical and economic files; s4, adopting a self-adaptive material semantic matching algorithm to examine equipment and material data, and carrying out market information price matching; s5, constructing a mapping relation between the project amount and the typical cost, and generating a deviation report; s6, establishing a data center to support multi-dimensional statistical analysis and a real-time cost report; and S7, combining historical and real-time data, performing early warning through a deep learning algorithm, and generating an intelligent review report. According to the invention, automation and intellectualization of electric power engineering technical-economic review are realized, and data processing efficiency and review accuracy are optimized.

The invention discloses a multi-defect synchronous detection device for copper product processing and a use method thereof, and belongs to the technical field of copper product detection.The multi-defect synchronous detection device comprises a mounting base and a defect detection system applied to the detection device, a positioning groove is formed in the upper surface of the mounting base, and a copper pipe is placed in the positioning groove; according to the device, firstly, a pair of detection mechanisms are installed below an I-shaped pipe and an L-shaped pipe correspondingly, so that synchronous detection of the inner wall and the outer wall of the copper pipe is achieved, the problem that the inner wall and the outer wall are detected and separated in a traditional detection method is solved, and the detection efficiency and accuracy are greatly improved; the detection system can emphatically identify and evaluate the risk of the position again, the problem of product quality misjudgment caused by dependence on a single detection standard is prevented, potential safety risks existing in products are eliminated in time, the product quality level is improved, meanwhile, the use safety of the products is enhanced, and the detection system has wide application prospects and market value.

The invention belongs to the technical field of video live broadcast monitoring management, and particularly relates to a live broadcast content real-time compliance detection method and system based on deep learning, and the method breaks through the limitation that traditional text detection only depends on dominant keywords by constructing a text semantic analysis chain matched with original words and mapping sensitive words through constructing metaphor words. The identification of deep semantic violation contents such as metaphor and private obscure expression is realized; the method comprises the following steps: performing linkage verification on text violation clues and image behavior characteristics through behavior characteristic analysis of image frames before and after time positioning association of text violation based on a time positioning cross-modal collaborative detection mechanism, and forming a multi-modal judgment logic for text triggering image confirmation; a progressive detection strategy of hierarchical thresholds is designed, a first threshold is set as violation confirmation and a second threshold is set as suspected violation early warning, and real-time preliminary detection and precise secondary verification are combined, so that stepped processing from suspected violation to violation confirmation is realized, and the real-time performance and accuracy of detection are balanced.

The invention relates to the technical field of equipment energy consumption monitoring and fault diagnosis, and discloses a large-scale equipment energy consumption anomaly detection method and a storage medium. The method comprises the following steps: synchronously acquiring equipment energy consumption, state and environment data, fusing to generate multi-dimensional initial features, and performing multi-scale transformation. And performing preliminary detection on each scale feature subset to generate abnormal confidence and mode description. And screening the feature subsets according to the confidence coefficient, and dynamically selecting a corresponding detection strategy according to the mode description. And performing deep feature extraction on the feature subset based on the selected strategy, inputting the obtained high-dimensional feature vector into a corresponding anomaly evaluation network, calculating a standard state matching degree, comparing with a dynamic threshold to generate a fine-grained judgment result, and finally finishing positioning and attribution analysis in combination with anomaly description to form a detection report. According to the method, efficient and accurate large-scale equipment energy consumption anomaly detection is realized.

The present invention provides a method for improving the accuracy of event detection based on multi-model fusion, comprising the following steps: performing multimodal data preprocessing on a monitoring video stream to generate a standardized feature matrix; performing feature-level fusion of the outputs of a YOLOv5 improved model, a ResNet Transformer hybrid model, and an LSTM spatiotemporal analysis model through an adaptive weighted fusion module; performing multidimensional event detection based on the fused feature matrix to generate a preliminary detection result set; establishing a three-level logical judgment mechanism to verify the confidence of the detection result; and outputting the verified final event detection result to a monitoring system. The present invention can improve the accuracy of event detection and is suitable for monitoring event detection in complex environments.

