67results about How to "Reduce misclassification" patented technology

The invention discloses a hyperspectral remote sensing image SVM classification method by combining spectrum and texture features and a hyperspectral remote sensing image SVM classification system thereof. The method comprises the following steps that S1, original hyperspectral images to be classified and a ground survey data sample set are inputted; S2, the image elements of the corresponding coordinate positions in the original hyperspectral images are extracted so as to form a reference data sample set; S3, a training sample set is randomly selected for each ground feature class; S4, principal component transformation is performed, and first principal component images are extracted; S5, a region segmentation image is acquired; S6, filtering images are acquired; S7, statistics of spectrum feature information and texture feature information of each segmentation region are performed; S8, a support vector machine model is solved; S9, the original hyperspectral images are classified so that the classified hyperspectral images are obtained; and S10, the classified images are outputted. The new strategy for combining the spectrum and texture features is provided so that the hyperspectral image classification precision can be effectively enhanced.

The present invention discloses a pedestrian repeat identification measurement learning method. The method comprises the following steps: establishing the set of pedestrian target features in different cameras; employing the Mahalanobis distance measurement mode, adding the constraint condition to restrain the feature measurement distance of the same target in different cameras to be zero and restrain the feature distance of different targets to be a constant [Mu] ([Mu]>0); and finally, establishing a loss function and optimizing the structure, obtaining an optimal measurement matrix satisfying the constraint condition through iteration by employing the projection gradient descent method, and completing the measurement learning process. The pedestrian repeat identification measurement learning method can effectively solve the problems of the overfitting phenomenon and measurement matrix sensitive to the noise in the current measurement learning method, and is suitable for the pedestrian repeat identification application occasions in complex scenes.

The invention belongs to the technical field of SAR (synthetic aperture radar) image change detection, and discloses an SAR image change detection method based on a support vector machine and a discriminative random field. The SAR image change detection method based on the support vector machine and the discriminative random field includes the steps: normalizing gray values of two original time phase images, and extracting corresponding gray characteristic differences and textural characteristic differences in the processed images; forming difference characteristic vectors; extracting boundary strength of each pixel in a difference image by the aid of weighted average ratio operators; selecting training samples in the difference image, and expressing the training samples by the aid of the corresponding difference characteristic vectors to obtain initial category labels of testing samples and posterior probabilities of the category labels of the testing samples by the aid of the training support vector machine; obtaining initial support vector machine-discriminative random field models; updating the support vector machine-discriminative random field models to obtain final category labels and change detection results of the corresponding testing samples.

The invention discloses a multiscale hierarchical processing method for extracting object-oriented high-spatial resolution remote sensing information. The method comprises the following steps: at the global level, determining a texture sampling interval and a template window size by utilizing image semi-variance statistical calculation, and dividing the images into several local areas with smooth textures in a course scale on the basis of image texture characteristics; at the local level, setting local segmentation scale parameters through space statistical calculation by taking the local areas as units, and carrying out fine segmentation by utilizing the geometric and spectral information of the images so as to obtain refined image objects which are capable of embodying more details; and carrying out object-oriented remote sensing image classification by taking the global or local areas as units through utilizing a global or local area sample training classifier. According to the method provided by the invention, multiscale hierarchical processing is carried out, so that both the macro and micro characteristics of the images are considered and the ground features can be divided more accurately; and the specific characteristics of the remote sensing images are combined to decide whether to partition to carry out image classification, so that the classification accuracy of the whole image is improved.

The invention relates to an unsupervised Markov random field image segmentation method, which is technically characterized in that: the scope of the classification number of an image M is firstly determined, the image M is then carried out the K-means segmentation from m min to m max type, the BICm which is corresponding to the image M under each type is calculated according to the BIC criteria, a potential energy function is applied to calculate the total energy of the image M, and finally the ICM method is selected to complete the Markov random field image segmentation. The method has the advantages that: the new potential energy function is applied to image segmentation, the anti-noise performance thereof is obviously improved, the segmentation result can remove most of the noise of the image with great noise, the error classification phenomenon can be simultaneously effectively reduced, the error classification rate is reduced by more than 60 percent; the classification number does not need to be manually determined, thus basically realizing the unsupervised image segmentation; and the judgment accuracy of the classification number can achieve more than 95 percent by the artificial synthetic image test.

