115results about How to "Rich image information" patented technology

The invention relates to a Bayer image compression method based on YUV conversion, which belongs to the technical field of an aerospace optical remote sensor and is a Bayer image compression method used for an aerospace optical remote sensing CMOS camera. The method comprises the following steps: carrying out space conversion on a Bayer-format image from a three-color space to a brightness-chroma space, eliminating the relativity among the three color components to be ready for different compression ratios used by brightness and chroma, self-adaptively determining the compression ratios based on the complexity of the image, carrying out line conversion and encoding. The invention greatly improves the quality of the compressed image by carrying out space conversion to eliminate the relativity among the three color components, and further enhances the adaptive capacity of the camera by using different compression ratios for images of different complexities.

The invention provides an attention mechanism-based image description generation method. Firstly, more accurate image description is generated by using an attention mechanism according to significantobject information in an image and information of a relationship between objects, then, finer-grained image description is generated by using a double-layer language generation model, and finally, thewhole model is optimized by using reinforcement learning. The method has the advantages that image information can be enriched through fusion of the relation features and the object features, the double-layer language model can generate image description with finer granularity, and the problem of exposure deviation can be relieved by further optimizing the training model through reinforcement learning.

The invention discloses a double-wire pulsed welding melten pool bidirectional near-infrared synchronous vision sensing method and device. The method is as follows: the light of electric arcs reflected by the melten pool and the radiation light of the melten pool are used as the light source; the near-infrared technology and the CCD vision sensor parameter adjusting devices are used to obviously reduce the twin-arc light intensity; the simultaneous display of the images of the back and side bidirectional melten pool and the data storage are completed through the bidirectional melten pool simultaneous same-image collection and storage method based on VC++ and OpenCV; and characteristic information reflecting the welding quality can be obtained through digital image processing. The sensing device of the invention comprises two near-infrared CCD vision sensors, two near-infrared compound filter systems, two image acquisition cards, a computer, a bidirectional visual system assembling jig, two high-coaxiality near-infrared compound filter system assembling components and two CCD vision sensor parameter adjusting devices. The simultaneous same-image vision sensing of the double-wire pulsed MIG and MAG welding bidirectional melten pool of the same kind of metal or different kinds of metals such as light alloy, stainless steel, carbon steel, copper steel and aluminium magnesium alloycan be realized.

The invention discloses a three-dimensional beautification method and device of an image. The beautification method comprises the following steps: acquiring a basic image and an auxiliary image corresponding to a current scene, wherein depths of fields and parallaxes of the basic image and the auxiliary image are different; determining a differentiated pixel point of the basic image and the auxiliary image; performing beautification processing on the basic image, and performing image processing on the auxiliary image according to the determined differentiated pixel point; combining the processed basic image and auxiliary image into a three-dimensional image. Through the method and the device, the problem that the picture quality of the beautified image is affected by the fat that fine beautification is unfavorably or cannot be performed due to over-dark environment on the two sides of a human face region and too less image information is solved, and the quality of the image is improved.

The invention relates to an auxiliary assessment method for prognosis of nasopharynx cancer based on enhanced MRI radiomics. The auxiliary assessment method comprises the steps of: (1), performing MRIimage processing; (2), extracting imaging features; (3), screening the imaging features; (4), establishing a radiomics scoring formula; (5), screening clinical risk factors; and (6), establishing a prognostic survival model: establishing a prognostic observation model through combination of a radiomics score and the clinical risk factors of a patient with nasopharynx cancer, performing qualitative and quantitative prediction on the PFS (progression free survival) of the patient, and furthermore, assessing the performance of the prognostic survival model. The auxiliary assessment method in theinvention has little harm to the image examination of the patient; qualitative and quantitative analysis on the survival time of a specific patient is carried out; therefore, a doctor is assisted tomake an individualized treatment and follow-up visit scheme; furthermore, the doctor is assisted to assess the survival and recurrence time of the patient; simultaneously, the performance of the obtained prognostic survival model is verified; and thus, the accuracy of a prognostic prediction model is ensured.

The present invention relates to a remote sensing image contrastive analysis method and system. The method comprises: S1: carrying out identification and segmentation on surface features in two remote sensing images that are taken at different times in the same area through the full convolution network, and obtaining segmentation images of all surface features in two remote sensing images, wherein the full convolution network comprises a plurality of convolution layer groups and a plurality of deconvolution layers, and the convolution layer group comprises convolution layers and parse convolution layers that are arranged alternately; and S2: carrying out contrastive analysis on the segmentation images of the same surface feature in the two remote sensing images so as to obtain a contrastive analysis result. The beneficial effects of the method and the system disclosed by the present invention are that: the technical scheme has good fault tolerance of the interference factors such as the atmosphere, the season, and the like, and also has a relatively high identification rate of dense surface features.

