Patents
Literature
Hiro is an intelligent assistant for R&D personnel, combined with Patent DNA, to facilitate innovative research.
Hiro

130results about How to "Improve image recognition accuracy" patented technology

An image recognition and recommendation method based on neural network depth learning

The invention provides an image recognition and recommendation method based on neural network depth learning. The method obtains pictures and classification from an image database, inputs to a convolution neural network, trains the neural network through repeated forward and backward propagation, improves image recognition accuracy, and extracts a 20-layer neural network model. By using this model, the object recognition and classification is carried out by collecting static pictures. Results are recognized, and by combining with the personalized characteristics of the input, the input probability of interest is analyzed. By using the machine learning model based on the effective recognition and classification of the material cloud database, and using the recommendation system algorithm, the predicted content material is pushed to the image inputter for cognitive learning. The method of the invention has the advantages of high image recognition rate, multiple recognition types and accurate content recommendation, and can be applied to the electronic products of a computer with a digital camera, a mobile phone, a tablet and an embedded system, so that people can photograph and recognize the objects seen in the eyes and actively learn the knowledge of recognizing the objects.
Owner:广州四十五度科技有限公司

Radar and video fused large-scale monitoring system and method

InactiveCN105376538ARealize all-round continuous monitoringAccurate trackingClosed circuit television systemsVideo monitoringRadar
The invention discloses a radar and video fused large-scale monitoring system and method. The device comprises a central control device and monitoring terminals which are connected with the central control device and respectively arranged at needed monitoring areas; the monitoring terminals comprise a controller, a video monitoring module and a radar monitoring module composed of more than one radar monitor, wherein the video monitoring module and the radar monitoring module are respectively connected with the controller; all radar monitors are arranged at the needed directions where the needed monitoring areas can be completely covered within each corresponding coverage area; when a target radar monitor monitors a target, the controller controls the video monitoring module to adjust an acquisition direction and transmit an acquired target image to the central control device to perform recognition and tracking. The method uses the system to realize large-scale intelligent monitoring. The radar and video fused large-scale monitoring system and method provided by the invention have the advantages that omnidirectional intelligent monitoring of a large-scale area can be realized, the target can be quickly and accurately tracked and recognized and the monitoring precision is high.
Owner:湖南纳雷科技有限公司

Method for locating face landmarks in an image

A method for locating face landmarks (e.g., eyes, nose, etc.) from an image is provided. The method comprises preprocessing an input image for alignment; comparing the aligned input image with a reference image located with face landmarks; calculating distances of pixels and pixel rows of the images; finding a correspondence between pixel rows of the reference image and that of the input image; and using the correspondence and the face landmarks of the reference image to find face landmarks of the aligned input image.
Owner:IND TECH RES INST

Teaching assisted method and teaching assisted system using method

The invention discloses a teaching assisted method and a teaching assisted system using the method. In the teaching assisted method, a trained depth tensor column network model is used to detect behaviors of students in a classroom image, so that the image recognition accuracy is higher and hardware requirement of an algorithm is reduced; the method can be used in embedded equipment to reduce theusing cost of the teaching assisted method; and the teaching assisted system using the method has the same advantages.
Owner:SHENZHEN INST OF ADVANCED TECH CHINESE ACAD OF SCI +1

Solar energy powered intelligent flowerpot and using method

The invention discloses a solar energy powered intelligent flowerpot and a using method. The solar energy powered intelligent flowerpot comprises a pot body, a lifting support and a moving plate, wherein the pot body comprises an inner pot and an outer pot; a temperature sensor and a humidity sensor are arranged on the wall of the inner pot; the inner pot is divided into an upper layer and a lower layer by a sealed partition; a water tank is arranged between the upper layer of the inner pot and the outer pot; a water pump, a singlechip and a storage battery are mounted on the lower layer of the inner pot; a solar cell panel and an illumination sensor are distributed on the upper surface of the moving plate; an LED light source and a water spraying hole are arranged on the lower surface of the moving plate; the water pump is connected with the water spraying hole through a water tube; the temperature sensor, the humidity sensor and the illumination sensor are connected with the singlechip; and the water pump, the lifting support and the LED light source are controlled by the singlechip. Temperature, humidity and illumination can be detected automatically, lighting supplementing and watering are carried out automatically, and therefore, a potted plant can survive normally under the condition that the potted plant is not attended by people in a short time; and the solar energy powered intelligent flowerpot not only can be used as the flowerpot, but also can be used as a flowerpot type desk lamp.
Owner:HUNAN INSTITUTE OF SCIENCE AND TECHNOLOGY

