10results about How to "Reduce classification error" patented technology

The invention discloses a plant disease identification method and system based on unsupervised category incremental learning, and a storage medium, relates to the technical field of plant disease identification, and aims to solve the problems that an existing deep learning model cannot meet the application requirements of real-time updating and optimization of the model, the timeliness is poor, and the efficiency is high. And the problem of failure of the class incremental learning strategy when the new class data is not labeled by the label exists. Comprising the following steps: S1, dividing a data set into category sets of a plurality of stages according to categories; s2, performing interpolation between the image of the current task and the random image uniformly sampled and selected from the cache region of the old category data by using a hybrid playback algorithm to generate a synthetic sample; s3, constructing a SimSiam model which comprises two identical neural network branches and shared parameters of the two branches, constructing a loss function based on the negative cosine similarity of the two network branches, and training the SimSiam model in multiple stages; s4, evaluating the performance of the model based on a KNN algorithm; and S5, classifying the plant diseases by adopting the trained SimSiam model.

The invention relates to a target classification method and device, computer equipment and a storage medium. The method comprises the steps of performing dual-channel feature extraction on a plurality of targets in a to-be-processed image to obtain semantic feature information and spatial feature information corresponding to each target, performing fusion processing on the semantic feature information and the corresponding spatial feature information of each target, determining to-be-processed feature information of each target, and determining the to-be-processed feature information of each target based on the to-be-processed feature information of each target. The targets are correspondingly classified into different types of target sets to complete classification processing, and the difference value between the different types of targets in the multiple targets is smaller than a preset difference threshold value. By the adoption of the method, fine-grained classification can be achieved, the fine-grained classification difficulty is reduced, in the processing process, the multi-type feature information, namely semantic feature information and spatial feature information corresponding to all the targets, of all the targets can be obtained, then accurate classification is achieved based on the multi-scale feature information of all the targets, and the classification efficiency is improved. Therefore, the accuracy of fine-grained classification can be improved.

The application discloses a bridge collision damage rapid evaluation method based on a conditional variational autoencoder. The method first acquires a vehicle-bridge collision dataset, determines a bridge damage state category and takes it as an output label; then, the dataset is layered according to the bridge damage state; a cVAE model is trained by using a training set, high-dimensional features are reduced and purified by an encoder, and the high-dimensional features are recovered to high-dimensional features by a decoder to realize effective representation; meanwhile, end-to-end training is carried out in combination with a damage classification task, feature coding, reparameterization, feature reconstruction, damage classification, loss calculation and parameter updating are sequentially completed, grid search is used to optimize hyperparameters, and the optimal cVAE model is used for actual evaluation. The application effectively overcomes the problems of traditional methods in high-dimensional bridge monitoring data processing, such as difficulty in feature extraction and insufficient classification accuracy; and has stronger generalization ability and higher recognition accuracy under the conditions of small samples and unbalanced damage samples.

The present application relates to the technical field of agricultural product origin identification and traceability, and particularly discloses an agricultural product origin identification and traceability system and method based on characteristic fingerprints, which comprises the following steps: collecting agricultural product samples to obtain a target sample set; marking the target sample set with dominant characteristic fingerprints according to the dominant characteristics of the agricultural products, and performing one-class partitioning on the target sample set according to the dominant marking result to obtain a plurality of one-class sample clusters; marking the target sample set with recessive characteristic fingerprints according to the recessive characteristics of the agricultural products, and performing two-class partitioning on the target sample set according to the recessive marking result to obtain a plurality of two-class sample clusters; comparing the one-class partitioning result and the two-class partitioning result to determine whether to start a successive verification mode or a separate verification mode; based on the successive verification mode, obtaining each verification result, and determining whether to output an actual origin traceability result according to the verification results.

The invention discloses a bark traditional Chinese medicine identification method and system based on a lightweight neural network, and relates to the technical field of medicinal material identification, and the method comprises the steps: collecting RGB images of various bark traditional Chinese medicines under different conditions, and forming a database; classifying and marking the images of various skin type traditional Chinese medicinal materials in the database, and establishing an original skin type traditional Chinese medicinal material data set; constructing a ShuffleNet V2-lite network model on the basis of a positioning attention mechanism and a cross-stage local network; the classification result of the bark type traditional Chinese medicinal materials in the ShuffleNet V2-lite network model is corrected; and the skin type traditional Chinese medicinal materials are detected and classified by using the updated ShuffleNet V2-lite network model. When the method is used for detecting the bark type traditional Chinese medicinal materials, different varieties of bark type traditional Chinese medicinal materials can be accurately distinguished, classification errors caused by similar appearances are reduced, intelligent classification and high-precision identification of the bark type traditional Chinese medicinal materials are achieved, the accuracy and reliability of classification are improved, and meanwhile the method has great significance in improving the quality control level of the traditional Chinese medicinal materials.

