30results about How to "Avoid misclassification" patented technology

The invention relates to the technical field of point cloud data modeling, and discloses a point cloud data processing and modeling method, system and device for power distribution equipment and a storage medium, and the method comprises the steps: collecting the point cloud data of the power distribution equipment, and constructing a topological relation graph according to the topological relation of the power distribution equipment; semantic segmentation is performed on the point cloud data based on the topological relation graph, a component association model is constructed, and missing data detection is performed in combination with the topological relation graph; and generating a high-precision three-dimensional model based on the component correlation model. According to the method, semantic segmentation is guided through the topological relation graph, the accuracy of point cloud data classification is improved, and the problem of misclassification in a traditional method is avoided. Meanwhile, a part correlation model is generated based on the segmentation result, the geometric and functional connection relation between the devices is comprehensively described, and deep fusion of the device connection information and the three-dimensional geometric model is achieved.

The invention belongs to the technical field of computer vision and artificial intelligence, and discloses a combined zero sample image classification method based on hierarchical feature fusion. According to the method, a solution based on hierarchical feature fusion is innovatively provided for the technical difficulty of combined zero sample image classification. According to the method, features of different depths in a CLIP model visual encoder are selected and subjected to hierarchical fusion, multi-level and multi-scale visual information is effectively extracted, the fused features are further used for visual and semantic cross-modal interaction, and the understanding ability of the model for complex combinations is remarkably improved. In addition, by introducing a specific loss item, attributes and object features are effectively separated, and the feature aliasing problem is avoided. In a combined zero sample image classification task, the method shows excellent generalization ability and recognition effect, and can significantly improve the classification performance of the model for unknown complex combined images.

The invention provides a traditional Chinese medicinal material intelligent storage and prescription filling control system based on AI image recognition. According to the system, the AI image recognition technology, automatic storage and intelligent control are combined, the functions of automatic recognition and classified storage of traditional Chinese medicinal materials and automatic medicine dispensing by recognizing a handwritten prescription of a doctor through AI or inputting the prescription through characters and voice are achieved, and the efficiency and accuracy of traditional Chinese medicinal material storage and medicine taking are greatly improved.

The present application provides a kind of forest tree species classification method, device, electronic equipment and storage medium, belong to artificial intelligence technical field, method includes: forest remote sensing image is input to forest tree species classification model, obtains the output tree species classification result;Forest tree species classification model includes preliminary feature processing module, high-order feature extraction module and tree species classification result output module, high-order feature extraction module is sequentially connected first residual block stack layer, first down-sampling layer, second residual block stack layer, second down-sampling layer, third residual block stack layer, third down-sampling layer and fourth residual block stack layer, stacked residual block embedding dimension rearrangement layer, dimension recovery layer and convolution block attention module.The present application improves the dimension compatibility of CBAM and convolution architecture by embedding explicit dimension rearrangement layer and dimension recovery layer in the residual block embedded with CBAM, effectively improves the classification accuracy of forest remote sensing image for forest tree species classification task.

The invention belongs to the technical field of data processing and analysis, and particularly relates to an industrial system operation mode description method based on a dynamic clustering algorithm DeepDPM. The method comprises the following steps: firstly, collecting actual operation data of a system; preprocessing the data, and dividing a training set and a verification set according to a fixed proportion; and then establishing an AE model, pre-training the AE model by using the training set, storing the trained AE model and the weight, and verifying the effect of the trained AE model by using the verification set. Inputting the preprocessed data into the trained AE model for feature extraction; then, the extracted feature data are input into a DeepDPM model to be clustered; and finally, carrying out visual display on a clustering result by using a dimension reduction visualization algorithm: t-distributed random neighborhood embedding (t-SNE for short). According to the invention, the efficiency and accuracy of operation mode prediction in an industrial system are improved.