The present invention discloses a photographing method and device and belongs to the electronic information technical field. The invention aims to solve the problems of complicated operation and poor user experience in the display of a photographed object by using multi-dimensional photographing of a user in the prior art. The photographing method includes the following steps that: a photographed picture obtained by a first camera is displayed for the user through a first view-finding frame, and a photographed picture obtained by a second camera is displayed for the user through a second view-finding frame, wherein the photographed picture in the second view-finding frame, is the locally enlarged image of the photographed picture in the first view-finding frame; and the pictures in the first view-finding frame and the second view-finding frame are photographed simultaneously according to a photographing instruction.

The invention relates to a home security system and an implementation method thereof. The home security system comprises a data processing module, an infrared detector, an image acquisition module, an image display module, a voice module, a wireless communication module, a visual intercom module and an entrance guard module, wherein the infrared detector is used for sensing whether a visitor appears; the image acquisition module is used for video monitoring and for transmitting the monitored video to the data processing module; the voice module is used for voice communication; the wireless communication module is used for making a call; the visual intercom module is used for the video call; and the entrance guard module is used for controlling the opening of an entrance guard. By adopting the home security system, when a user is not home when the visitor visits the home of the user, the function of remotely opening the entrance guard for the visitor after the identification of the visitor is confirmed by the user can be realized.

The invention discloses an attention mechanism-based natural language generating method for remote sensing images, and belongs to the technical field of computer vision. A CNN part comprises a commonconvolutional pooling layer, a C-S layer and a full connection layer; an RNN part structurally comprises a multi-layer model unit, a GRU unit and an attention unit; the remote sensing images at different moments in the same area are input into the initialized CNN part to obtain eigenvectors; the eigenvectors are input into the GRU unit and the attention unit, the attention unit also receives hidden layer states from the GRU unit, and the attention unit maps the eigenvectors and the hidden layer states to low dimension for further compression and normalization to obtain eigenvectors subjected to weighted average; and the eigenvectors subjected to the weighted average and the hidden layer states of the GRU unit are integrated through the multi-layer model unit, and then maximum likelihood isperformed through a hidden layer and a normalization layer to obtain an integer sequence. The problem that a result of generating a natural language by the remote sensing images at present is not ideal is solved.

The invention provides a pathological classification method and system based on multi-modal deep learning. The method comprises: extracting pre-selected attributes from the electronic medical recordsto serve as feature representation vectors of structural data, randomly discarding the feature representation vectors according to a preset proportion after being averagely amplified, and replacing the discarded parts with numbers 0 to serve as medical record feature vectors of the structural data in the electronic medical records; obtaining a histopathology image corresponding to the electronic medical record, performing global average pooling on the feature map of each convolutional layer of the convolutional neural network, and splicing the feature maps into a one-dimensional vector to serve as a rich image feature vector of the histopathology image; and splicing the image feature vector and the medical record feature vector together to obtain a multi-mode fusion vector, and inputting the multi-mode fusion vector into a full connection layer to obtain a binarized pathological classification result. The technical problem that the accuracy of pathological benign and malignant classification through single-mode feature representation is not high is solved.

The invention provides blood glucose test method and system. The test method comprises the steps of: S1) acquiring a hyperspectral image data of a to-be-tested position; S2) extracting vein region spectral image data from the hyperspectral image data; S3) on the basis of the vein region spectral image data, calculating blood glucose concentration. Based on the fact that blood glucose is distributed in vein of human body, the blood glucose test method and system are used for quantitatively analyzing blood glucose data by recognizing the human body vein region from the collected hyperspectral image data and extracting characters pixel spectral data from the vein region; compared with a conventional method for analyzing the blood glucose data with whole-zone spectral data, the method can reduce influence on analysis results due to spectral data of human body tissue, such as fat, muscle and the like, which contain no blood glucose, so that spectral analysis is more accurate and stable.

The invention provides a license plate character identification method and apparatus. The method includes steps: obtaining a to-be-identified colorful license plate character image; performing image processing on the obtained colorful license plate character image; identifying characters in the colorful license plate character image after image processing through a pre-trained license plate identification model, and obtaining similarities between the characters in the colorful license plate character image and preset characters in the license plate identification model; and determining the fact that the preset character with the highest similarity and the character in the colorful license plate character image are consistent. According to the license plate character identification method and apparatus, the identification rate of the license plate characters can be increased.

