76results about How to "Improve image recognition efficiency" patented technology

The invention discloses a convolutional neural network-based cervical cell image recognition method. The method is characterized by comprising the following steps of: (1) preparing a training sample; (2) constructing a convolutional neural network layer; (3) constructing a second classifier; and (4) obtaining a recognition result: inputting a to-be-tested cervical cell image into an improved convolutional neural network so as to be automatically recognized and classified by the improved convolutional neural network. The method is high in automation degree and strong is adaptive ability, not only can improve the cervical cell image recognition efficiency, but also can improve the cervical cell image recognition correctness.

The invention relates to an image recognition method, an image recognition device and an image recognition terminal. The image recognition comprises the steps of: acquiring a first class light source image and a second class light source image to be recognized; respectively inputting the first class light source image and the second class light source image to a trained convolutional neural network model, and determining a first feature of the first class light source image and a second feature of the second class light source image; and inputting class parameters corresponding to the first feature and the second feature to a trained classification model to recognize image classes, wherein the image classes are used for indicating whether an object contained in the first class light source image is consistent with an object contained in the second class light source image. By applying the image recognition method, the image recognition device and the image recognition terminal, the first class light source image and the second class light source image are subjected to classification and recognition through the trained convolutional neural network model and the trained classification model, and whether the object contained in the first class light source image is consistent with the object contained in the second class light source image can be recognized quickly and accurately.

The invention discloses an image identification method based on a Spiking neural network, belongs to the technical field of image processing, and aims to solve the problems of low image identification efficiency and incapability of accurately identifying image features in the prior art. The method comprises the following steps: separating the image features on the basis of a Gaussian differential filtering thought; performing coding with a phase retardation method; and after coding is finished, learning an obtained feature sequence with a Spiking neural network learning algorithm to obtain an identification output result finally. Compared with a conventional Spiking image identification method, the method has the advantages that the image features are refined, local characteristics of images are specifically optimized by means of dimension reduction, integration, separation, extraction and the like, and the Spiking neural network learning algorithm based on a membrane voltage is used, so that the efficiency and accuracy of an identification process are increased. The image identification method is applied to image identification, image classification, image feature extraction and Spiking learning algorithm application, and relates to the fields of machine learning, Spiking neural networks, phase retardation coding and the like.

The invention relates to an image recognition system and method. The method includes the following step of data storage, the step of image recognition, the step of image marginalization, the step of local feature analysis and the step of comparison, wherein in the step of data storage, local features of multiple target images are stored in advance, and a local feature library of the target images is built; in the step of image recognition, images to be recognized are analyzed and extracted in input real-time video streams or streaming medium files; in the step of image marginalization, binary edge images corresponding to the images to be recognized are obtained by means of a dynamic edge analysis method; in the step of local feature analysis, the binary edge images corresponding to the images to be recognized are analyzed, and local features of the images to be recognized are obtained; in the step of comparison, whether the images to be recognized and the target images are similar is determined by comparison through a local feature template, if the images to be recognized and the target images are similar, the images to be recognized are output, and if the images to be recognized and the target images are not similar, the step of image recognition is repeated. By means of the image recognition system and method, image recognition can be efficiently carried out on the real-time video streams or the streaming medium files.

The invention provides an image identification method and apparatus. The method comprises steps of obtaining an image to be processed; according to a preset strategy, obtaining positional information of an edge identifier in the image to be processed; according to the positional information of the edge identifier and a preset shape of a sensitive area, determining the sensitive area in the image to be processed; and identifying an image in the sensitive area by a preset algorithm. The invention firstly determines the sensitive area in a small range by the edge identifier, and then identifies the image in the sensitive area. As the image identification area is narrowed, the image can be rapidly positioned, so that the image identification speed is accelerated, the image identification time is saved, and the image identification efficiency is improved.

The invention provides an image recognition model training method and device and electronic equipment, and relates to the technical field of deep learning, and the method comprises the steps: inputting a training sample marked with a sample label in advance into an image recognition model; in an iterative training process of an image recognition model, determining a fine-grained feature map corresponding to the training sample based on a network layer of the image recognition model, inputting the fine-grained feature map into a preset deep learning network to enable the deep learning network to learn fine-grained feature information from the fine-grained feature map, and distilling the learned fine-grained feature information into the image recognition model; wherein the fine-grained feature map is an image marked with a discriminative region corresponding to the sample label; and repeatedly executing the above training steps until the training is finished, and obtaining a trained image recognition model. According to the invention, the image recognition efficiency of the trained image recognition model can be improved.

