281results about How to "Improve recognition rate" patented technology

The invention relates to the technical field of data processing, and discloses a double-end fault positioning method based on signal enhancement, which is used for improving the efficiency of double-end fault positioning based on signal enhancement. The method comprises the following steps: acquiring and preprocessing electrical signals at two ends of a power transmission line to obtain preprocessed synchronous electrical signals; performing adaptive enhancement processing on the preprocessed synchronous electrical signal to obtain an enhanced fault feature signal; performing time-frequency analysis and feature extraction on the enhanced fault feature signal to obtain a fault feature vector and a key parameter; performing machine learning analysis on the fault feature vectors and the key parameters to obtain a fault type identification result and initial fault position estimation; performing multi-algorithm fusion analysis on the preliminary fault position estimation to obtain an accurate fault position coordinate; and performing geographic information mapping and field verification on the accurate fault position coordinates to obtain a fault positioning result and system optimization data.

The invention provides a high-density RFID tag group scanning system based on a multi-antenna array and beam forming, and relates to the technical field of the Internet of Things. The system comprises a multi-antenna array, an RFID reader-writer, a beam forming unit, a signal processing unit, a feedback unit and a data management unit. Through a digital beam forming technology, the system dynamically adjusts the direction and width of a beam, and achieves the precise scanning of a target area. And in combination with real-time signal feedback, a density peak clustering algorithm and a gradient descent algorithm are adopted to optimize beam parameters, so that the recognition efficiency is improved. The system further utilizes an anti-collision algorithm to dynamically adjust frame parameters, and performs logic grouping based on spatial positions and type characteristics of tags, so that signal collision and missed reading rates are remarkably reduced. Meanwhile, the data management unit stores and outputs the label data in a classified mode, and intelligent management capacity is provided for multi-scene application.

The invention relates to the technical field of path planning, in particular to a pipeline robot path planning and detecting method based on multi-data fusion, which comprises the following steps: collecting data from vision, ultrasound, magnetic flux leakage and laser radar, performing timestamp alignment and space coordinate correction on the data, performing data fusion and checking the consistency of the data, an environmental data set is generated. According to the method, by constructing the three-dimensional pipeline model, not only is visual representation of the pipeline provided, but also the understanding of the internal structure of the pipeline is enhanced, so that a maintenance team can make a repair strategy more effectively, and the robot is allowed to check in the most effective sequence through the access priority determined through cost-benefit analysis and the optimized access path; the detection time is shortened, the operation cost is reduced, instant fault diagnosis and quick response are ensured, the detection accuracy and the operation efficiency of the robot are remarkably improved, the pipeline detection efficiency and accuracy are remarkably improved, manual intervention is reduced, and the maintenance cost is reduced.

The invention relates to the technical field of medical systems, in particular to an intelligent gastrointestinal postoperative intestinal obstruction diagnosis and intervention decision-making system and method, and the system comprises a data preprocessing module, a multi-modal data fusion module, a gastrointestinal function recovery model construction module, a self-adaptive progressive reinforcement learning optimization module AHRL and a diagnosis and intervention recommendation module. The self-adaptive progressive reinforcement learning optimization module optimizes an intervention strategy layer by layer through a five-layer progressive algorithm from the steps of dynamic adjustment of feature weights, recursive state propagation, layered accumulation optimization, symptom similarity matrix generation, progressive weight reverse feedback and the like, deep fusion of key symptoms by multi-modal data is achieved, and the accuracy of the key symptoms is improved. And the intervention scheme is updated in real time according to the patient state change. In the intervention decision recommendation module, the system generates personalized diagnosis suggestions based on the optimized strategy, carries out real-time risk assessment, and provides continuous and dynamic intervention support for the patient.

The invention provides a textile defect type identification method, and relates to the technical field of textile defect identification, and the method specifically comprises the steps: collecting original images of a plurality of textile samples, and carrying out the detailed marking; adjusting the original image to 224 * 224 pixels, constructing a CNN model, extracting feature maps of different convolutional layers, and generating a fusion feature vector in combination with the weights of the feature layers; performing pooling processing on the fused feature map to generate a single-channel space attention map, and obtaining a weighted feature map in combination with a channel attention weight; constructing a segmentation model, outputting a mask after training, and identifying a normal region and a defect region; analyzing and identifying a defect area and extracting related features through a connected component to form a feature set; and a comprehensive scoring coefficient is constructed based on the features, so that the textile defect type can be accurately identified. According to the method, the recognition capability on fine defects can be effectively enhanced, and finer deviations can be captured, so that the recognition rate is remarkably improved.