The invention relates to a target object detection method and device, computer equipment and a readable storage medium, and the method comprises the steps: inputting a to-be-detected image into a target detection model for preliminary detection, and obtaining a preliminary detection result of a target object; inputting the preliminary detection result into a category discrimination model for prediction to obtain a probability score of a non-target object; comparing the preliminary detection result with a pre-stored feature queue to obtain a similarity score; the feature queue is used for storing non-target features in the historical detection result; correcting the probability score based on the similarity score to obtain a target score; determining a false detection result from the preliminary detection result based on the target score; and the false detection result is eliminated from the preliminary detection result to obtain a final detection result, so that the problem of poor target object detection effect is solved, the false detection recognition accuracy is improved, and the quality of the target detection result is improved.

The invention discloses a vacuum cavity pressure maintaining test and leak rate rapid estimation method and system, and the method comprises the steps: achieving the test preparation of multiple sections through a public metering manifold and a multi-way isolation valve, and building a section volume database; the pressure and temperature data are synchronously collected in a fixed sampling period by exhausting air to preset pressure and isolating a pump end; obtaining an equivalent pressure change rate through temperature normalization and deflation baseline deduction in combination with filtering and regression algorithms, and performing conversion to obtain a standardized leak rate; step-by-step detection is carried out on each functional section of the cavity through a segmented bipartite isolation strategy and volume correction, and rapid positioning of a suspicious leakage section is realized; through double-threshold determination of an early warning threshold and an unqualified threshold, a detection report is automatically generated and written into a historical database, and a detection-determination-maintenance closed loop is formed. The method can improve the precision of leak rate detection, and is suitable for vacuum coating equipment, semiconductor process equipment and industrial vacuum treatment devices.

The invention discloses a complex scene target detection method based on a model dynamic joint mechanism, and the method comprises the steps: carrying out the preliminary detection through an RT-DETR model, retaining more potential targets through dynamic threshold adjustment, and projecting a generated detection frame to a feature space of an improved YOLOv12 model through dual-mode feature mapping; the improved YOLOv12 model integrates an SEAM attention module and a rejection loss function so as to enhance feature representation and positioning compactness of an occluded target. Then, a model joint mechanism is adopted to process preliminary results of the two models; through difficult case mining and online learning, missing detection targets are supplemented, and the RT-DETR model is optimized; and intelligently fusing the detection results of the two models through dynamic weight distribution based on scene complexity and hierarchical fusion of a decision tree. And finally, post-processing is carried out by using an improved non-maximum suppression algorithm, and mistaken deletion is reduced. According to the method, the problems of missing detection, false detection and inaccurate positioning of the target in a complex scene are effectively solved, and the recall rate and the accuracy rate of detection are remarkably improved.

The invention discloses an industrial control network flow abnormity real-time detection method based on a P4 programmable switch, and the method comprises the steps: extracting network flow characteristics related to a timestamp in a data plane in real time, carrying out high-line-speed preliminary detection on the traffic by using an industrial control network traffic anomaly detection algorithm based on a threshold rule to discover suspicious traffic, and uploading a preliminary detection result to a control plane; the control plane obtains suspicious traffic characteristic data based on the periodically maintained key traffic characteristic data and uploads the suspicious traffic characteristic data to the detection plane; the detection plane performs fine-grained detection on the suspicious traffic feature data by using a traffic anomaly detection model based on machine learning to obtain a final detection result; and when the difference between the final detection result and the preliminary detection result exceeds a given value, generating a new preliminary detection threshold rule based on an optimization algorithm or deep reinforcement learning, and updating to optimize the detection accuracy. According to the method, the real-time detection delay of the flow abnormity of the industrial control network can be greatly reduced, and the detection accuracy is remarkably improved.

The invention discloses an abnormal data processing method and system for a virtual game, and relates to the technical field of big data processing. Receiving an interaction data frame sequence uploaded by the terminal equipment, and analyzing to obtain an input equipment sampling point set and a displacement coordinate vector; performing feature extraction on the sampling point set of the input equipment to obtain a high-dimensional feature vector; mapping the high-dimensional feature vector to a preset behavior feature space, calculating a multi-dimensional statistical distance of the high-dimensional feature vector in the behavior feature space, and generating a first abnormal confidence coefficient; calling space mapping data corresponding to the virtual scene; performing discretization ray stepping detection in the space mapping data based on the displacement coordinate vector to generate a second abnormal confidence coefficient; performing fusion calculation on the first abnormal confidence coefficient and the second abnormal confidence coefficient based on a dynamic weighting algorithm to obtain a comprehensive abnormal score; and judging abnormal data through the comprehensive abnormal score, and executing a state rollback operation for the virtual object.