Neural network training method, device and system, and storage medium

The invention provides a neural network training method, device and system for image recognition, a storage medium and an image recognition method based on a neural network. The training method comprises the steps of receiving sample images of category labels marked with N image classification levels, wherein the category of the next level in the adjacent image classification levels is the sub-category of the category of the previous level, and N is an integer larger than 1; training the neural network by utilizing the sample images, wherein based on the classification labels of the image classification levels, the classification losses for the sample images are calculated respectively to serve as the level losses of the image classification levels; calculating the total loss according tothe level losses of the image classification levels; and training parameters of the neural network by using the total loss as an objective function. According to the technical scheme, the neural network capable of more accurately performing image recognition is obtained.
Owner:BEIJING KUANGSHI TECH

Image recognition method and apparatus

The application discloses an image recognition method and apparatus. The method includes: obtaining a standard image and the position coordinate of the standard image by running a test script; determining, according to the position coordinate, a first image with a preset size in a test terminal screen; determining a second image in the first image by subjecting the first image and the standard image to feature matching; identifying a target image corresponding to the standard image by comparing the second image with the standard image. The method and apparatus improve image recognition accuracy when running the test script in different test terminal devices for performing an APP automatic test through a dual recognition process of feature matching and image comparison.
Owner:BEIJING YUNCE INFORMATION TECH CO LTD

Underwater image enhancement method based on adaptive histogram stretching of different color spaces

The invention relates to an underwater image enhancement method based on the adaptive histogram stretching of different color spaces, and the method comprises the steps: carrying out the equalizing of GB channels based on the Gray-World theory, and initially obtaining an adjusted underwater image; analyzing the distribution features of the RGB channels of the underwater image and the attenuation conditions of RGB channels during propagation in water, proposing an adaptive histogram stretching method in an RGB color space based on the automatic parameter obtaining, and employing a bilateral filter for the three channels so as to reduce the impact from noise; converting the RGB color space into a CIE-Lab color space, and carrying out the adaptive histogram stretching of the three components 'L', 'a', and 'b'; finally and enabling the enhanced underwater image to be high in in contrast, to be balanced in saturation and to be high in brightness. The method is lower in calculation complexity of image enhancement, and is mainly suitable for various types of underwater images, such as the images of underwater creatures, underwater fishing and target detection.
Owner:SHANGHAI OCEAN UNIV

Text clustering method based on weak supervised deep learning

The invention discloses a text clustering method based on weak supervised deep learning. The method comprises the following steps: (1) by means of an image data set with text click information, imagevisual information and image category labels are utilized, and adopting image amplification and clustering to construct an image category click characteristic matrix of each text; And (2) obtaining asmooth image click feature map on the initial class click matrix by using a sorting and propagation method. Performing text clustering on the feature map to obtain an initial text category, and initializing text weight by utilizing click priori; (3) under the condition of minimizing an intra-class mean square error, building a deep text clustering model to learn deep text characteristics; (4) performing joint optimization on the depth model and the text weight by using a weak supervised learning method, and iteratively updating the depth model and the text weight; (5) deep text features are extracted through the deep text model, and K-based text feature extraction is achieved. And clustering the means method. The method has very high universality, and the semantic gap in image recognitionis effectively solved.
Owner:HANGZHOU DIANZI UNIV