The invention relates to the technical field of infrared target detection, in particular to a light-weight real-time infrared target detection method and system for an unmanned ship, and the method comprises the steps: collecting infrared data through an infrared thermal imager carried by the unmanned ship, and carrying out the preprocessing, marking and dividing. And a lightweight GLAF-MobileNetV4 network is adopted as a backbone of the detection method, so that the real-time performance is improved, and model parameters are greatly reduced. Wherein the lightweight global and local adaptive fusion attention can ingeniously integrate global and local information, and local information fusion is adjusted and optimized through dynamic weight. Meanwhile, the designed dynamic adaptive multi-scale feature fusion encoder adaptively adjusts the fusion proportion according to the importance of the multi-scale features by using an attention guidance mechanism. The method is simple and convenient to deploy, improves the target detection precision and real-time performance of the unmanned ship under the low-light condition, and is remarkable in detection efficiency.

The invention discloses a laser welding spatter real-time recognition method based on OCT keyhole depth measurement. The method comprises the steps of S1, keyhole depth data collection based on OCT measurement and data set construction and S2, construction training of a deep learning model and spatter defect recognition. According to the method, an SD-OCT measurement technology is combined with a composite deep learning model based on 1DCNN-BiLSTM splash defect recognition, a Dropout technology and adaptive moment estimation are adopted to prevent model overfitting, and a TPE algorithm is adopted to carry out hyper-parameter optimization on the model, so that the trained 1DCNN-BiLSTM splash defect recognition model can accurately predict the splash state in the dynamic laser welding process, and the accuracy of the dynamic laser welding process is improved. Manual intervention is not needed, and the laser welding spatter defect can be rapidly and accurately recognized.

The invention relates to the technical field of target detection, in particular to an aquatic organism multi-target detection method and system based on YOLO. The method comprises the following steps: carrying out aquatic organism image acquisition on a target water area to obtain an aquatic organism image; performing resolution enhancement on the aquatic organism image to obtain an enhanced aquatic organism image; performing multi-branch hierarchical feature extraction on the enhanced aquatic organism image to obtain a residual enhanced biological feature image set; performing cross-scale feature fusion on the residual enhanced biological feature image set to obtain a fused aquatic organism feature image; performing historical detection data acquisition on the target water area to obtain an aquatic organism historical detection data set; and performing anchor frame generation on the fused aquatic organism feature map based on an aquatic organism historical detection data set to obtain an initial anchor frame set. The defects in aquatic organism monitoring and recognition in the prior art are effectively overcome, and the accuracy, efficiency and practicability of detection are improved.

The invention discloses an oracle bone text image recognition method, device and equipment based on a multi-modal large model, and a medium, and relates to the technical field of computer vision. The oracle bone text image recognition method comprises the following steps: firstly, extracting an oracle bone text single character image from an OBIMD data set, and retrieving basic information from an original oracle bone platform to form a single character data set; thirdly, generating a first dialogue data set containing information such as character paraphrases by utilizing a large language model, and finely adjusting the Qwen2VL model to obtain an initial recognition model; and then, performing second fine tuning through the position information dialogue data set to obtain a positioning sequence recognition model which can recognize characters and complete positioning labeling. Furthermore, a modern Chinese translation dialogue data set is generated, third fine tuning is carried out on the model, a semantic recognition model is obtained, and characters can be combined into sentences and translated. And finally, inputting a to-be-recognized oracle-bone image, respectively outputting character identifiers, paraphrases or combined sentences by the model according to a single character or multiple characters of the image, and translating the sentences into modern Chinese.

The invention discloses an overhead line system cable state real-time monitoring method and sensing system based on distributed optical fiber sensing. The Mach-Zehnder interference and phi-OTDR fusion type distributed optical fiber sensing system comprises a first light source, a second light source, a Mach-Zehnder interferometer, a phi-OTDR and a data acquisition and processing module. Raman amplifiers are arranged at the two ends of the Mach-Zehnder interferometer, so that the sensing distance is prolonged; the optical fiber sensing system extracts an interference optical signal and a backward Rayleigh scattering optical signal corresponding to a fault event after phase modulation when the fault event occurs; then processing the extracted backward Rayleigh scattering light signal to obtain intensity change information of a fault event, and demodulating the position of the fault event; and sending the two-dimensional optical signal into a pre-trained Swin Transform network model to extract a time domain feature, a frequency domain feature and a spatial domain feature, and identifying the type of a fault event through a classifier.