The present invention relates to a leakage detection method for a distribution box, and the technical field of leakage of distribution boxes includes the following steps: S1: real-time acquisition of current signals of each line in the distribution box, using a current transformer or a Hall effect sensor to acquire data; S2: using an adaptive sensitivity adjustment algorithm to automatically adjust the threshold of leakage detection according to real-time environmental parameters and historical data; S3: fast Fourier transform processing of the collected data; S4: when a suspected leakage event is initially detected, a secondary verification procedure is immediately started; S5: after confirming the leakage, the power supply is quickly cut off and an alarm message is sent to a user terminal. The present invention maintains optimal sensitivity and accuracy under different working conditions, avoiding the problem of missed detection or false alarm that may be caused by the traditional fixed threshold method in a complex environment.

The invention discloses a drainage pipe network data cleaning and intelligent repairing method fusing multiple models. The method comprises the following steps that S1, original monitoring data flow is collected; s2, carrying out the preliminary detection and elimination of the abnormity based on IQR; s3, performing upstream and downstream multi-source feature extraction and XGBoost weight analysis; step S4, multivariable depth autoregression prediction based on PatchTST is carried out; s5, generating and publishing continuous high-quality data; according to the method, the key technical problems of various abnormal types, incomplete abnormal detection, low cleaning and repairing precision, poor data continuity and trend consistency, high system integration and real-time processing difficulty and the like commonly existing in the actual collection and transmission process of the online monitoring data of the drainage pipe network are solved.

The invention relates to the technical field of steel wire rope defect identification and positioning, in particular to a steel wire rope defect identification and positioning system and method based on real-time target detection. According to the technical scheme, the system comprises a sensor array, a synchronous detection execution mechanism, an NPU neural computing unit, a multi-modal weight management unit, a model library, a multi-core acquisition mainboard, a communication module and a power supply module. On the basis of a multi-modal real-time target detection method, the multi-modal cooperative training is realized by acquiring surface and internal key features of the steel wire rope through a multi-sensor array. The method not only improves the data characterization capability, optimizes the problems of model lightweight and boundary fuzziness, but also designs a self-defined dynamic allocation method, effectively improves the accuracy and precision of defect identification and positioning, realizes effective closed loop of model training, and provides a more efficient and reliable technical scheme for steel wire rope operation and maintenance enterprises.

An object of the present invention is to obtain a step detection device that can accurately detect a step without being affected by weather, environmental conditions, and the like. A step detection device (2) according to the present invention is a step detection device 2 that detects a step (20) in a case where a tire (10) of a vehicle (100) runs over the step, including: a plurality of strain detection elements (1) disposed on a tire inner circumferential surface of the tire at predetermined intervals in a circumferential direction of the tire; and a signal processor (4) that processes detection signals of the plurality of strain detection elements. The signal processor detects the step based on detection signals of two or more adjacent strain detection elements among the plurality of strain detection elements.

The present invention relates to the field of real-time data filtering technology, and in particular to a real-time data filtering method and system with multi-dimensional feature fusion. The method comprises the following steps: obtaining data transmission logs and extracting original transmission data features, parsing data frame field structure information; identifying multi-dimensional abnormal coupling of data transmission based on field time gradient change rate and dynamic offset status; further detecting abnormal dilution of transmission data, predicting the data transmission complexity exceeding limit state, and detecting the overload condition of data filtering chip structure; then evaluating the attenuation trend of data real-time filtering efficiency, and optimizing the data real-time filtering path accordingly, completing real-time filtering processing of data transmission complexity exceeding limit condition, and outputting filtering results; the present invention achieves higher credibility of data filtering by filtering data in real time.