The invention discloses an image recognition method, a related device and equipment, and the method comprises the steps: obtaining at least one to-be-recognized medical image obtained through scanning, and respectively determining a target region, corresponding to a target organ, in each to-be-recognized medical image; respectively carrying out feature extraction on the image data of each target area to obtain individual feature representation of each to-be-identified medical image; fusing the individual feature representations of the at least one to-be-identified medical image to obtain a global feature representation; and determining a scanning image category to which each to-be-identified medical image belongs by utilizing the individual feature representation and the global feature representation of each to-be-identified medical image. According to the scheme, the image recognition efficiency and accuracy can be improved.

The invention relates to an image recognition method and device, a storage medium and computer equipment. The method comprises: acquiring a page image to be recognized; Dividing the page image into page sub-images corresponding to the page elements according to an area where the page elements in the page image are located; Determining an image type corresponding to each page sub-image; And identifying each page sub-image according to an identification mode matched with the respective corresponding image type to obtain an identification result corresponding to each page sub-image. According tothe scheme provided by the invention, the image recognition efficiency is improved.

The invention belongs to the technical field of image processing and discloses a remote sensing image target detection method based on cloud computing storage and deep learning. The method in the invention reduces the training time through parallel deep learning, optimizes the training effect, effectively improves the image recognition efficiency and achieves the purpose of real-time recognition by using Storm real-time recognition; and is conducive to improving the recognition speed and accuracy by dividing a remote sensing image into a plurality of segmented sub-regions. Since the method inthe invention adopts a mean shift algorithm for segmentation, the algorithm is a fast and efficient cluster segmentation algorithm which can quickly and accurately obtain segmented sub-regions. In theinvention, after the mean shift algorithm is used for segmentation, a K nearest neighbor method is used for recognition. The method is the simplest among the data mining classification methods, and the operation efficiency is high.

The application discloses a model configuration and image identification method and device; the method comprises the following steps: using a client end to obtain at least one identification parameter from a server, and configuring at least one identification model according to the obtained at least one identification parameter and fixed parameters pre-stored in local. The client end can dynamically obtain identification parameters used for configuring the identification models from the server, so different identification models can be configured according to the obtained different identification parameters; the configured identification models can identify different objects; in an image identification process, each identification model only needs partial the video frames of a video for fast image identification; when the partial video frames are determined not to have the object, another identification model can be employed on partial video frames of the video for image identification.

The invention relates to a system and a method for preventing a full-intelligent computer display screen from being illegally shot by a mobile phone. The system mainly comprises a video monitoring sampling module, a main server and an alarm screen locking module. In the video monitoring and sampling module, a micro camera monitors the working state of an operator in front of a computer display screen in real time, samples in real time to obtain a video image, stores the video image in a video memory and outputs the video image; in the main server, an image processing module receives a real-time video image and carries out image processing, a mobile phone detection module receives an image processing result and carries out mobile phone detection, and a mobile phone identification module receives a mobile phone detection result, carries out mobile phone shape identification and outputs a monitoring identification state; it is determined that mobile phone shapes are stored in the criterion library, and the mobile phone shapes are obtained through learning and training of mobile phone images in video storage; in the alarm screen locking module, the alarm module receives the monitoringidentification state, and if the monitoring identification state is that a mobile phone exists, the alarm module sends an alarm signal; and the screen locking module receives the alarm signal and performs screen locking processing on the operating system.

The invention relates to an image local feature extraction method and device, and belongs to the technical field of image processing. The method comprises the following steps of constructing a Siamesenetwork; selecting an included angle cosine between the output feature vectors of the Siamese network as the similarity between the images; constructing an error function by using the similarity, training the Siamese network by using the sample image, and selecting the Siamese network with the minimum error as an image local feature extraction network; performing image block detection on the twoto-be-extracted whole images, combining the image blocks into image pairs, wherein each image pair comprises the image blocks belongs to the two images respectively, inputting the image pairs into theimage local feature extraction network, and selecting the matched image blocks as the local features of the to-be-extracted whole images. According to the method, the similarity of the image pairs iscalculated by adopting the included angle cosine between the feature vectors, the efficiency of selecting the matched image blocks is improved, and the image recognition efficiency is further improved.