The application discloses a kind of semantic boundary guided robot operation scene point cloud segmentation methods. Firstly, the three-dimensional point cloud of robot operation scene is obtained;Then, the semantic boundary point in the scene three-dimensional point cloud is predicted using point cloud learning network;The scene three-dimensional point cloud is input into point cloud Transformer model for feature extraction, and the semantic boundary point is used as the extended key point of self-attention mechanism in Transformer model, wherein for the point feature of each boundary region, a contrast learning loss function is added to reduce the semantic ambiguity of boundary point feature;Finally, based on the point feature of each point, the semantic class of each point is predicted to obtain the point cloud semantic segmentation result. The application enhances the learned point feature by introducing semantic boundary information in the point cloud feature extraction process, and improves the misclassification problem caused by boundary ambiguous features, effectively improves the point cloud segmentation quality of robot operation scene, and has good engineering practical value.

The utility model relates to a dispensing tool for carton forming processing, which comprises a main frame body, a rolling dispensing assembly is rotatably mounted on the main frame body, and the rolling dispensing assembly is connected with a pressurized liquid supply assembly mounted on the main frame body through a liquid supply supporting pipe. The rolling dispensing assembly comprises a rotating disc, supporting rib rods, a mounting ring plate and a telescopic dispensing mechanism, the rotating disc is rotatably connected with the main frame body, and the rotating disc is coaxially connected with the mounting ring plate through the supporting rib rods which are annularly arranged at intervals; and a plurality of telescopic dispensing mechanisms which radially coincide with the mounting ring plate are inserted into the mounting ring plate in a rotational symmetry manner. According to the utility model, the dispensing spacing distance can be changed according to requirements, so that the dispensing quality and the dispensing efficiency are improved, and meanwhile, the pressure-contact glue discharging can be realized, so that the overflow and splashing of glue liquid are reduced.

The invention discloses a sensor product classification method based on a multi-modal condition, and relates to the technical field of product classification, and the method comprises the steps: defining the multi-modal condition and a standard value of a sensor; performing preprocessing operation on the multi-dimensional measurement characteristic data of the sensor, and screening effective samples to obtain a standardized deviation rate matrix; constructing a multi-dimensional similarity component for any two effective samples; setting a similarity component weight, and performing weighted fusion to obtain a comprehensive similarity; and the comprehensive similarity is converted into distance metric, effective samples are clustered by adopting a DBSCAN algorithm, a clustering result is output, and classification operation of sensor products is realized. Through the multi-dimensional similarity component, the classification result can be consistent with the actual performance of the sensor, the weight of the similarity component can be adjusted according to business requirements, classification targets of different application scenes can be adapted, abnormal samples seriously deviating from the standard are eliminated by screening effective samples, the difference between the dimension and the baseline is eliminated through standardization, and the classification accuracy is improved. And the interference of noise on the clustering result is reduced.

The application provides a lightweight millimeter wave radar two-dimensional feature map classification method and system for FPGA hardware deployment, and solves the technical problems of poor real-time performance and high power consumption of existing millimeter wave radar two-dimensional feature map classification methods in edge computing scenarios. It includes obtaining a feature map and preprocessing it to obtain a preprocessed feature map, constructing a dataset, and dividing the dataset into a training set and a validation set; building a classification network model based on LeNet, training the classification network model using the training set, and obtaining a trained classification network model; verifying whether the performance of the trained classification network model meets the performance indicators using the validation set, and if so, obtaining a validated classification network model, deploying it to an FPGA, classifying the feature map based on the classification network model, and obtaining a classification result; otherwise, continue training to update the parameters of the classification network model until the performance after verification meets the performance indicators. The application can be widely applied in the field of image classification technology.

This specification provides a parcel sorting and detection method and system. The parcel sorting and detection method includes: receiving parcel sorting information of parcels to be sorted; collecting parcel monitoring information of parcels to be sorted during the sorting process; identifying parcel transportation information of parcels to be sorted based on the parcel monitoring information and the parcel sorting information; predicting the parcel compartment prediction information based on the parcel transportation information; and determining the abnormal compartment of the parcel to be sorted at the detection time of the parcel to be sorted when the predicted compartment information does not match the target compartment information in the parcel sorting information, and generating sorting abnormal information of the abnormal compartment.