The invention discloses a layered random forest model-based copper-nickel sulfide ore deposit mineralization prediction method, which comprises the following steps of: S1, collecting multivariate geological data in an area, and establishing a copper-nickel sulfide ore deposit geology information database; S2, analyzing the mineralization rule of the copper-nickel sulfide ore deposit in the region,and extracting ore finding information; S3, selecting non-mineral points, constructing a training sample set in combination with known mineral points, and trainin a layered random forest model; S4, optimizing the hierarchical random forest model, and performing mineralization prediction by using an optimization model; and S5, verifying a prediction result, and evaluating the importance of the prospecting information. According to the method, the layered random forest model is used for carrying out mineralization prediction on the copper-nickel sulfide ore deposit, and the problems of ore point misclassification and prediction accuracy reduction caused by imbalance of training samples can be effectively solved, so that the mineralization potential of the copper-nickel sulfide ore deposit in the region is evaluated more objectively and accurately, and a foundation is laid for the next exploration and development work.

The invention discloses a semi-supervised text sentiment classification method based on random feature subspace. The method comprises steps as follows: 1 acquiring and pre-processing comment text data and constructing a global feature set; 2 expressing all comment texts in a vector form; 3 marking part of comment texts and acquiring a marked sample set and an unmarked sample set; 4 calculating feature weight of all feature words in the global feature set; 5 constructing random subspace; 6 performing cooperative training with unmarked samples to obtain Z classifiers finally; 7 integrating the Z classifiers in a main voting manner to obtain a final integrated classifier. The method solves the problems that a large number of misclassified samples exist in the training process with traditional cooperative training algorithms and the difference between classifiers is small in semi-supervised text sentiment classification methods, so that the accuracy of the text sentiment classification method is improved.

A method is described for determining a change in an object or class of objects in image data, wherein the method comprises: receiving a first image data set of a geographical region associated with a first time instance and receiving a second image data set of the geographical region associated with a second time instance; determining a first object probability map on the basis of the first image data set and a second object probability map on the basis of the second image data set, a pixel in the first and second object probability maps having a pixel value, the pixel value representing a probability that the pixel is associated with the object or class of objects; providing the first object probability map and the second object probability map to an input of a neural network, preferably a recurrent neural network, the neural network being trained to determine a probability of a change in the object or class of objects, based on the pixel values in the first object probability map and in the second object probability map; receiving an output probability map from an output of the neural network, a pixel in the output probability map having a pixel value, the pixel value representing a probability of a change in the object or class of objects; and, determining a change in the object or class of objects in the geographical region, based on the output probability map.

The invention provides an image classification method, an image classification device, a storage medium and electronic equipment, and relates to the technical field of artificial intelligence. The method comprises the following steps: extracting intermediate features of a to-be-processed image by utilizing a pre-trained image classification network; determining the similarity between the intermediate feature of the to-be-processed image and the reference features of each category, wherein each category is each image classification category associated with the image classification network; matching the similarity between the intermediate feature of the to-be-processed image and the reference feature of each category with predetermined similarity probability distribution of each category toobtain a classification result of the to-be-processed image, wherein the classification result comprises the step of determining that the to-be-processed image belongs to a target category in the categories, or determining that the to-be-processed image does not belong to the categories. According to the disclosure, the accuracy of image classification can be improved, and particularly, the misclassification situation for out-of-class images can be reduced.

The invention discloses a high-spectrum image segmentation method based on pixel space information, mainly solving the problem that similar physiognomies can not be favorably segmented by the prior method. The high-spectrum image segmentation method comprises the following steps: solving and normalizing a pixel characteristics matrix and a pixel space Euclidean distance matrix of high-spectrum data; weighting the pixel characteristics matrix and the pixel space Euclidean distance matrix, adding the two weighted matrixes to form a joint dissimilarity matrix and adjusting weighted parameters toacquire a plurality of groups of joint dissimilarity matrixes; using an isometric mapping algorithm to reduce the dimension of each group of joint dissimilarity matrix and acquiring a plurality of groups of mapping results; counting and analyzing each group of mapping result, finishing the primary segmentation of a high-spectrum image; and carrying out category correction to primarily segmented boundary points to acquire a final image segmentation result. The method can effectively find the nuance of different physiognomies in the high-spectrum image and can be applied to martial object recognition, mineral exploration and environmental condition analysis.

