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

The invention discloses a power terminal loophole attack detection method based on message features, and belongs to the technical field of intelligent power grid terminal equipment safety. The methodcomprises the following steps: S01, acquiring communication message data between power terminal equipment and a master station, and classifying the communication message data of the power terminal equipment in a normal working state and an attacked state into a positive sample and a negative sample; s02, performing feature extraction on the positive sample and the negative sample, and forming a sample feature vector; s03, based on the sample feature vector, selecting a classifier to perform deep neural network training, and generating a vulnerability attack detection model; s04, collecting real-time communication message data between the power terminal equipment and the main station in work; s05, performing feature extraction on the real-time communication message data, and forming a detection feature vector; and S06, inputting the detection feature vector into a vulnerability attack detection model to detect whether the power terminal is attacked or not and the attack type. Accordingto the method, the power terminal equipment is subjected to safety monitoring from a network layer, and the power grid safety is improved.

The invention discloses a vehicle detection and identification method based on an improved DSOD model. The DSOD is improved on the basis of an SSD algorithm, can be simply understood as SSD + DenseNet= DSOD, adopts proposal-free detection model SSD, and adds a DenseNet thought. The DSOD model is divided into two parts, namely a Backbone for feature extraction and a Front-end for target prediction. The Backbone sub-network is similar to a DenseNet, is composed of a layer of Stem block (main module), four layers of Dense blocks (Dense modules), two layers of Translation layers (transition layers) and two layers of Translation w/o pooling layers (transition w/o pooling layers), and is used for extracting image features. The Fort_end sub-network_network (front-end detection sub-network) implements a border frame detection effect through Dense Connetion.

The invention discloses a remote control intelligent garbage can system based on the internet of things technology. A garbage can is connected with a communication host through radio waves, the communication host is connected with a garbage station remote control background through a GPRS network, and the garbage station remote control background is connected with a mobile phone APP client side through a wireless network; and the system improves garbage recycling, and transmits information to a garbage station in time, special supervision by people is not needed, a large number of material resources and a large amount of manpower are saved, and it is ensured that the garbage can can be treated at any time.

The invention discloses a method for optimizing a Parsytec 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 Parsytec on-line surface defect detection system is 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 multi-feature optimization and fusion method for crop planting structure extraction, and the method comprises the steps: collecting a time sequence satellite remote sensing data set which is not greater than the monthly scale, and completing the pre-obtaining of sample data in a research region; describing spectral and texture characteristics of various crops; calculatingexpressions of different samples on spectral information, vegetation indexes, texture characteristic quantities and the like, counting mean values and variances of the characteristic quantities, andcalculating distinguishable capabilities of the different samples on the characteristic quantities; establishing a multi-feature optimization formula, and determining feature quantities participatingin classification and proportions of the feature quantities in the classification process by utilizing the formula; constructing a new image; and performing fine identification on the crop type of theresearch area by utilizing a random forest classifier, generating a space-time distribution thematic map of the crops, and verifying the precision. According to the method, the problem that the timecomplexity and the computer running speed are increased due to the fact that screening of the classification characteristic quantity is ignored in a traditional remote sensing information extraction method is solved.

The invention discloses an image classification algorithm and a system based on manifold learning. The method includes S1, selecting a training sample set and a testing sample set; S2, extracting feature points of image of two sample sets by using a sift algorithm; S3, respectively reducing the dimension of the feature points in the two sample sets by using the local linear embedding or Labras feature mapping in the manifold learning method; S4, inputting that reduced dimension feature point of the training sample set into a support vector machine classifier for training; S5, classifying the test sample set by using the trained support vector machine classifier. The invention combines the SIFT feature extraction algorithm with the non-linear manifold learning dimension reduction algorithm,extracts the middle-level feature of the image, and then classifies and processes the image by using the SVM classifier, thereby effectively improving the computing speed and the classification accuracy.

The invention discloses an image classification method and a device. The method comprises the following steps: obtaining an image set to be classified; A set of images to be classified is inputted into a classification model sequence for classification, The classification model sequence includes at least two classification models arranged in order from a lower level to a higher level, wherein theset of images to be classified is input to a first classification model in the classification model sequence, and a non-noisy image to be classified output from the lower classification model is usedas an input of an adjacent higher classification model; A classification result of the image to be classified by the highest-level classification model in the classification model sequence is outputted. In the present application, classification accuracy is ensured while classification speed is improved in the above manner.