The invention discloses an image classification method based on multi-sample dictionary learning and local constraint coding, and belongs to the technical field of artificial intelligence and the field of image classification. The method comprises the following steps: firstly, generating a virtual training sample and an initialization dictionary for a training sample by utilizing a K-SVD algorithm; learning a dictionary and a coding coefficient of the training sample by using all the training samples and the initialized dictionary; and then learning the coding coefficient of the test sample byusing the learned dictionary, learning the linear classifier coefficient by using the coding coefficient of the training sample, and finally performing classification identification on the coding coefficient of the test sample by using the linear classifier coefficient and outputting a classification result. When the method is used for image classification and recognition, the learned dictionaryand coding coefficient matrix can have better representation capability, and the classification accuracy can be obviously improved.

An embodiment of the invention discloses a surgery image processing method and a surgery image processing device. The surgery image processing method comprises the steps of acquiring a surgery image of a surgery area, wherein the surgery image of the surgery area is acquired by image acquisition equipment; identifying image characteristics of the surgery image, and determining lesion tissue information which is contained in the surgery image according to the identified image characteristics; generating lesion mark information according to the lesion tissue information contained in the surgery image; and displaying the surgery image which contains the lesion mark according to the lesion mark information. Compared with an existing surgery image processing method, the surgery image processing method and the surgery image processing device have advantages of preventing artificial lesion tissue identification by a doctor, improving image identification efficiency, realizing no requirement for relatively high lesion tissue identification capability, and reducing labor cost.

The embodiment of the invention provides an image processing method and device and electronic equipment, and belongs to the technical field of data processing, and the method comprises the steps: obtaining an image set comprising a target area, with the image set comprising one or more sample images with fixed sizes; establishing a feature matrix matched with a labeling result according to the labeling result aiming at a target area in the sample image; constructing a minimized objective function for training a recognition model through the image set and the feature matrix so as to obtain a trained recognition model for recognizing the target area in the image; and after the size of the to-be-identified image is adjusted to a target size, carrying out target area identification on the to-be-identified image by utilizing the identification model to obtain a first feature matrix and a second feature matrix for carrying out target area calculation on the to-be-identified image. Through the scheme of the invention, the image recognition efficiency is improved.

The invention provides an intelligent access control system based on a sensor network. The intelligent access control system based on the sensor network comprises a central processing device, a temperature sensor, a signal processing circuit, a fingerprint acquisition module, a camera, a first trigger module, a temperature identification module, a fingerprint identification module, a second trigger module, n image recognition module, a third trigger module, a gate control module, a non-gate circuit and an alarm module, wherein the temperature sensor is used to confirm the human body entering the access control, and the signal processing circuit is used to improve the temperature monitoring accuracy; when detecting that there is a human body, the fingerprint collection module is triggered to perform fingerprint collection operations; when the collected fingerprint is the fingerprint of the person who is allowed to enter the access control, the camera is further triggered to perform image collection, and then the image recognition module extracts the feature information of the image for further judging whether the person who enters the access control is allowed to enter the access control.

The invention discloses a positioning method, a device, equipment and a computer-readable storage medium. The method comprises collecting a building image of the current position of the user; identifying the building image using a pre-established area image database of the initial positioning area in order to determine position information of the building; according to the position information ofthe building and the position relationship between the user and the building, the target positioning of the current position of the user being determined. The method, the device, the device and the computer-readable storage medium provided by the invention can accurately locate the current position of the user.

The invention relates to an image recognition method and device, computer equipment and a storage medium, and relates to artificial intelligence, and the method comprises the following steps: receiving an image recognition request for a target concept label; obtaining a to-be-identified target image corresponding to the image identification request; inputting the target image into a target image recognition model corresponding to an entity label for image recognition, a target image recognition result set is obtained, an image recognition result in the target image recognition result set is animage recognition result corresponding to a target entity label, and the target entity label is an entity label corresponding to the target concept label; and obtaining an image recognition result corresponding to the target concept label according to the target image recognition result set. The method can improve the image recognition efficiency.

The invention discloses a wild animal information acquisition method and device based on an unmanned aerial vehicle. The method comprises: controlling the unmanned aerial vehicle to obtain remote sensing image data of wild animals; performing feature extraction on the remote sensing image data by using the trained feature extraction neural network model, wherein layers in each module of the feature extraction neural network model are densely connected; utilizing a trained object recognition neural network model to recognize a plurality of feature images of the remote sensing image data obtained by the feature extraction neural network model according to a plurality of preset feature image scales, the plurality of preset feature image scales being compression ratios of the feature images relative to the remote sensing images; and obtaining an identification result according to the object identification neural network model, and determining the species number of wild animals and the number of various wild animals in the remote sensing image data. According to the technical scheme provided by the invention, the wild animal information acquisition efficiency can be improved.