The utility model discloses paper product detection equipment. The paper product detection equipment comprises a visual equipment module; the double-station discharging module is provided with two incoming material pipelines which are in butt joint with the forming machine, and it is ensured that the paper cups stably fall onto the clamp assembly through a multi-layer support assembly; the rotary platform disc module comprises a rotary platform disc, a plurality of stations are arranged on the rotary platform disc, the stations correspond to different detection positions, and the visual equipment module is used for carrying out omnibearing quality detection on the paper cups; and the non-defective product and defective product discharging module is used for respectively blowing qualified products and unqualified products to corresponding collecting areas by utilizing an electromagnetic valve control pipeline according to the judgment result of the visual equipment module. According to the utility model, a fixed shooting mode is adopted, the paper cup is firstly stopped and then shot, the problem of fuzzy and unstable imaging of mobile shooting is solved, one camera is adopted to shoot the paper cup at one station, and the problems of mutual interference of light sources and poor imaging are solved.

The present invention discloses a method and system for adjusting navigation information for a navigation robot, wherein the method comprises: generating a navigation path based on an input tour location combined with the location information of the navigation robot; when the navigation robot moves to the corresponding tour location, dividing the obtained navigation area image into a plurality of image blocks; extracting features from each image block to obtain image block features, and encoding the features to obtain context information features; determining the corresponding navigation scene based on the context information features; comparing the similarities and differences between the scene label text and each navigation scene information, obtaining the target navigation scene information for display; the navigation robot obtains the operation instruction input by the user, obtains the corresponding operation information based on the operation instruction, and adjusts the navigation scene information based on the operation information. The present invention improves the reliability of the navigation service of the navigation robot, thereby being able to provide users with better navigation services.

The application relates to the technical field of lithology intelligent classification, in particular to a lithology intelligent classification system and method based on multi-modal data fusion, which comprises a data acquisition module, a feature extraction module, a cluster analysis module, a classification prediction module and a visualization module. Through preprocessing, sensitivity quantization and weighted feature space construction of the original drilling data of the tunnel, the improved K-means++ algorithm is combined to dynamically adjust the cluster number, generate the lithology category label, and display the distribution through the visualization interface. The application can effectively reduce noise interference, reasonably allocate feature weights, improve the lithology identification accuracy and robustness under complex geological conditions, and provide intuitive support for geological decision-making.

The present application relates to infusion pump technology field, especially infusion device control method, before the user starts to use the device, the self-locking type switch circuit is disconnected, the system is in zero power consumption complete power-off state, when the user starts to use the device, the trigger signal of opening the device is generated, the switch circuit is turned on and self-locked, and the control circuit is powered; thereafter, the system controls the subsequent process by accumulating the number of pulses sent to the drive mechanism: when the pulse number reaches the first preset value, the initialization is completed; when the second preset value is reached, the trigger of the needle assisting mechanism is completed, and the infusion needle is implanted into the human body; finally, according to the user setting infusion instruction starts infusion. The control method ensures the safety of the infusion device through the hardware barrier, realizes the automatic and accurate control of the initialization and needle assisting mechanism by using single pulse counting, and realizes the true zero standby power consumption, which significantly improves the safety, reliability and user experience of the infusion device.

The application provides a forest resource remote sensing data spatio-temporal change analysis method and system, relates to the technical field of data processing, and comprises the following steps: according to the result of spectral unmixing processing, performing ground object type classification processing on each pixel to generate a forest coverage distribution map; using the forest coverage distribution maps at multiple time points, extracting all pixel center points on the boundary of a change region as an input point set; finding out the points located at the outermost periphery in the input point set through iterative calculation and connecting the points to form a convex polygon; normalizing and integrating the shape and range of the change region based on the convex polygon to obtain a normalized description result; calculating the annual change rate of the forest area based on the normalized description result; and calculating the vegetation coverage and forest biomass ecological indexes based on the annual change rate of the forest area. The application can realize more accurate and continuous monitoring and evaluation of the spatio-temporal change of forest resources.