The invention discloses a convolutional neural network emotion image classification method combining emotion category attention loss, and relates to the technical field of intelligent media calculation and computer vision. The method comprises the following steps: firstly, performing category weight calculation on a training sample to obtain an emotion category attention weight vector; secondly, modifying a final classification layer and a loss function of the convolutional neural network according to the emotion category number and the emotion category attention loss; then preprocessing the training sample and then transmitting the preprocessed training sample into a network, so that the network achieves convergence after iterative updating of parameters by a loss function and an optimizer, and training is completed; and finally, sending the preprocessed test image into a network, and calculating the emotion image classification accuracy of the obtained model and the prediction category of the model for the test emotion image. According to the convolutional neural network emotion image classification method, when sentiment classification is carried out on the sentiment image through the convolutional neural network, a classification result more conforming to data set sample distribution characteristics can be obtained in a self-adaptive mode, and training and using of a sentiment classification algorithm in different actual application scenes are facilitated.

The invention discloses an SAR image segmentation method for a bee colony and gray association algorithm on the basis of superpixel blocks. The method comprises the steps of (1) pretreating an SAR image to be segmented; (2) calculating the superpixel blocks; (3) initializing the gray space of each superpixel block; (4) carrying out optimization in each superpixel block through a bee colony algorithm to find the optimal cluster center; (5) initializing the distances between the cluster centers of the superpixel blocks and the cluster centers of the other adjacent pixel blocks; (6) defining the correlation degree of the pixel blocks in a domain and the other adjacent domains according to a weighing formula; (7) repeatedly executing the step (6) until each pixel block is classified into the category with the optimal weight; (8) obtaining a segmentation result. According to the method, the SAR image to be segmented is segmented into the superpixel blocks, so that pixel points in the image have position information restriction, the image cannot be disordered, and the local integrity of the image is improved.

The invention discloses a high-spectral unsupervised classification method for constructing a generic dictionary based on confidence degrees. The method comprises the steps of firstly, constructing a two-dimensional spectral-pixel matrix; performing row and column standardization processing; performing feature extraction and selection to obtain dimension reduction features of pixels; performing coarse classification and confidence degree assessment, namely, classifying the pixels by utilizing the dimension reduction features, calculating Euclidean distances between the spectral pixels and a coarse classification category center to serve as the confidence degrees, and obtaining high-confidence-degree classified samples and low-confidence-degree classified samples; and finally, performing secondary classification based on kernel sparse representation, namely, forming the generic dictionary by the high-confidence-degree classified samples, performing the kernel sparse representation on the low-confidence-degree classified samples, and determining classification tags of the low-confidence-degree spectral pixels. According to the method, the problems of low classification sub-space description precision and excessively high computing complexity due to construction of the dictionary by directly utilizing all spectral data are solved; the dictionary sub-space identification performance is improved; and the misclassification error rate is reduced.

The invention relates to a high-resolution remote sensing image city function partitioning method based on multi-feature fusion. The method comprises the steps of 1, image preprocessing; step 2, distributing the feature values in each image to the visual word most similar to the feature value, and counting the corresponding word frequency of each visual word to form a visual word feature; constructing a multi-feature BoW visual dictionary; 3, constructing an LDA probability topic model, and mining a high-dimensional semantic vector of the image by utilizing the LDA probability topic model; 4,training an SVM classifier according to the high-dimensional semantic vector obtained in the step 3; and step 5, carrying out urban function partitioning on the test set by using an SVM classifier. The method has the beneficial effects that the POI data is introduced, so that the wrong score of the remote sensing data caused by metamerism and foreign matters in the same spectrum is reduced; multiple features of the image, which comprises local features, spectral features, texture features, surface temperature features, spatial three-dimensional features and POI features, are comprehensively utilized, and higher classification precision can be obtained under the condition that the single feature of the image is not obvious.

The invention discloses a method for identifying mildewed peanuts through near-infrared hyperspectral images based on a deep learning algorithm, and the method comprises the steps: carrying out the near-infrared hyperspectral imaging of to-be-identified peanuts, and obtaining the data of the near-infrared hyperspectral images; constructing a deep belief network (DBN) model for identifying the mildewed peanuts according to different spectral responses of the mildewed peanuts and the healthy peanuts, and generating a peanut mildewing information distribution diagram; judging whether one to-be-identified peanut particle is mildewed or not according to the mildewing pixel number of each peanut particle in the mildewing information distribution diagram and a threshold value (beta), and generating a mildewed peanut identification result diagram; the method can efficiently and accurately identify the mildewed peanuts, is high in identification precision, and is beneficial to improving the quality of oil extracted from the peanuts and corresponding food.