Image recognition method and apparatus

The present disclosure discloses an image recognition method and apparatus, and belongs to the field of computer technologies. The method includes: extracting a local binary pattern (LBP) feature vector of a target image; calculating a high-dimensional feature vector of the target image according to the LBP feature vector; obtaining a training matrix, the training matrix being a matrix obtained by training images in an image library by using a joint Bayesian algorithm; and recognizing the target image according to the high-dimensional feature vector of the target image and the training matrix. The image recognition method and apparatus according to the present disclosure may combine LBP algorithm with a joint Bayesian algorithm to perform recognition, thereby improving the accuracy of image recognition.
Owner:TENCENT TECH (SHENZHEN) CO LTD

Image recognition method combining convolutional neural network and gradient lifting tree

The invention discloses an image recognition method combining a convolutional neural network and a gradient lifting tree, and belongs to the technical field of mechanical learning. According to the method, a group of image patches are obtained according to the size of an input image. The image patches and an original image serve as input data together. Five branches are adopted. Each branch employs an improved VGG-19 model to extract features of an input image. Then, features are inputted to a gradient boosting tree for training to acquire a basic classifer. A weighted voting method is used for combining the basic classifier into a strong classifier to be used for classifying the input image. According to the method, medical images are identified and proved, the input medical images can beaccurately and rapidly classified, doctors are assisted to diagnose diseases, the diagnosis efficiency is improved, and therefore the misdiagnosis rate is effectively reduced.
Owner:HARBIN UNIV OF SCI & TECH

Quantum neural network method and system for image recognition and medium

The invention provides a quantum neural network method and system for image recognition, and a medium, and relates to the technical field of quantum neural network methods, and the method comprises the steps: quantum state coding: preprocessing MNIST data set data, and converting the MNIST data set data into angle information corresponding to the operation of a rotating door according to a processing result; constructing a quantum neural network: optimizing the learning process of the quantum neural network through decomposition calculation; and characterizing a measurement result: finding out the quantum state with the maximum probability through the measurement result to realize image recognition. According to the invention, the structure logic of the quantum neural network is clearer, the realization is easy, and the learning efficiency is higher.
Owner:EAST CHINA INST OF COMPUTING TECH

Benign and malignant pulmonary nodule identifying apparatus and method based on improved capsule network

The invention discloses a benign and malignant pulmonary nodule identifying system and method based on an improved capsule network, which are characterized in that a new activation function operation(shown in the description) is proposed in a full connection structure of the capsule network, wherein vj is an output vector, and sj is an intermediate vector after dynamic routing adjustment. The beneficial effects are that the activation function can improve the convergence speed of the capsule network and the image identifying accuracy thereof; meanwhile, based on the advantages of the capsulenetwork on the image identifying of the convolution neural network, the pulmonary nodule identification effect of the benign and malignant pulmonary nodule identifying system and method is greatly superior to that of the traditional recognition tools.
Owner:SOUTHWEST UNIVERSITY

System for collecting growth information of crops in greenhouse

A system provides a collecting growth information of crops in greenhouse. In view of the above, it is possible to estimate the growth and yields of crops depending on the size of the greenhouse and the number of the crops in the greenhouse by collecting growth information of the crops such as plant lengths, leaf areas, internode lengths, fruit color, and the number of fruits of the reference crops.
Owner:ELECTRONICS & TELECOMM RES INST

Training method and device, recognition method and device, equipment and medium

The invention discloses an image recognition model training method and device, an image recognition method and device, equipment, a medium and a program product, and relates to the field of artificial intelligence, in particular to the computer vision and deep learning technology. According to the implementation scheme, a sample image and annotation information thereof are acquired, and the annotation information comprises initial annotation information annotated according to positive and negative sample dimensions and at least one kind of fine-grained annotation information obtained by dividing positive samples in the sample image according to different fine-grained dimensions; the sample image is input into a pre-built image recognition model, with the image recognition model comprising at least two independent convolution layers, and the different convolution layers being used for extracting feature vectors of a feature map of the sample image from different dimensions; and according to the annotation information of different dimensions of the sample image, supervised training is performed on the image recognition model by using a loss function corresponding to each dimension, and the loss function is used for returning to the convolution layer of the corresponding dimension. The invention can improve image recognition precision.
Owner:BEIJING BAIDU NETCOM SCI & TECH CO LTD