The invention relates to the technical field of soil humidity prediction, and discloses a high-precision soil humidity prediction method based on a space-time convolutional neural network coupling model, which comprises the following steps of: deploying soil humidity detection equipment with a GPS (Global Position System) function to obtain measured soil humidity data of different depths of a target area, and obtaining a satellite remote sensing image of the target area; after the image is corrected, humidity data of the whole area is obtained, and errors of remote sensing humidity data are corrected through sensors and earth surface factors. According to the regional soil humidity dynamic monitoring method based on multi-source data fusion and deep learning, limitation caused by a single data source is avoided by fusing a soil humidity sensor carrying a GPS and satellite remote sensing data, and the monitoring precision of soil humidity is improved by introducing earth surface factors to correct remote sensing data. The result shows that the early warning forecast accuracy is not lower than 90%. Utilizing a machine learning model to expand measured data of the humidity sensor, and predicting humidity change of a certain time window in the future.

The invention discloses a pedestrian recognition method based on video data, and belongs to the field of data processing, and the method comprises the steps: a cloud database sets a plurality of pedestrian IDs and data representations corresponding to the pedestrian IDs; obtaining video data, electing a key frame, and identifying whether the contour feature of the key frame is missing or not; if the key frame is not missing, obtaining a first representation and directly matching the data representation, if the key frame is missing, marking the key frame as a sensitive image, taking N coherent frames associated with the sequence information of the sensitive image to obtain N extended images, performing grid division on the extended images into a plurality of local image blocks to obtain a first image set and a second image set, integrating the feature vectors of the sensitive images into a plurality of vector sets, fusing the first image set and the second image set with the plurality of vector sets according to fusion attributes to obtain a second representation, and matching through the second representation; a video data processing mode is expected to be optimized on the basis of existing pedestrian re-identification, and the problem of feature fusion deviation of the video data in pedestrian re-identification is solved.

The invention discloses a forest fire audio recognition method based on a BP neural network, a medium and a product in the technical field of forest fire recognition. The method comprises the following steps: acquiring forest fire audio data to be detected; extracting a Mel-frequency cepstrum coefficient (MFCC) based on the frequency domain features of the forest fire audio data to be detected; inputting the Mel-frequency cepstral coefficient MFCC into a pre-trained forest fire audio recognition model based on a BP neural network, and outputting a forest fire category; wherein the forest fire category comprises a forest fire and a non-forest fire. Compared with a traditional visual monitoring means, the method can cover a wider forest area, in the same monitoring range, in addition, model training based on the BP neural network can efficiently process a large amount of audio data and rapidly output an identification result, the efficiency and accuracy of forest fire identification are further improved, and the method is suitable for popularization and application. The problems of limited coverage range, high cost and low accuracy in the existing forest fire identification technology are solved.

The invention discloses a weather early warning method and device based on radar data and meteorological data and storage equipment, and the method comprises the following steps: S1, obtaining high-temporal-spatial-resolution wind field data, an inversion wind shear field and a visibility field through a high-precision high-temporal-spatial-resolution coherent laser radar; s2, detecting atmospheric parameters of an overhead area of the airport in real time to form airport auxiliary observation data; s3, based on the wind shear field, the visibility field and the airport auxiliary observation data, artificial intelligence machine learning is used for data processing, and a forecasting scheme is formed. Compared with the prior art, the method is based on the high-precision and high-temporal-spatial-resolution coherent laser radar, atmospheric parameters of the area above the airport are detected in real time, high-temporal-spatial-resolution wind field data are obtained, and the prediction accuracy is improved. And inverting a wind shear field and a visibility field, and processing by using methods such as artificial intelligence machine learning and the like to form a forecasting scheme.

The invention provides a deep reinforcement learning mobile target defense system and method based on intrusion detection, and mainly solves the problem that the existing method consumes too many resources and cannot perform targeted defense for frequent IP hopping in a complex network environment. The scheme comprises the following steps: 1) constructing an SDN network topology and monitoring network traffic in real time by using an SDN controller; 2) constructing an intrusion detection model based on CNN + LSTM and introducing an attention mechanism, and identifying potential threats and abnormal behaviors; 3) initializing IP resources and maintaining a dynamic mapping table and an IP use record; 4) constructing an IP jump model based on deep reinforcement learning and defining an action and reward function; 5) completing model optimization through strategy updating; and 6) moving target defense is realized by using the optimal IP hopping strategy. The method can improve the recognition rate of complex network attacks, reduces the network security risk, reduces the delay of data processing, and effectively optimizes the utilization rate of network resources.