The invention discloses a trade product classification method and system. The method comprises the steps of obtaining multi-modal data of a target product; the multi-modal data comprises text data and image data; screening product keywords from the text data, constructing text features, and performing first hierarchical classification on the product keywords to obtain a first classification result; according to the first classification result, selecting a target neural network from pre-trained convolutional neural networks for different categories of products, and extracting image features from the image data by adopting the target neural network; performing feature fusion on the text features and the image features to obtain multi-modal features; and according to the multi-modal features, carrying out second-level classification on the target product to obtain a second classification result. By adopting the embodiment of the invention, the advantages of different data can be fully utilized, the classification error is effectively reduced, and the practicability and reliability are improved.

The invention discloses a coastal aquaculture pond extraction method based on a DBCE-Net image and a Sentinel-2 image. According to the method, a DBCE-Net framework adopting a double-branch structure is provided, and the fine features of a high-resolution small image and the global features of a down-sampling large image are combined, so that classification errors caused by context information loss are remarkably reduced, and the extraction precision of the coastal aquaculture pond is improved; and through the dynamic feature fusion module, features extracted by the main branch and the auxiliary branch are effectively fused, the robustness and stability of the model are improved, and a good extraction effect can be obtained under different regions and different image conditions. Compared with a traditional method in combination with machine learning, the method based on deep learning has higher generalization ability, multiple threshold values and parameters do not need to be set for different areas, and the method can be applied nationwide more conveniently and efficiently.

The present invention relates to the technical field of crop classification, and in particular to a crop classification method based on multi-angle satellite observation data, comprising S1. acquiring zenith reflection data and multi-angle observation factor data of a study area; S2. using a land cover data set to screen crops in the zenith reflection data and multi-angle observation factor data, and obtaining single-angle observation type crop images and multi-angle observation type crop images; S3. constructing a sampling point database of different types of crops in the study area; S4. constructing a training data set and a verification data set based on the sampling point database, single-angle observation type crop images and multi-angle observation type crop images, and training four machine learning classifiers provided by a GEE platform; S5. using the trained four machine learning classifiers to classify single-angle observation type crop images and multi-angle observation type crop images. The present invention can improve the accuracy of crop distribution mapping.

This invention relates to the field of text classification technology, and more particularly to a text classification method, medium, and device. It dynamically determines the first parallel inference quantity M using computational resources, achieving a balance between computational resource utilization and inference stability. By initiating multi-threaded parallel inference, it generates M candidate output sequences containing a first intermediate semantic representation and a first text classification result, covering different inference logics of the model for the target text. This approach balances result diversity and inference efficiency, reducing classification errors caused by single-sequence inference bias. Through compliance verification and confidence assessment, it calculates a first quality score to obtain the target classification result. This not only filters sequences containing security risks or structural defects, ensuring the security of the classification process, but also distinguishes the reliability differences of the classification results, avoiding one-sided judgments that ignore reliability based solely on labels. In scenarios without real labels, it achieves objective quality quantification, ensuring the accuracy of the text classification results.

The invention discloses a stratigraphic structure recognition method and system based on K-means unsupervised learning, and relates to the technical field of stratigraphic structure recognition. The method comprises: step S100, obtaining data of various drilling parameters of a drilling rig during drilling through a drilling test device, and constructing an initial data set; step S200, normalizing the initial data set to obtain a normalized data set for K-means unsupervised learning; step S300, setting a category K of a prediction result, a maximum number of iterations, and a deviation; step S400, randomly selecting K data points as initial clustering centers; step S500, respectively calculating the distance from each data point in the normalized data set to the initial clustering center, and classifying points whose distance is less than or equal to a preset distance threshold into one category; step S600, reselecting K data points as updated clustering centers; step S700, repeating step S500 for iteration; step S800, outputting a classification result after reaching the maximum number of iterations in step S300, thereby completing stratigraphic structure recognition.