Method, system and device for recognizing and locating medical slide based on machine vision

The invention discloses a medical slide identification and positioning method based on machine vision, a system and a device thereof, comprising: 1) photographing the slide, obtaining a target area image of the slide, and performing expansion and erosion operation on the target area image to smooth the image; 2) converting the image processed in the step 1) into a Mat type, and carrying out gray-scale conversion operation and thresholding operation on the Mat type image to obtain a clear target region image; 3) establishing pixel coordinates of the target area, drawing a minimum circumscribedrectangle and a center point of the target area image, and outputting the target image coordinates. The invention obtains the positioning coordinates of the slide through the image capturing picture processing of the slide, has the advantages of simple operation, high positioning accuracy, and can further cooperate with the four-axis manipulator to operate, thereby effectively avoiding the error caused by the irregular behavior of the operator in the operation process.
Owner:QILU UNIV OF TECH +1

Image recognition method and system based on deep learning

The present disclosure discloses an image recognition method and system based on deep learning. The image recognition method trains a recognizing engine for high resolution images and a recognizing engine for low resolution image separately. The two recognizing engines generate two independent feature groups specifically for high resolution and low resolution images respectively. A categorizing engine categorizes a testing image to a high resolution mode or a low resolution mode, and then the testing image is recognized by the appropriate recognizing engine. Therefore, the present disclosure increases the accuracy of image recognition under various conditions.
Owner:PIXART IMAGING INC

Capsule endoscope image recognition method and device based on deep learning and medium

The invention provides a capsule endoscope image recognition method and device based on deep learning and a medium. The method comprises the steps of collecting N original images through a capsule endoscope according to a time generation sequence; segmenting the N original images into M groups of original image sequences with the same size by adopting a sliding window segmentation method; analyzing the N original images or analyzing the M groups of original image sequences to form M groups of RGB image sequences, and analyzing the N original images or analyzing the M groups of RGB image sequences to form M groups of optical flow images, wherein each RGB image sequence is composed of image data in an RGB format, and each optical flow image sequence is composed of image data formed by calculating optical flow fields of adjacent RGB images; and respectively inputting the RGB image sequences and the optical flow image sequences into a 3D convolutional neural network model to output an identification result, wherein the identification result is a probability value of occurrence of a preset parameter. The image recognition precision is improved.
Owner:安翰科技(武汉)股份有限公司

Adversarial sample defense method based on Bayesian convolutional neural network

The invention discloses an adversarial sample defense method based on a Bayesian convolutional neural network. The method comprises the following steps: selecting a plurality of traffic signal board pictures as a picture training set and an initial training set according to a traffic signal recognition task of an automobile automatic driving image recognition system; constructing a Bayesian convolutional neural network model of the automobile automatic driving image recognition system, and training the model to determine model parameters; setting a disturbance value and a disturbance value increasing step length, and generating a plurality of adversarial samples; taking the adversarial sample as training set data, and training the model in combination with the initial training set to update model parameters; and improving the automobile automatic driving image recognition system based on the updated model parameters. According to the method, adversarial training is performed on the neural network model by mixing the adversarial samples generated under different disturbance values so that the model is enabled to learn more features, the robustness of the model can be effectively enhanced and thus the recognition precision of the automobile automatic driving image recognition system can be enhanced.
Owner:NAT INNOVATION INST OF DEFENSE TECH PLA ACAD OF MILITARY SCI

Locomotive overhaul system and method based on video identification realized by hologram dual cameras