The embodiment of the invention provides an image retrieval method and device and electronic equipment, and belongs to the technical field of data processing, and the method comprises the steps: carrying out the feature extraction of a to-be-retrieved image through employing a first layer group of a first network model, and obtaining a first feature image; performing feature calculation on the target image through a second network model to obtain a feature matrix related to the target image; and based on the feature matrix and the first feature image, performing a retrieval operation by using a second layer group of the first network model, with the retrieval operation being used for determining whether the to-be-retrieved image contains the target image. Through the scheme of the invention, the retrieval of the target image can be realized without marking the training data, and the image recognition efficiency is improved.

The embodiment of the invention discloses an image recognition method and device, electronic equipment and a computer readable storage medium. According to the embodiment of the invention, the method includes: after a face image sample is obtained, carrying out spatial feature extraction on the face image sample, training a preset face recognition model based on extracted image features and face label information, obtaining a trained face recognition model, then carrying out domain feature extraction is carried out on the image features, acquiring category features of a face image sample, determining a feature classification result of the face image sample on the basis of the category features and category label information, performing correcting according to the feature classification result and a trained face recognition model, and recognizing a to-be-recognized face image by adopting the corrected face recognition model. According to the scheme, the image recognition efficiency can be improved.

The invention discloses a data processing method, device and equipment and a readable storage medium. The method comprises steps of obtaining first image data sent by a first client, and storing the first image data in a receiving queue; performing image recognition processing on the first image data in the receiving queue, and in the image recognition processing process of the first image data, storing continuously acquired second image data sent by the first client in the receiving queue to obtain an updated receiving queue; and when a first object area where the target object is located in the first image data is extracted through image recognition processing, sending the first object image data contained in the first object area to the target cloud application server, and synchronously performing image recognition processing on second image data with the latest receiving timestamp in the update receiving queue. According to the method, the image transmission time delay can be reduced, and image recognition efficiency is improved.

The invention relates to a method and system for discriminating effective images of cervical liquid-based cells, wherein, the method comprises the following steps: acquiring an input RGB image, and judging whether a dark dye binary image and a light dye binary image exist in the RGB image; if the RGB image has the dark dyeing binary map and the light dyeing binary map, calculating the proportion of the light dyeing points in the neighborhood of the target dark dyeing points for the target dark dyeing points; determining whether the RGB image is a valid image according to the calculated numberratio. The technical proposal provided by the application can improve the image recognition efficiency.

The invention discloses an image traversal algorithm for rapid image recognition and feature matching. The method comprises: enabling a computer to obtain coordinates of all pixel points on an image and RGB values of corresponding colors, converting graying of the color image into a binary grayscale image, carrying out recognition from the upper left corner of the image, and finding out a pixel point corresponding to a certain pixel point of a feature graph to serve as a recognition point; selecting the identification point as a traversal coordinate origin, and designing the image in a regional and layered manner; and dividing the image into a plurality of regions by taking the traversal coordinate origin as a center, carrying out traversal from a certain region according to a clockwise direction, and searching other pixel points of the feature graph, wherein a traversal coordinate point of each layer of the image is a coordinate point set of the same level of each region. According tothe invention, the new identification point is searched from the coordinate close to the last identification point, so that the efficiency of image identification and feature matching is greatly improved.

The invention relates to a target recognition method and system based on a convolutional neural network algorithm, and relates to the technical field of image processing, and the method comprises the steps: S1, a collection module collects a to-be-processed image; S2, a processing module carries out image processing on the to-be-processed image to obtain a target image; S3, a judgment module judges whether the target image meets the requirement or not; and S4, an output module outputs the target image meeting the requirement. When the processing module carries out image processing, the processing module draws a continuous gray scale boundary line for a to-be-processed image according to the gray scale difference, and when the processing module draws the gray scale boundary line, the processing module sets the gray scale difference according to the hue number B of a to-be-acquired target; and the processing module generates a plurality of target areas according to the gray scale boundary lines, sets similar side lines according to the graph similarity C of the target areas and the target image, and then sets target frame lines according to the similar side lines. According to the invention, the image target recognition efficiency is effectively improved.