The invention relates to a land utilization classification method and device based on land parcel units, equipment and a medium, and relates to the technical field of land utilization type classification, and the method comprises the steps: obtaining a boundary vector file and remote sensing image data of a target classification region, carrying out the feature conversion based on the remote sensing image data, and obtaining multi-dimensional array data, geometric transformation is carried out according to the boundary vector file, geometric shape data are obtained, the geometric shape data comprise land parcel unit information, classification is carried out through a preset land classification model according to the multiband remote sensing images, the geometric shape data and the multidimensional array data contained in the remote sensing image data, and a land utilization classification result is obtained. The land utilization classification is realized by taking the land parcels as basic classification units, the classification error that the interior of the same land is divided into different land types is avoided, the refinement degree of land utilization classification is improved, and the problems of low identification and classification efficiency, insufficient accuracy and the like of an existing land utilization classification method are solved.

The application provides a medical waste automatic low-temperature throwing and storing device, which comprises a cold storage, a plurality of recovery cages, a data entry device and a grabbing assembly. An annular track is arranged on the top wall of the cold storage and penetrates through the front side wall of the cold storage. The plurality of recovery cages are arranged in the cold storage along the direction of the annular track at intervals. The recovery cages have upward openings and are located below the annular track. The data entry device is arranged on the front side of the cold storage and is used for entering the information of the waste in a garbage bag. The grabbing assembly is slidingly connected to the bottom of the annular track and is used for grabbing the garbage bag and throwing it into one of the recovery cages. The medical waste automatic low-temperature throwing and storing device provided by the application solves the problems of rough classification of medical waste recovery stations, lack of low-temperature storage for special medical waste and low efficiency of manual sorting.

The application discloses a kind of based on multi-modal feedback and feature purification ancient building point cloud model completion method, including calculating the local density of each three-dimensional coordinate point in ancient building defect point cloud model, obtain adaptive neighbor quantity, utilize dynamic graph convolution network to process feature descriptor, obtain the component class to which three-dimensional coordinate point belongs;Retain the component class to which the three-dimensional coordinate point corresponding to the component class of classification confidence higher than dynamic confidence threshold belongs, obtain final belonging component class, obtain purified feature based on the class;Construct two-dimensional uniform grid point set, splice purified feature with grid point coordinates in two-dimensional uniform grid point set, obtain high-dimensional feature vector, utilize folding decoder to obtain the component point cloud set after completion, and then obtain the point cloud set after smoothing, utilize multi-modal large language model to process the set, complete point cloud completion.The application realizes high-fidelity fine repair, guarantees the completion stability under full scene.

The invention belongs to the technical field of radar signal processing, and discloses a low-altitude surveillance radar target intelligent identification system. The method comprises the following steps: analyzing and standardizing radar original echoes or detection output, associating discrete plots based on space-time correlation to form a target track, and generating track motion state parameters; performing space-time-frequency combined feature extraction on the trace point and / or track sequence, and constructing a comprehensive feature set comprising one or more of rotor micro Doppler, multi-scale time sequence, time-frequency non-stationary and physical constraint; the features are input into a multi-expert dynamic fusion recognition model, the model at least comprises time sequence modeling experts and statistical learning experts, and output of all the experts is fused in a self-adaptive mode through a dynamic weight fusion mechanism; and performing track-level aggregation and stabilization processing on the prediction result, and outputting a track-level target category label and confidence, thereby avoiding misclassification caused by a single model under the condition that the track is incomplete, and enhancing the stability of an identification result in a complex environment.

This invention discloses a method, system, and storage medium for capacity compensation of lithium-ion battery cells. The method includes: constant current charging of the lithium-ion battery cell under test to a charging cutoff voltage, then switching to constant voltage charging until the charging current drops to the cutoff current; constant current discharge of the fully charged cell to a first cutoff voltage using a first discharge current, recording the released capacity as the main discharge capacity; after the main discharge, allowing the cell to rest for a preset time; continuing constant current discharge of the rested cell to a second cutoff voltage using a second discharge current, recording the released capacity as the reference discharge capacity; and calculating the compensated capacity using a capacity compensation formula based on the main discharge capacity, the reference discharge capacity, and a pre-determined reference discharge capacity benchmark value. This invention can eliminate the influence of polarization differences caused by ambient temperature fluctuations on the capacity test results during capacity testing, improving the accuracy of capacity testing.