The invention provides an expander abnormity identification method and system based on a transformer substation patrol image, electronic equipment and a computer readable storage medium. The method comprises the following steps: acquiring the transformer substation patrol image; using a YOLOv11-Marge model used for expander position detection to carry out expander area detection on the transformer substation patrol image, and identifying and extracting an expander area image; performing expander defect detection on the expander area image by using a YOLOv11-Marge model used for expander defect detection; wherein the YOLOv11-Marge model used for expander defect detection is obtained after training of an expander defect expansion sample, and the expander defect expansion sample is obtained after expansion of the expander defect sample is carried out on the expander area image by using a diffusion model; and identifying and outputting the abnormal category of the expander. According to the method, the technical effects of effectively improving the defect recognition rate of the expander, reducing the missed recognition rate and enhancing the generalization performance of the model are achieved.

The invention discloses an operation stage identification method based on time-frequency feature fusion, which is realized based on a constructed TFRCDLFormer network, a multi-level frequency domain feature module performs multi-layer cavity convolution RCDL processing on an input picture, then discrete wavelet transform is performed to obtain multi-level frequency information containing high frequency and low frequency, and the multi-level frequency information is subjected to time-frequency feature fusion. Processing the frequency information and then carrying out discrete wavelet inverse transformation; the multi-scale time domain feature module carries out RCDL processing on an input picture, then carries out average pooling and up-sampling processing to obtain feature information of different scales, and carries out comprehensive processing on the feature information of each scale by adopting RCDL and Transform methods; and the multi-level frequency domain feature enhancement module carries out feature fusion on the multi-level frequency domain features and the multi-scale time domain features, carries out frequency domain feature enhancement on the fused features, and outputs and obtains an operation stage identification prediction result after passing through a full connection layer. And the recognition speed and precision are effectively improved.

The invention discloses a human eye-imitating three-dimensional point cloud data compression method, which comprises the following steps of: acquiring original point cloud data, and performing Gaussian filtering denoising processing on the original point cloud data; performing region-of-interest extraction on the denoised point cloud by using a human eye-imitated target-of-interest extraction algorithm; performing high-resolution retention on the extracted region of interest by using a human eye-imitated interested target compression algorithm, and performing non-uniform compression on a non-interested region; and finally, displaying a compression result by using a display program. The region of interest is obtained according to the human eye retina attention perception mechanism, then high resolution is performed on the region of interest, non-uniform redundancy compression is performed on the non-region of interest, the problem of low efficiency of fixed resolution when a current point cloud data processing algorithm processes the region of interest and the non-region of interest is solved, and the accuracy of point cloud data processing is improved. Therefore, the quality of the three-dimensional point cloud data is effectively improved, and the recognition precision and speed in the field of three-dimensional target recognition are improved.

The invention discloses a multi-signal joint fault diagnosis method and system for a wind power variable pitch bearing. The multi-signal combined fault diagnosis method for the wind power variable pitch bearing comprises the following steps: signal acquisition: acquiring a vibration signal, a strain signal, a temperature signal and a sound signal during operation of the variable pitch bearing; signal denoising: transmitting the collected vibration signal, strain signal, temperature signal and sound signal to a signal denoising module for variational mode decomposition; signal processing: transmitting at least one of the denoised vibration signal, strain signal, temperature signal and sound signal to a signal processing module to obtain a corresponding two-dimensional time-frequency image; and multi-signal combined fault diagnosis: transmitting the plurality of one-dimensional time-domain signals and the at least one two-dimensional time-frequency image to a combined fault diagnosis module, and carrying out combined fault diagnosis through a plurality of one-dimensional convolutional neural networks and a two-dimensional convolutional neural network, thereby realizing multi-signal combined fault diagnosis and improving the recognition rate of fault diagnosis.