The invention relates to the technical field of photoelectric measurement, and discloses a data acquisition and processing method and system for photoelectric measurement. The method comprises the steps of obtaining photoelectric data of a target area, performing preprocessing to realize multi-source data space-time registration, and extracting a space-time fusion photoelectric signal; spectral features are obtained according to the signal, a preliminary detection area is divided, and fine target segmentation and feature optimization are completed; constructing a feature extraction model based on time-frequency analysis to perform target classification, and generating an acquisition result; energy distribution changes are obtained through the multi-period data, and dynamic calibration parameters are generated. The system comprises a memory and a processor, and executes corresponding programs to realize the steps. The method solves the problems of multi-source data fusion, target identification, dynamic calibration and the like in the prior art, improves the precision and reliability of photoelectric measurement, and is suitable for photoelectric measurement scenes in multiple fields.

The invention discloses an image linear target high-precision parameter detection method, and relates to the technical field of image recognition, and the method comprises the steps: carrying out the super-resolution reconstruction of a to-be-detected original image based on a perception optimization SRGAN algorithm, and obtaining a super-resolution reconstruction image; using an adaptive Canny edge detection algorithm to obtain contour information in the ROI region of the super-resolution reconstruction image; based on contour information in the ROI region of the super-resolution reconstruction image, performing linear fitting through an RANSAC algorithm for dynamically adjusting the number of iterations and a tolerant error to obtain a preliminary detection result; and constructing a multi-error-source joint model, and performing error compensation and quantization on the preliminary detection result based on the multi-error-source joint model to obtain a final straight line detection result, thereby completing detection. The method effectively improves the accuracy and stability of straight line detection in a complex environment, and is especially suitable for image processing tasks with low contrast, high noise or significant distortion.

The invention provides a signal detection method for a multi-antenna orthogonal time-frequency-space sparse code division multiple access system, which is characterized in that dimensionality reduction is carried out on time domain and time delay Doppler domain signals by utilizing sparsity of a sparse code division multiple access codebook, and prior distribution of the time domain signals is approximated by using Gaussian distribution. Preliminary detection is realized by acquiring cavity distribution of a signal in a time domain, and the posterior probability of the signal is calculated by using a sparse code division multiple access codebook in a time delay Doppler domain. Based on a transformation relation in a multi-antenna orthogonal time-frequency-space sparse code division multiple access system, information exchange between a time domain and a time delay Doppler domain is realized. The sparse structure of the time domain matrix in the multi-antenna orthogonal time-frequency-space sparse code division multiple access system is utilized, and the complexity of matrix inversion in the algorithm is reduced. By using the method, the complexity of algorithm implementation can be reduced while the signal detection performance is ensured by utilizing the dual sparsity of the multi-antenna orthogonal time-frequency-space system and the sparse code division multiple access codebook.

The invention discloses a dual-frequency-band weak target detection method based on Hough transformation, and the method comprises the following steps: S1, selecting wave beams of two frequency bands of a radar, carrying out the alignment of the wave beams of the two frequency bands in space and time, and scanning a target region; s2, acquiring echo data of wave beams of two frequency bands, and respectively setting an initial detection threshold to carry out target initial detection; s3, performing wide constraint detection on the two groups of echo data subjected to preliminary detection in the step S2 by adopting pre-installed information, and screening out detection units conforming to motion feature constraints; s4, performing dual-frequency fusion detection on the detection units screened out in the S3 based on Hough transformation, and determining a target point in a parameter domain transformation spectrum after Hough transformation; s5, for the target point determined in the S4, using Hough inverse transformation to output each dimension measurement value; the method is high in efficiency, and the detection performance is stable when the external interference noise condition changes greatly.