The invention relates to an equal-part quantitative automatic packaging method and system for toy parts. The equal-part quantitative automatic packaging method comprises the following steps that all the toy parts are conveyed into a conveying belt in a scattered and tiled mode through vibration of a vibration disc; a CCD camera is adopted to shoot the toy parts on the conveying belt; the obtained part image is preprocessed; a mechanical vision module is adopted to carry out part type identification on the preprocessed image; classifying the identified part types; generating a classification signal based on the identification classification result; a mechanical arm module is controlled to grab toy parts according to the target category, and the toy parts are put into a quantitative disc; after the weight reaches a set standard, the parts in the quantitative disc are emptied, packaged and sent out; according to the equal-part quantitative automatic packaging method and system for the toy parts, the accuracy, efficiency and flexibility of automatic packaging are effectively improved, and multiple technical defects in a traditional toy part packaging method are overcome.

The invention is suitable for the technical field of data recognition, and provides a task processing method and device for smart office, equipment and a medium, and the task processing method for smart office comprises the steps: obtaining to-be-recognized text data and first-level classification information of the to-be-recognized text data; extracting a current title part and / or a current abstract part in the to-be-recognized text data; determining target secondary classification information of the to-be-recognized text data according to the current title part and / or the current abstract part; according to the keyword corresponding to the target key content category, extracting a text region corresponding to the target key content category in the to-be-recognized text data; and in the to-be-identified text data, labeling target key content categories corresponding to the plurality of text regions. According to the technical scheme, through intelligent document classification, key content extraction and region labeling, the automation degree and accuracy of document processing can be greatly improved.

The utility model discloses a full-automatic winding machine mistaken-supporting-preventing device which comprises a fixing base, a belt conveyor and a yarn drum, an annular fixing plate is installed in the hollow portion of the bottom of the yarn drum, two fixing side plates are installed on the fixing base, a soft magnet sheet is installed on a conveying belt of the belt conveyor, and the soft magnet sheet is installed on the belt conveyor. Two guide side plates are installed between the two fixed side plates, the two guide side plates are located on the two sides of the soft magnet sheet correspondingly, a plurality of notches are formed in the two guide side plates, a vertical side plate is installed on the fixed base and located on one side of the belt conveyor, and the vertical side plate is located on the other side of the belt conveyor. A plurality of multi-stage telescopic air cylinders are installed on the vertical side plate, and connecting push plates are installed at the telescopic ends of the multi-stage telescopic air cylinders. The yarn drum sorting device has the advantages that the yarn drums can be moved to the designated positions and automatically sorted and placed, manual picking is not needed, working efficiency is improved, friction force to the yarn drums is reduced, yarn on the yarn drums is prevented from being damaged, and production quality is improved.

The invention discloses an information management method of an information management system, and the method comprises the following steps: determining an information type according to the existing database information, determining an information type label according to the information type, and finally dividing the information importance level according to the information type. The collected information is input, and the system automatically carries out information data preprocessing and quality verification; and after the quality verification is completed, the system automatically identifies the information type and classifies the information according to the known category label. According to the method, the mathematical model is established according to the known database, and the mathematical model is used for automatically classifying the information, so that a large amount of information data can be conveniently and quickly processed, and meanwhile, the method also has the functions of information data quality inspection, repeated data elimination and encryption of different levels of information, so that the information management is more efficient, accurate and safe.

According to the problem integration method and device, the storage medium and the computer equipment provided by the invention, a to-be-integrated problem is acquired, key feedback information in the to-be-integrated problem is identified, and a target service node is determined according to the key feedback information; according to the target service node, obtaining a target prompt template corresponding to the to-be-integrated question, and combining the key feedback information with the target prompt template to generate a prompt text corresponding to the to-be-integrated question; inputting prompt texts corresponding to the to-be-integrated questions into the question classification large language model to obtain target question types of the to-be-integrated questions; determining a high-dimensional vector of the to-be-integrated problem, and determining a high-dimensional vector of each problem library in the target problem type; and calculating the vector similarity between the high-dimensional vector of the to-be-integrated problem and the high-dimensional vector of each problem library, and determining a target problem library of the to-be-integrated problem in each problem library according to each vector similarity. Therefore, manual participation can be remarkably reduced, the possibility of human errors is reduced, and the problem processing efficiency is improved.