The invention discloses a cart right turn must stop detection method based on a DeepSort algorithm, and belongs to the technical field of computer vision target detection and tracking, and the method comprises the following steps: inputting a video stream, preprocessing the video stream, and inputting the preprocessed video frame into an improved YOLOV7 neural network for preliminary target detection; performing target tracking on the detected suspicious target through a DeepSort algorithm; and judging whether the suspicious target does not stop when turning right or not according to the change of the pixel gray value of the suspicious target, thereby realizing the detection and identification of the large vehicle which must stop when turning right in a complex traffic environment. According to the method, the improved YOLOV7 neural network and the DeepSort algorithm are applied, the peccancy detection accuracy is improved, the peccancy detection speed is increased, the method is applied to the problems of vehicle overlapping, reflection and the like in the complex traffic environment, and intelligent detection and recognition of the cart turning right and stopping in the complex environment are achieved.

The invention provides an identification and early warning method and system for dangerous behaviors of inland ships. The method comprises the following steps: acquiring infrared monitoring video image data of an inland river water area to obtain a standard data set; then, performing ship main shaft identification according to the standard data set, performing bow and stern identification according to a ship gray value, and performing ship motion direction judgment; then, according to a method based on sample generation, a large number of ship target tracking virtual samples are generated; then, inputting the generated ship target tracking virtual sample into a DVIS model tracking training framework to realize tracking training, and obtaining a tracking model; and finally, according to the obtained tracking model and input real sample data, analyzing motion track data and a motion mode of the ship, and carrying out identification and early warning on dangerous behaviors of the ship. According to the method, tracking of ship behaviors under the complex ship shore environment condition is achieved, early warning is conducted on the ship dangerous behaviors, and the recognition rate of the ship navigation dangerous behaviors is further increased.

The invention relates to the technical field of raw silk inspection, in particular to a method for simultaneously identifying feed silkworm rearing raw silk, mulberry leaf silkworm rearing raw silk and fresh cocoon raw silk / dry cocoon raw silk, which comprises the following steps: grinding the feed silkworm rearing raw silk, mulberry leaf silkworm rearing raw silk, fresh cocoon raw silk and dry cocoon raw silk from the same source, adding equal amount of hydrochloric acid after grinding, mixing and reacting, stirring by using a glass rod, and standing for 24 hours; then carrying out ultrasonic treatment and centrifugation, and removing supernate; removing the supernatant, adding the same amount of deionized water, stirring, washing, standing, pouring out the supernatant, repeating for several times, centrifuging, taking the precipitate, drying, grinding and crushing to obtain an extract; and determining the obtained extract by adopting a stable isotope ratio mass spectrometer, and judging the silkworm rearing raw silk of the feed, the silkworm rearing raw silk of the mulberry leaves, the fresh cocoon raw silk and the dried cocoon raw silk according to the N content and the 15N relative content. The method for detecting the total N content and the < 15 > N relative content in the raw silk is reliable, high in recognition rate, visual and reliable in experimental result and high in accuracy.

The invention provides a path sorting question parsing method based on information extraction, and belongs to the technical field of natural language processing. From the perspective of reducing the labor cost, the method considers that the same question may have different questioning modes, related sub-paths are extracted according to a central entity in combination with a knowledge base, and path sorting is performed by using semantic similarity, so that a question parsing task is completed. The method mainly comprises the steps of candidate entity recognition, entity linking, question classification, candidate path generation, candidate path sorting and answer retrieval. The method comprises the following steps: firstly extracting a central entity in a question, then querying a knowledge base according to an extracted candidate entity set, reserving a subset with a query result, performing question classification, then performing subgraph extraction by taking the candidate entity as a center, generating candidate paths, finally sorting the candidate paths, and obtaining a path with the highest score as an answer query path. And constructing a query statement for query and returning an answer.

According to the electric bicycle digital identity information and multi-video identification data matching method provided by the invention, one or more key point locations are set in the monitoring target area, the radio frequency identification equipment is set at each key point location, the radio frequency identification equipment does not need to be installed at other point locations, and only a low-cost pure video camera needs to be installed to capture video images, so that the monitoring efficiency is greatly improved. The digital identity information of the electric bicycle in the monitoring target area can be effectively identified by sharing the data acquired by the approval reading equipment of the key point location, so that the implementation cost of the scheme is greatly reduced.

The invention relates to the technical field of text detection and recognition of images, in particular to a text detection and recognition method for medical bill images. The method comprises the following steps: S1, performing feature enhancement on a medical bill image through a feature enhancement module, and reconstructing a table structure of the medical bill image; s2, after the table structure is reconstructed, feature fusion of the medical bill image is further carried out, and a pre-detection table image is formed; s3, integrating a pre-detection table image and a visual language pre-training model MDETR to enhance feature expression; s4, recognizing a text region from the pre-detection table image by using a connection time sequence classification CTC method improved through an Attention mechanism; and S5, further performing deep analysis on the detected text region through a fusion model of the SVTR network and the PP-HGNet. According to the invention, through feature enhancement, table structure reconstruction, feature fusion and integration with a visual language pre-training model MDETR, a text region can be identified in a complex medical bill image.