The invention relates to a cooling tower noise monitoring system and method, and belongs to the field of cooling tower systems. The system comprises a signal acquisition unit, a controller and a terminal display, wherein the controller is connected with the signal acquisition unit and receives the signal parameters of the signal acquisition unit; the terminal display is connected with the controller and receives and displays signals of the controller. By means of the cooling tower noise monitoring system and method, components related to cooling tower noise can be monitored in real time, a quantum immune optimization neural network model is utilized to analyze and process the detected data related to the cooling tower noise, a quantum searching mechanism and an immune algorithm clonal selection principle are combined, a primitive population and a cloning subgroup are generated through cloning operation of the neural network algorithm to achieve population expansion, the local searching ability is improved, the optimal model parameter is obtained through good processing and analysis, the parameter data classification probability is increased, the false alarm rate is lowered, and the problem that in the prior art, the root cause of the cooling tower noise is neglected is solved.

The invention discloses a household garbage datamation classification recycling system. The system comprises a garbage recycling system and a garbage sorting part; the garbage recycling system comprises household garbage throwing intelligent equipment and a control system; the household garbage throwing intelligent equipment comprises a garbage can body, a garbage funnel and a garbage trolley; and a display screen, a recognition device, a garbage throwing opening, a garbage door and a pedal are arranged on the garbage can body, a special garbage bag storing and delivering mechanism is arranged in the garbage can body, a set of special garbage bags are stacked in the special garbage bag storing and delivering mechanism, and a special garbage bag opening and closing and garbage throwing opening opening and closing mechanism is arranged below the special garbage bag storing and delivering mechanism. The method has the advantages that the operation is simple, the garbage is convenient to recycle, the use method and the control mode are simple and practical, the classification method for decayed, burnt and dyed and transformed garbage is provided, the classification method is simple and efficient in management, and meanwhile, the user is promoted to accurately classify the garbage in a reward income mode.

The invention discloses an automatic classification method for images of an automobile gear shifting panel. The method comprises the following steps of collecting an image of the automobile gear shifting panel; carrying out binarization, expansion and hole filling, carrying out differential operation, and separating a target area; extracting eleven shape features of the target area and building the structure of an MLP neural network model to acquire model parameters and the trained MLP neural network model; extracting eleven shape features from an automobile gear shifting panel image of an unknown type and inputting the features into the MLP neural network model, and outputting the type of the automobile gear shifting panel. Different types of automobile gear shifting panels in the image can be classified, the classification rate of the automobile gear shifting panels is increased, and the method has good classification robustness and accuracy.

The invention belongs to the technical field of medical instruments and discloses a medical interventional therapy device for medical oncology, and a control method thereof. The device comprises a linear motor. The rotating ring of an output shaft of the linear motor is connected with a sliding connecting rod through a screw structure. The sliding connecting rod is connected with a one-way syringethrough a joint; the lateral side of the unidirectional syringe is clamped with a transverse fixing support; vertical fixing supports are fixed at both ends of the transverse fixing supports; the other end of the unidirectional syringe is connected with a syringe tube through a three-way one-way valve; the other side of the three-way one-way valve is connected with a burette through the one-way valve; the other end of the burette is connected with a medicine storage bottle through a medicine adding valve; a linear motor is wirelessly connecte with a signal acquisition control module; and thesignal acquisition control module is connected with a display module and a control module. The invention realizes the treatment of radiation and injection medicine, saves the treatment time, reduces the pain of the patient, improves the treatment effect and saves the treatment cost.

A side information calculating unit (110) calculates side information for assisting either identification processing or classification processing. When there is a discrepancy between a processing result of either the identification processing or the classification processing, and the side information, the multilayer neural network (120) changes an output value of an intermediate layer (20) and performs either the identification processing or the classification processing again.

The invention discloses a pulse diagnosis five-internal-organ state classification method and device and a storage medium, and relates to the field of intelligent traditional Chinese medicine, and the method comprises the steps: obtaining a to-be-processed data set and coding label data corresponding to five-internal-organ state categories; performing data processing on the to-be-processed data set to obtain a pulse condition signal data set; performing feature processing on the pulse condition signal data set to obtain two-dimensional array data; inputting the two-dimensional array data and the coded label data into a pulse diagnosis five-internal-organ state classification model for classification processing to obtain pulse diagnosis five-internal-organ state classification data; wherein the pulse diagnosis five-internal-organ state classification model is obtained through training according to LightGBM optimal parameters, and the LightGBM optimal parameters are obtained after parameter optimization is conducted on a LightGBM algorithm according to a genetic algorithm. By means of the five-internal-organ state classification method for pulse diagnosis, the accuracy rate of conducting pulse diagnosis classification and recognition on the five-internal-organ state is improved, and the rate of classifying five-internal-organ state data for pulse diagnosis is increased.