This invention discloses a method and system for identifying land cover in desert areas based on remote sensing image interpretation. It acquires multi-band reflectance and various remote sensing index data from satellite remote sensing images of the target area and performs band merging using a band combination tool to obtain merged image data. Based on statistically obtained remote sensing index threshold ranges corresponding to different land cover types, it constructs a hierarchical land cover separation and extraction rule. Based on the constructed hierarchical land cover separation and extraction rule, and combined with mask extraction, it separates and extracts different types of land covers from the band-merged image data. The interpretation method of this invention is relatively flexible, using different remote sensing index threshold ranges for different land cover types to effectively extract land cover feature information. Based on a rule-oriented object-oriented workflow, it can effectively extract various types of land covers, and the use of mask extraction effectively avoids misclassification of land covers.

The application provides a power user power consumption behavior portrait method considering power consumption sensitivity, and belongs to the technical field of industrial data processing. The method solves the problem that in most power user portrait methods, sensitive loads and basic loads influenced by other factors are mixed for analysis, which makes the portrait result not fine enough and unable to accurately grasp the power consumption behavior characteristics of users. The method comprises the following steps: obtaining a power load data set, drawing a load curve, and performing preliminary clustering; labeling the preliminary clustering cluster; inputting the preliminary clustered load data set into an STL model for decomposition and division into basic loads and sensitive loads; performing clustering analysis on the basic loads and the sensitive loads respectively, defining a basic load label library and an external factor sensitive load label library according to the clustering results; and labeling each user with two types of labels according to the basic load label library and the external factor sensitive load label library, combining the labels of the initial clustering results, and generating a precise portrait of the power user.

The present application relates to the field of image recognition technology, and in particular to a method for extracting invoice information and determining invoice categories based on a large language model, the method comprising obtaining an invoice file, extracting the invoice text of the invoice file; designing a first-level prompt word, inputting the first-level prompt word and the invoice text into a preset large language model, and the large language model outputs the key information of the invoice text under the guidance of the first-level prompt word; for each commodity in the key information of the invoice text, designing a second-level prompt word, the second-level prompt word includes the commodity name, the corresponding commodity category and the invoice category, inputting the second-level prompt word into the large language model, and the large language model outputs the invoice category coding result of each commodity under the guidance of the second-level prompt word; integrating and outputting the key information of the invoice text and the invoice category coding result of each commodity. The present application can achieve a significant improvement in the efficiency and accuracy of invoice processing.

The application provides a single-cell rare cell type identification method and system based on a divide-and-conquer strategy, comprising the following steps: S1, using a variational autoencoder to encode single-cell transcriptome data into Gaussian distribution hidden variables; S2, performing spectral clustering on the Gaussian distribution hidden variables to obtain a single-cell coarse-grained clustering result; and S3, optimizing the single-cell coarse-grained clustering result based on intra-class cluster loss and inter-class cluster loss until the intra-class cluster loss and the inter-class cluster loss are minimized to obtain a single-cell clustering result containing rare cell types. The application can overcome the deficiency that previous single-cell clustering and identification methods do not optimize rare cell types, leading to misclassification or difficulty in identifying rare cell types.

The invention relates to the technical field of text processing, and particularly discloses a non-standard document slicing processing method which comprises the following steps: S1, determining a first keyword and a second keyword; s2, obtaining a first word vector and a second word vector, and determining the association degree between the sub-paragraph and the chapter name based on the first word vector and the second word vector; s3, determining paragraphs to be determined according to the association degree, and sorting the paragraphs to obtain a paragraph sequence; determining a chapter name corresponding to the target paragraph, and determining an undetermined paragraph again according to a new chapter name; and when the to-be-determined paragraph does not exist, judging that the document is normalized, and slicing the normalized document. According to the method, the recall accuracy is improved by normalizing and slicing the non-standard document.