The quality of signal sorting directly affects the signal recognition performance of a radar radiation source and the subsequent decision judgment of a combat system, so that the comprehensive performance of EDW (Emitter Descriptions Words) of signal sorting needs to be evaluated. The invention provides a comprehensive evaluation method for an electronic warfare passive reconnaissance signal sorting result, and the method comprises the steps: carrying out the comprehensive performance evaluation and confidence judgment of an EDW through introducing an AHP (Analytic Hierarchy Process) according to the parameters, such as the frequency, the pulse width, the repetition period and the associated pulse number, after the sorting of a radiation source signal, carrying out the trace point association and target recognition based on the judgment result, and carrying out the recognition of the target. And the reliability and the accuracy of a signal sorting result are greatly improved. The technology provides a stable and reliable comprehensive evaluation method for the signal sorting result (EDW) in the complex electromagnetic environment, improves the accuracy and target recognition probability of signal sorting in the complex electromagnetic environment, and reduces the batch increasing rate of the system.

The embodiment of the invention provides a face recognition method and device based on adaptive learning, equipment and a medium, and belongs to the technical field of financial science and technology. The method comprises the following steps: acquiring a current face image of a target user; comparing the current face feature vector of the target user with a historical face feature vector of the target user, and determining a feature vector variable quantity; adjusting a first target model parameter of the face recognition model through an incremental learning module under the condition that the feature vector variable quantity exceeds a first preset change threshold value; under the condition that the feature vector variable quantity exceeds a second preset change threshold value, adjusting a second target model parameter of the face recognition model through an online learning module; and optimizing the face recognition model according to the adjusted first target model parameter or the second target model parameter. According to the embodiment of the invention, self-adaptive learning can be carried out on the face change of the user, and the long-time accuracy of face recognition can be ensured in a financial scene.

The invention discloses a multi-person deep squat real-time counting method based on human body key point measurement, and the method can improve the detection precision and robustness, supports the simultaneous detection of multiple persons, and is suitable for multiple scenes. The method comprises the following steps: based on a multi-target real-time detection and tracking technology of a CSTrack algorithm, identifying and tracking a plurality of detected persons at the same time, and recording and analyzing the motion trail and position of each person; based on a BlazePose algorithm, carrying out attitude estimation on a plurality of detected persons, detecting and tracking the human body attitude data of each detected person in real time, and measuring the distance and angle change between joint points to judge the start and end of the action; kNN similarity discrimination and CNN depth feature learning are combined to analyze and optimize the posture of the human body in real time, whether a detected person performs a deep squat action or not is judged, the deep squat category is determined, and then deep squat counting and action standard scoring are performed.

The invention relates to an iteration-based undersampling cost-sensitive ensemble classification method, belongs to the technical field of medical data analysis, and is particularly suitable for processing a disease prediction model in an unbalanced data set. According to the method, the sampling probability of the majority class samples is dynamically adjusted through an iterative undersampling strategy, it is ensured that the features of the minority class samples are focused on in the training process, and meanwhile important information of the majority class samples is prevented from being lost. In combination with cost-sensitive learning and ensemble learning, the invention provides a classification method capable of remarkably improving the minority class sample prediction performance while keeping the overall accuracy of a classifier. The method has good adaptability, is suitable for various unbalanced data set scenes, and shows an excellent application effect in medical data analysis and disease early warning.

The invention discloses a DC arc discharge fault detection method based on multi-scale feature extraction and clustering screening, and relates to the technical field of artificial intelligence algorithms and photovoltaic fault diagnosis, and the method comprises the following steps: S100, signal collection; s200, carrying out multi-scale analysis; s300, time domain feature extraction is carried out; s400, frequency domain feature extraction and clustering screening are carried out; s500, generating a feature vector; s600, training a random forest model; and S700, analyzing a series direct current arc discharge fault. According to the invention, the identification rate of direct-current arc fault detection is effectively improved, the false alarm rate of direct-current arc fault detection is reduced, and the overhaul and maintenance cost of photovoltaic equipment is reduced.