The invention discloses a locomotive overhaul system based on video identification realized by hologram dual cameras. The system comprises an upper monitoring terminal and mobile video acquisition equipment used for acquiring video information of each overhauled object on an overhauled locomotive, wherein the mobile video acquisition equipment comprises a casing, a dual-camera video shooting device and an electronic circuit board; a data processor, an image synthesis unit and a clock circuit are arranged on the electronic circuit board; for the main and auxiliary holographic cameras, all that is needed is to start the main camera under most shooting scenes, when an auxiliary camera is started, multi-angle information of images can be acquired, and multi-angle holographic images are obtained through processing of main and auxiliary images. The system is simple in structure, simple and convenient to operate and good in use effect, can simply and rapidly complete a locomotive overhaul process and has the advantages of simple steps, reasonable design, high overhaul speed and good overhaul effect, and the implementation is convenient.
Owner:XIAN YINSHI TECH DEV CO LTD

Video analysis system based on deep learning

The invention relates to the technical field of fire fighting monitoring and provides a video analysis system based on deep learning. The system comprises a camera, a video analysis terminal in communication connection with the camera, a video management server, and a video analysis deep learning server in communication connection with the video analysis terminal and the video management server, wherein the video analysis deep learning server is used for acquiring image data at a fire fighting monitoring point through a first recognition model camera obtained by neural network training and transmitting the image data to the video analysis terminal, and the image data comprise video data and / or image data; and the video analysis terminal is used for receiving the image data transmitted by the camera, carrying out classification and recognition on the image data through the first recognition model, acquiring a danger recognition result and transmitting the result to the video managementserver. The deep learning method is adopted to improve the image recognition accuracy, false or missed detection is avoided maximally, and the practicability and the usability are strong.
Owner:ZDST COMM TECH CO LTD

Image identification system, central meter reading data system with image identification system, and remote central meter reading method

The invention discloses an image identification system capable of improving image identification accuracy. The image identification system comprises a file server, a picture processing module, a digital transmission module, and a learning training library; the file server is used for storing a picture received by an input end and data transmitted from a digital transmission module; the picture processing module is used for processing the picture in the file server at fixed time; the digital transmission module is used for judging the output of the picture processing module, and transmitting the output of the picture processing module to the file server or the learning training library according to a judgment result; the learning training library is used for performing depth learning on thedata transmitted from the digital transmission module, continuously extracting data features, and transmitting the features to the picture processing module, thereby enhancing the processing capacityof the picture processing module. The invention further discloses a central meter reading system with the image identification system, the central meter reading system is strong in adaptability and intelligent in management; the invention further discloses a remote central meter reading method which is high in accuracy, convenient and feasible.
Owner:深圳蜜獾智抄科技有限公司

MaskRCNN-based substation equipment anomaly recognition and positioning method and system

The invention discloses a MarkRCNN-based substation equipment anomaly recognition and positioning method and system. The method comprises the steps: collecting real-time data of power equipment, and carrying out the preprocessing; recognizing an abnormal value of the preprocessed operation data by using an iForest algorithm, and marking the abnormal value in combination with a Kmeans clustering strategy; constructing a Mask RCNN target recognition network model based on a convolutional neural network; inputting the marked abnormal value into the Mask RCNN target recognition network model for preliminary recognition, and outputting a target recognition result; training and parameter optimization are carried out on the LSSVM, precision requirements and threshold values are set, and a positioning model is output after training is completed; and importing the target recognition result into the positioning model to obtain abnormal position information of the power equipment. According to the invention, while the image recognition accuracy of the power transformation equipment is greatly improved, the positioning recognition of the abnormal position of the equipment is improved, and thefault-tolerant capability, the positioning efficiency and the accuracy of the fault positioning information are improved.
Owner:SHANGHAI HENGNENGTAI ENTERPRISE MANAGEMENT CO LTD PUNENG ELECTRIC POWER TECH BRANCH +1
Who we serve
  • R&D Engineer
  • R&D Manager
  • IP Professional
Why Patsnap Eureka
  • Industry Leading Data Capabilities
  • Powerful AI technology
  • Patent DNA Extraction
Social media
Patsnap Eureka Blog
Learn More
PatSnap group products