The invention discloses a malicious traffic detection method in a data imbalance scene. The method comprises the following operation steps: S1, obtaining a training set and a test set: manually collecting traffic of each malicious software sample as a negative sample. The invention relates to the technical field of network information security. According to the malicious traffic detection method in the data imbalance scene, each network flow is identified by using a tetrad, statistical features of each flow are extracted and standardized, port information and the statistical features are combined to serve as feature vectors, and a training set and a test set are formed; in the algorithm level, a k-means clustering algorithm is combined with an SMOTE oversampling technology, noise is avoided only by performing oversampling in a safe area, the emphasis is on imbalance between classes and imbalance in the classes, the problem of minority class sample dispersion is solved by increasing minority class samples in a minority sample sparse area, the minority class sample size is expanded, and good training data is provided for a classification model.

The invention relates to a Landsat TM remote sensing image data cloud removal method. The method includes: inputting Landsat TM remote sensing image data; carrying out multi-scale segmentation on the above-mentioned Landsat TM remote sensing image data; establishing an operating feature ''ThickC'' for the above-mentioned Landsat TM remote sensing image data after the multi-scale segmentation according to a spectroscopic characteristic of thick clouds; classifying objects, which meet a threshold value condition, of the operating feature ''ThickC'' from "unclassified" into ''Thick Cloud''; establishing a relation feature ''ThinC'' for the above-mentioned Landsat TM remote sensing image data after the multi-scale segmentation according to a spectroscopic characteristic and a distribution characteristic of thin clouds; in remaining ''unclassified'' objects, utilizing a double threshold value classification method to classify the objects, which meet a condition, from the ''unclassified'' into ''Thin Cloud''; and uniformity classifying the above-mentioned objects, which are classified into the ''Thick Cloud'' and the ''Thin Cloud'', into ''Cloud'', and completing data cloud removal. The invention also relates to a Landsat TM remote sensing image data cloud removal system. According to the method and the system, influences of the clouds can be effectively reduced or removed, the amount of error classification and leaking classification is reduced, and the classification efficiency and precision are improved.

The invention discloses an analogue circuit fault diagnosis method based on an improved type clone selection algorithm. The method includes the first step of canceling the decision rules that an original decision algorithm determines which diagnosis radiuses faults belong to based on experience, and adopting the minimum Euclidean distance as a diagnosis decision condition, the second step of modifying an affinity calculation formula and utilizing a formula f=1 / (1+d) to replace an original formula f=1 / d so as to prevent overflowing in calculation and standardize the affinity within a fixed range of (0,1], and the third step of modifying an overall affinity calculation mode and utilizing an average value expression method of the affinity of all individuals in a species group to replace a sum expression method of the affinity of all the individuals in the species group. Through the first step, failure switch-off can be eliminated, false switch-off and excessive switch-off can be reduced and the fault diagnosis rate can be improved. Through the second step, calculation is convenient and the comparability of the affinity is higher. The improved type clone selection algorithm is applied to analogue circuit fault diagnosis and has superior performance.

The invention discloses a tidal creek extraction method based on a high-resolution remote sensing image, and the method comprises the following steps: collecting the high-resolution remote sensing image, and eliminating the atmospheric interference of the remote sensing image through radiation calibration and atmospheric correction. The strategy of the method is to extract wide tidal trenches andsmall tidal trenches separately, where the tidal trenches having a width greater than or equal to a predetermined number of pixels are the wide tidal trenches and the tidal trenches having a width less than or equal to the predetermined number of pixels are the small tidal trenches; and the wide tidal creeks are extracted by using a normalized water body index (NDWI) and an OTSU (Out-of-cluster variance) method. the SEaTH algorithm and the J-M (Jeffries-Matusita) distance are adopted to calculate the separability between classes, and the tidal creek and tidal flat are selected for the tidal trench extraction. Aiming at a wave band screened out by an SEaTH algorithm, an improved fuzzy C-means algorithm is utilized to homogenize a heterogeneous background, and then multi-scale Gaussian matching filtering is utilized to enhance a fine tidal creek with a Gaussian shape. And then the small tidal creeks are extracted by using self-adaptive threshold segmentation based on a global mean valueand a standard deviation. And then chippings and plaques are removed in the small tidal creeks. And finally, the small tidal trenches and the wide tidal trenches are combined by adopting trench logicor operation to form a complete tidal trench.

