31results about How to "Improve classification rate" patented technology

This invention aims to enhance efficiency of a sieve and to greatly improve productivity of a sifting operation. There is provided a sieve comprising a metal plate including long holes, wherein the long holes are plurally provided such that lines extending in length directions thereof cross one another.

The invention discloses an undesirable image detecting method based on connotative theme analysis, which is substantially used for solving the problem of wrong judgment on normal images resulting from semantic information consideration failure in the present undesirable information detecting method. The scheme is as follows: extracting a skin region of an image by a double-blending Gaussian model; generating a codebook base containing distinguishing features in the skin region by a word bag model, and representing each training image to a group of word co-occurrence vectors with weights via aword frequency-inverse identification file frequency method; forming all co-occurrence vectors to a co-occurrence matrix, performing LDA model creation on the co-occurrence matrix to obtain the themeof the image; inputting the mixed theme of the training image in a BP neural network to train an undesirable image classifier; and obtaining the theme of an image to be measured, inputting the theme to the undesirable image classifier, and judging whether the theme is an undesirable image so as to finish the undesirable image detection. As shown in the test, the invention can be used for better distinguish the undesirable images and the normal images, so that the invention can be used for filtering the erotic information in the images.

The invention discloses a method for optimizing a hot rolling production line on-line surface defect detection system. The method mainly comprises the following steps of: (1) maintaining hardware; (2) calibrating equipment; (3) grouping materials; (4) verifying defect; (5) establishing and optimizing a classifier; and (6) performing post-treatment. By the method, the hot rolling production line on-line surface defect detection systemis optimized, so that the surface defect detection rate of the system can be increased to over 95 percent in a short time, the surface defect classification rate is increased to over 90 percent, and instantaneous, stable and accurate on-line surface defect detection and classification are realized.

The invention discloses a method for quickly classifying seismic waveforms based on a semi-supervised algorithm, comprising the following steps: S1, obtaining seismic waveform data along horizons, and adopting an SSDR algorithm based on linear transformation to reduce the dimensionality of seismic waveform data; S2, using seismic The label data in the waveform data trains a distance measurement matrix; S3, using the semi-supervised Kmeans classification algorithm to classify the seismic waveform data to generate a seismic phase map. The invention uses a semi-supervised dimension reduction method to process the original seismic waveform data, eliminates redundant data, increases the similarity of the same type of waveform data, and enhances the differences of different types, making the classification results more accurate; using the existing A suitable distance measurement matrix is ​​trained from the well logging data, and the distance measurement method is introduced into the subsequent classification method, and then a weighted semi-supervised Kmeans classification method is proposed, which makes full use of the well logging data and improves the Classification accuracy, and speed up the classification rate.

The invention discloses a method for detection and classification of internal circular spline products in harmonic reducers. The method comprises the steps of (1) product detection, (2) channel product classification and (3) maintenance and overhauling. The invention provides a technical proposal related to the method for detection and classification of the internal circular spline products in theharmonic reducers. The overall structural design is rational; all-automation processing of product detection and classification is realized; manual labor strength is effectively reduced; safety performance of operations is improved; classification of the products is more accurate and convenient; the error rate of product classification is effectively reduced; the waste rate of resources is reduced; and data is transmitted to a control system in time during full-automatic classification, so ratio data of qualified and rejected products detected in classification can be updated in time and thusthe reject rate of the equipment production products can be controlled.

An apparatus and method for controlling a convergence of a projection television includes a pattern generator which generates R, G, and B video patterns used for controlling the convergence and then outputs the R, G, and B video patterns to a display, and a controller which compares position coordinates and luminance levels of the R, G, and B video patterns with position coordinates and luminance levels of reference R, G, and B video patterns to adjust the luminance levels of the R, G, and B video patterns to the luminance levels of the reference R, G, and B video patterns.

The invention discloses a method for classifying flash memory of high, middle and low order, which comprises the following steps: firstly, using the method of automatically judging and testing the data rate of flash memory, and setting double data rate or single data rate according to the result; if the route is set as single data rate, Firstly, the bad line test is performed, and the flash memory is set to single-plane detection mode, then the flash memory is erased, and the read-write comparison classification test is performed, and finally the flash memory is classified as low-level flash memory; if the double data rate route is set, first Perform a bad line test, set the flash memory to multi-plane detection mode, then erase the flash memory and record the blocks that failed to erase, perform a read-write comparison classification test and record the blocks that failed to write; judge whether there is an erase failure block; judge whether there is a block that fails to write; judge whether there is a bad line in use, and classify it into mid-level flash memory and high-level flash memory respectively. This invention can efficiently classify high-end, high-end and low-level flash memory. nearly double.

The invention discloses a non-supervised clustering method for classification of spectral remote sensing images with a large amount of data. The original data is divided into several data blocks, and the cluster centers of each data sub-block are obtained by the peak density search method; the cluster centers are re-divided into several data blocks, and clustered again by the peak density search method to reduce the number of cluster centers; repeat In the block-clustering process, a two-dimensional matrix can be used to represent the similarity between any two cluster centers, and then the final classification result can be obtained. The advantage of the method of the present invention is: good applicability, not only can be used for classification of hyperspectral remote sensing images with more spectral segments, but also suitable for classification of multi-spectral remote sensing images with fewer spectral segments or hyperspectral remote sensing images after spectral segment selection; The operation efficiency is high, and the block processing reduces the calculation redundancy of the similarity matrix, and because the clustering processing of each data block is independent of each other, parallel processing can be used to speed up the classification rate.

The invention discloses an automatic classification method for algae cell images. The automatic classification method includes the following steps: preprocessing the algae image, obtaining the processed algae image; and extracting the visual features of the algae image according to the processed algae image, Obtain the feature vector represented by multi-feature fusion; classify all algae cells according to category, and extract a specific proportion of cell images in each type of cell as a training set for subsequent model learning; construct and solve the objective function of the multi-task learning mode; use training The resulting model makes predictions on the test data. The invention reduces the computational complexity and improves the classification rate of algae cells.