The invention discloses a synchronous detection system for surface density and wall thickness uniformity of a composite bulletproof helmet, which belongs to the technical field of bulletproof helmet detection and comprises a detector cooperative control module, a two-parameter synchronous acquisition module, a data preprocessing module, a two-parameter cooperative constraint module and an intelligent modeling output module. Through the integrated design that the detection unit integrates the 3D scanning detector, the microwave thickness measuring probe and the beta-ray density detector, and in cooperation with the synchronous triggering and double-channel signal acquisition mode of the double-parameter synchronous acquisition module, synchronous acquisition of the surface density and wall thickness data of the helmet is realized; the problems that traditional step-by-step detection is low in efficiency and poor in data relevance are solved, the detection efficiency and the accuracy of quality judgment are greatly improved, meanwhile, a detection report containing process optimization suggestions is output, the problems that traditional detection results are not visual, and connection with the production process is not smooth are effectively solved, and the product quality is improved. And accurate guidance can be provided for mass production quality control and process adjustment.

The invention relates to the technical field of data anomaly identification and bridge safety monitoring, in particular to a burr and step detection method, a data anomaly identification method and a bridge health monitoring method. The burr anomaly detection method provided by the invention comprises the following steps: firstly, dividing a time sequence data set into a plurality of data intervals with equal numerical value widths; taking the mean value of the data interval containing the maximum number of numerical values as a reference value; and calculating a difference value between a numerical value at each time point of the time series data and a reference value as a fluctuation at the time point, and counting a continuous time point of which the fluctuation is greater than a burr threshold and identifying the continuous time point as a burr. The reference value is calculated based on the trend distribution, and then the burr is judged according to the difference value between the sampling value and the reference value. Through the consideration of the trend distribution, the influence of sudden factors is well removed, data analysis is carried out only by considering the influence of long-term factors, a better burr detection effect is obtained, and the burr detection precision is improved.

The invention relates to the technical field of large models and target detection, and discloses an intrusion target detection method based on a large model, and the method comprises the steps: obtaining a to-be-detected image, and inputting the to-be-detected image to a first large model, so as to obtain a preliminary analysis result used for representing the content of the to-be-detected image; determining a target detection algorithm from a plurality of preset target detection algorithms according to the preliminary analysis result; running a target detection algorithm, and processing the to-be-detected image to obtain a preliminary detection result; and inputting the preliminary detection result into the second large model to verify the preliminary detection result so as to generate a final detection result. According to the method, the flexibility and the automation level of the detection process can be remarkably improved, and a single system can cope with more diversified detection tasks, so that the universality of the system is enhanced. And the cognitive and inference capabilities of the artificial intelligence large model are utilized to identify and correct possible misjudgments, so that the accuracy and reliability of the final output result are improved, and the probability of false alarm or missing alarm is reduced.

The invention discloses a radar weak maneuvering target tracking-before-detection method based on dynamic programming and Hough circle detection. The method comprises the following steps: constructing an accurate motion model and an echo measurement model in a target adjacent situation for detection and tracking in a complex multi-target scene; according to the DP-TBD method based on sequential deletion, preliminary detection and target point extraction of adjacent targets are realized, and false detection and missing detection caused by clutter interference are avoided. And the target measurement point is corrected by combining a Hough circle detection technology, so that the accuracy and the reliability of a detection result are greatly improved. The Hungary algorithm is adopted to solve the matching and distribution problems among multiple targets, detection and tracking under the condition of high target proximity are achieved, and accurate association of the multiple targets is achieved. The method not only improves the precision and robustness of target detection and tracking, but also provides an efficient solution for multi-target tracking in a complex radar scene.

The invention discloses a CC attack gateway layer detection method and system based on an intelligent sliding time window, and belongs to the technical field of network security. The method comprises the following steps: analyzing the HTTP traffic of a gateway layer in real time, and extracting multi-dimensional features of the traffic and user behaviors; performing real-time aggregation and multi-dimensional analysis on the features by using a sliding time window mechanism; based on a risk score accumulation mechanism, matching the aggregation result with a pre-configured rule base to obtain a preliminary detection result and a risk score; when the initial detection is negative, further performing verification detection by using a pre-trained machine learning model; and finally, according to the risk score and a machine learning result, judging an attack and executing a progressive disposal measure. Through streaming processing, rule and model cooperation, intelligent degradation and other mechanisms, the problem that detection precision and processing performance cannot be achieved at the same time in a gateway scene in the prior art is effectively solved, and high-precision CC attack defense under the requirements of high concurrency and low delay is achieved.