The invention discloses a high-resolution remote sensing image cloud snow identification method and device fusing topographic data and a deep neural network, and the method comprises the following steps: S1, inputting a high-resolution remote sensing image into a DeepLabv3 + semantic segmentation neural network model, and inputting the topographic data of the corresponding position of the remote sensing image into a topographic feature extraction network model; S2, outputting the topographic features extracted by the topographic feature extraction network model to a DeepLab v3 + semantic segmentation neural network model, and fusing the topographic features with deep features in the DeepLab v3 + semantic segmentation neural network model; and S3, enaling the DeepLab v3 + semantic segmentation neural network model to output cloud pixels and snow pixels in the high-resolution remote sensing image. The topographic features are fused through the topographic feature extraction network, and the channel attention module is introduced into the DeepLab v3 + semantic segmentation neural network model, so that in the application of mountainous area cloud snow identification, the model cloud snow identification precision can be improved, the condition of mutual wrong classification of cloud snow can be reduced, and meanwhile, the model prediction time is shortened.

The invention relates to a hyperspectral image classification method based on set empirical mode decomposition of image features. The method comprises the following steps: S1, the original hyperspectral image is dimensionally reduced by using principal component analysis algorithm; S2, adaptive total variational filtering is performed on the obtained 20-dimensional principal components to reduce the sensitivity of noise and obtain rough contour features; 3, each spectral band is decomposed into serial components by using the set empirical mode decomposition, and further the features of the hyperspectral image are fused in the conversion domain; S4, the final classification of the processed image is performed by using a support vector machine classifier. The method can effectively enhance the contour feature of an image, and has better classification performance.

The invention discloses a camera position optimization method based on visibility, and the method comprises the steps: firstly, representing the surface of an object as a sampling point, and parameterizing a potential camera position as a function of a relative posture; secondly, performing point visibility analysis by adopting a hidden point removal algorithm; then, in consideration of visibilityand ray angles between different camera positions, performing camera position optimization through a genetic algorithm, measurement accuracy will be greatly affected according to an accuracy theory in close-range photogrammetry, and then performing optimization circulation until the ratio of the coverage degree of surface points is satisfied. According to the method, the optimized positions of aplurality of cameras are obtained, and the method can be used for automatic manufacturing metering application.

The invention discloses a remote sensing image water body automatic extraction method and device, and belongs to the field of water body monitoring. According to the method, satellite remote sensing images and imaging parameters are obtained and preprocessed, and then, feature information such as NDWI, MSWI, ESWI, Yellow and Canny of a water body is constructed based on spectrum, gradient, statistical features and the like of the images; NDWI and p(NIR) features are utilized to identify a purified water body, then feature information is coupled, and misclassification of clouds, buildings and mountain shadows is comprehensively removed; a yellowness characteristic and a characteristic water body sample are constructed to supplement a high-sand-content water body and other abnormal water bodies; and finally, boundary processing and minimum pattern spot adjustment are carried out to obtain the spatial distribution of the pervasive water body. According to the invention, the problem of wrong division of the water body caused by shadows and clouds of buildings, mountains and the like is solved, the problem of division leakage of abnormal water bodies such as high-sediment-content water bodies and eutrophication is solved, and the limitation of water body information extraction under the background of a traditional single threshold value and complex ground features is overcome.

The invention provides a video classification method and device. The video classification method comprises the following steps: receiving a classification task for a target video; in response to the classification task, extracting video frames of the target video, and generating a video frame set; determining an initial classification result of the target video based on the video frame set; selecting a target video frame from the video frame set; and carrying out key region detection on the target video frame based on the initial classification result, and determining a target classification result of the target video. According to the video classification method provided by the invention, video classification and key region detection are combined, so that videos with similar pictures can be classified more finely, and misclassification of videos can be reduced through key region detection, and the success rate of video classification is improved.

The invention discloses a transfer learning indoor positioning method based on improved TrAdaBoost, and the method comprises the steps: 1, enabling an original scene of a collected fingerprint database to serve as a source domain, and defining a new scene or a scene with a changed internal environment as a target domain; 2, coding the CSI amplitude data after the linear transformation is eliminated by using a One-Hot algorithm; 3, performing cross matching on the processed amplitude data by using a One-vs-Rest algorithm; 4, using a TrAdaBoost algorithm to adjust weights of data of a source domain and a target domain, training a final multi-classifier, and combining fingerprint features of two scenes to construct a new fingerprint library for positioning the target domain; and 5, finally, estimating the position of the test point through confidence regression. According to the invention, the fingerprint database with changed scenes can be updated or the fingerprint database with new scenes can be established at a low cost, and the algorithm complexity is reduced on the premise of ensuring high positioning precision.

