16results about How to "Avoid misclassification" patented technology

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 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.

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 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 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.

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 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.