The invention relates to the technical field of image processing, and discloses an image recognition method and device based on figure fine-grained features. The method comprises the steps: obtaininga to-be-recognized figure image; performing feature extraction on the to-be-recognized figure image to obtain a figure feature layer; inputting the figure feature layer into a preset super-column feature recognition model to obtain a corresponding image recognition result; obtaining image recognition accuracy according to the image recognition result; and when the image recognition accuracy is greater than or equal to a preset standard threshold, taking the image recognition result as an image recognition result based on the fine-grained characteristics of the person. Compared with the prior art, when an attention mechanism network is used for image processing, key area information cannot be accurately acquired, and image category cannot be accurately identified; and when the figure feature layer is input into the preset super-column feature recognition model, and the key area of the image can be accurately positioned, so that the image recognition result corresponding to the figure image can be quickly and accurately obtained.

The invention discloses a method for determining an area where a sample graph is located in an interface graph and electronic equipment. The method and the electronic equipment are used for solving the problem that the area where the sample graph is located is determined inaccurately in the interface graph. According to the scheme, the area where the zoomed sample graph is located is determined inthe target interface graph on the basis of the sample graph, the width and height values of the source graph containing the sample graph and the position of the sample graph in the source graph, wherein zooming is carried out based on the original size of the sample graph, the zoomed sample graph in the target interface graph can be efficiently matched, and the accuracy of determining the area where the sample graph is located is improved. According to the method, the target area graph is intercepted from the target interface graph, and the zoomed sample graph is searched in the target area graph, so that the calculation amount can be reduced, and the efficiency of determining the area where the sample graph is located is improved. Besides, according to the scheme, the target area graph is intercepted by estimating the width and height values of the zoomed sample graph and referring to the estimated width and height values of the zoomed sample graph and the position of the sample graph, so that the zoomed sample graph is contained in the target area graph, and the situation that the intercepted target area graph contains an incomplete zoomed sample graph is avoided.

The invention relates to an image recognition method, device and system and calculation equipment. The method comprises the steps that firstly, a data center receives a first feature value transmittedby a first edge small station, the data center communicates with the first edge small station through a network, and the first feature value is obtained through the preprocessing of a first image obtained by the first edge small station; determining a first attribute according to the first characteristic value; a first label is sent to the edge small stations in the edge small station set, the first label comprises a target characteristic value and a first attribute, the target characteristic value is a characteristic value associated with the first attribute, and the edge small station set comprises the first edge small stations; receiving at least one image recognition result sent by the edge small stations in the edge small station set; and finally, determining the position of the target object according to each image recognition result. Therefore, the problems that the video data is processed by the data center, the load of the data center is large, and the image recognition efficiency is influenced are solved.

The invention provides a medical image processing method and device based on a pathfinder intelligent search algorithm, and the method comprises the steps: obtaining an input to-be-processed medical image I (i, j), segmenting the image into n windows, and removing the noise of each of the n windows through median filtering, and obtaining a noise-free image IF (i, j); based on a gray level co-occurrence matrix (GLCM) feature extraction method, extracting texture parameters of the image from the noiseless image IF (i, j), wherein the texture parameters comprise the contrast ratio, the correlation degree, the angle second moment, the inverse differential moment (IDM) and the entropy between adjacent pixels, and GLCM features are constructed; extracting important texture features from the constructed GLCM features by using a pathfinder intelligent search algorithm; and inputting the extracted important texture features into a trained kernel extreme learning machine KELM to obtain classification and recognition results of the to-be-processed medical images. According to the scheme of the invention, the accuracy of ultrasonic image recognition can be improved, and the efficiency of ultrasonic image classification and image recognition is improved.

The invention relates to an image recognition method and device, computer equipment and a storage medium. The method comprises the following steps: acquiring an image pair set, wherein the image pair set comprises target image pairs corresponding to a plurality of image categories respectively; inputting the images in the target image pair into a to-be-trained image recognition model for feature extraction to obtain image extraction features corresponding to the images in the target image pair; obtaining the feature similarity between the image extraction features corresponding to the target image pair; obtaining an image pair comparison loss value corresponding to each image category based on the feature similarity of the target image pair corresponding to each image category; carrying out statistics on the image pair comparison loss values corresponding to the image categories to obtain model loss values; and adjusting model parameters of the to-be-trained image recognition model based on the model loss value to obtain a trained image recognition model. By adopting the method, the identification range can be generalized, and the identification accuracy can be improved.