The invention discloses an audio classification method and device and a storage medium, and belongs to the technical field of internet. The method includes the steps of acquiring at least one target audio segment in target audio information; performing high-pass filtering and feature extraction on the at least one target audio segment to obtain at least one audio feature corresponding to the at least one target audio segment; based on an audio classification model and the at least one audio feature, determining a classification identifier of the at least one target audio segment, and determining a classification identifier of the target audio information according to the classification identifier of the at least one target audio segment, wherein a first identifier is used for indicating that the corresponding audio information is normal audio information, and a second identifier is used for indicating that the corresponding audio information is sensitive audio information. The high-pass filtering is performed before determining the classification identifier of the target audio information, the low-frequency noise of the target audio information can be filtered out, so that the casewhere the low-frequency noise is determined as the sensitive audio information does not occur, and the accuracy of the audio classification is improved.

The invention discloses an unbalanced data processing method for traffic accident analysis, and the method comprises the steps: firstly obtaining traffic accident data, and carrying out the data combination and data cleaning, and obtaining an original data table; performing grade division on the accident severity in the original data table, and discretizing the influence factors to obtain an optimized data table; then in the optimized data table, according to the grade distribution of the accident severity, dividing positive samples and negative samples, and carrying out unbalanced data resampling on the positive samples to obtain a balanced data table; and finally, inputting the obtained balance data table and optimization data table into a Bayesian classifier, and determining the final classification of the accident case according to the maximum posterior probability. According to the method, the problem that an analysis result is biased to majority class samples when reasoning and prediction classification are carried out through an original unbalanced traffic accident data set is effectively solved, and meanwhile, the recognition capability of minority class samples is kept at an acceptable level.

The invention discloses a terrain adaptive interpolation filtering method suitable for airborne LiDAR point cloud. The terrain adaptive interpolation filtering method comprises the steps of removing low abnormal points, selecting ground seed points, checking the ground seed points, progressively selecting the ground points and the like, wherein in the step of removing the low abnormal points, automatic removal of the low abnormal points is realized by adopting an elevation histogram, so that the manual intervention degree is reduced; in the step of selecting the ground seed points, a large number of ground seed points are obtained by utilizing a one-dimensional discrete smooth spline, so that the precision of an initial ground reference surface is improved; in the step of ground point progressive selection, a to-be-classified point value is estimated by adopting scale-independent interpolation so as to avoid the influence of a spatial position error; and in addition, a terrain adaptivefiltering threshold value is also adopted to adapt to point clouds of different terrain scenes so as to reduce the levels of wrong classification and missing classification of the point clouds of scenic spots of different terrain scenes. The terrain adaptive interpolation filtering method is beneficial to improving the precision of point cloud filtering.

The invention provides a ground cover classification and area calculation method based on a cloud platform and a random forest, a set of multiband remote sensing data preprocessing technology based on MOD09A1 and MCD12Q1 is formed based on a Google Earth Engine cloud platform, and the method mainly comprises the steps of data set selection, data quality evaluation, data cutting, multiband remote sensing data set band synthesis and the like. And a random forest surface cover classification model based on multi-band remote sensing data is constructed, an image reduction function and image spatial-temporal feature information are combined, surface covers in a large range can be simply, conveniently, efficiently and quickly identified in a cloud computing mode, and the area of the regional surface covers is calculated. According to the invention, by fully utilizing the strong computing power of the cloud platform, the land coverings in a large area can be quickly classified, the area can be estimated, and the change trend of the land coverings in a time sequence can be counted.

The invention discloses a camellia oleifera fruit shelling and screening processing method and equipment, and relates to the technical field of camellia oleifera fruit processing. The equipment comprises a supporting frame body; a pre-selection mechanism and a winnowing mechanism are fixedly installed in the supporting frame; the pre-selection mechanism comprises two groups of damping shock absorption pieces which are symmetrically arranged; the bottom ends of the two groups of damping shock absorption pieces are fixedly connected with the supporting frame; the top ends of the two damping shock absorption pieces are fixedly connected with a screening shell; the bottom surface of the screening shell is fixedly connected with two vibration motors which are symmetrically arranged; the inner wall of the screening shell rotationally communicates with a screen drum through a bearing. Through the design of the pre-selection mechanism and the winnowing mechanism, the equipment can efficiently complete the separation operation of camellia oleifera fruit shells and camellia oleifera seeds in an automatic form, and when the equipment is used, a traditional one-time separation structure is changed into a multi-time separation structure.