The invention relates to an optical system, an image capturing module and an electronic device. The optical system includes: a first lens element with negative refractive power, wherein an object-side surface of the first lens is concave in a paraxial region thereof; a second lens with refractive power , wherein an object-side surface of the second lens is convex in a paraxial region, and an image-side surface of the second lens is concave in a paraxial region; a third lens with positive refractive power, wherein an object-side surface of the third lens is convex in a paraxial region; a fourth lens with positive refractive power, wherein an object-side surface of the fourth lens is convex in a paraxial region, and an image-side surface of the fourth lens is convex in a paraxial region; a fifth lens with negative refractive power, wherein an object-side surface of the fifth lens is concave in a paraxial region thereof, and an image-side surface of the fifth lens is concave in a paraxial region; a sixth lens with positive refractive power, wherein an image-side surface of the sixth lens is convex in a paraxial region; and a seventh lens with negative refractive power. The optical system satisfies that FOV is greater than or equal to 120 degrees and less than or equal to 130 degrees. The optical system has an ultra-wide angle characteristic, and can acquire more scene contents to enrich imaging information of the optical system, thereby improving user experience.

The invention discloses a sleeper defect detection method based on step-by-step deep learning and a storage medium. The method comprises the steps: obtaining a track image, and carrying out the preprocessing of the track image to obtain a sleeper region; inputting the sleeper area into a trained sleeper damage detection model to obtain a detection result of the track image; the training method ofthe sleeper damage detection model comprises the steps that a sleeper training set is constructed, and the network structure and the network input channel number of a YOLO V3 network are adjusted; training the migrated YOLO V3 network by adopting the sleeper training set to obtain a sleeper positioning model; adopting a sleeper positioning model to intercept sleeper areas of the pictures in the training set a to train the Darknet-53 network; adopting a defect feature extraction model to extract sleeper features of the pictures in the training set a to train an SVM classifier to obtain a classification model; wherein the sleeper positioning model, the defect feature extraction model and the classification model which are connected in sequence are used as a sleeper damage detection model.

A color night vision system based on infrared and low light level image fusion technology is disclosed. The invention relates to a color night vision system and particularly relates to the technology with the fusion of infrared and low light night vision image fusion imaging. At present, main night-vision equipment is divided into an infrared thermal imager and a low light night-vision instrument. However, the infrared thermal imager has the disadvantages of blurred outline of an infrared image, low resolution and poor sense of depth and does not meet a human eye visual characteristic, and the eye observation is not facilitated. The low light night-vision instrument is easily affected by the environment conditions of light, weather and the like, in the bad conditions of rain, rain and smoke, the image detail loss is serious. The invention provides a color night vision system which is not affected by weather, light and smoke and has the advantages of high resolution, good real-time performance, rich details, the accordance with human eye visual characteristic, and easy reading understanding by people. According to the system, an image pyramid method is used, the two paths of night vision images of infrared and low light in the same scene and same visual angle are fused, then a color rapid transfer algorithm is used, and the tone in a day natural scene is transferred to a fused night vision image.

The invention discloses an optical lens, a camera module and electronic equipment. The optical lens comprises the following lens elements which are sequentially arranged from an object side to an image side along an optical axis: a first lens element with positive focal power having a convex object side surface and a concave image side surface; a second lens element with negative focal power having a convex object side surface and a concave image side surface; a third lens element with focal power; a fourth lens element with focal power having a concave object side surface and a convex image side surface; a fifth lens element; a sixth lens element; and a seventh lens element with negative focal power having a concave object side surface and a concave image side surface. The optical lens also satisfies the relational expression: 2.30 < FNO * TTL/ImgH < 2.38. According to the optical lens, the camera module and the electronic equipment provided by the embodiment of the invention, the design requirements of miniaturization, large aperture, large image plane and high imaging quality of the optical lens can be met.

The invention discloses a night target detection method based on millimeter wave radar and visual fusion. The method specifically comprises the steps of preprocessing original data based on millimeterwave radar detection, acquiring an original image corresponding to the original data by using a camera, projecting an effective target point detected by the millimeter wave radar onto the original image to generate a region of interest, conducting image brightening on the image in the region of interest, classifying effective targets in the image based on visual deep learning, and matching the target category with the track of the effective target measured by the millimeter wave radar. The camera collects original image data; meanwhile, the all-weather working characteristics of the millimeter wave radar are combined, image brightening and target detection are carried out in the region of interest generated on the image by utilizing the target reflection points, detection of targets suchas pedestrians, motorcycles and automobiles at night is realized by utilizing the advantages of visual deep learning, and the accuracy of night target detection and tracking is effectively enhanced.

The invention discloses a fixed area night fog all-weather color video imaging system and a building method thereof. The system comprises a visible light video collection system, a visible light image collection module, a background image updating module, a visible light background image bank, an infrared light video collection system, an infrared light image collection module, an image preprocessing module, an infrared light background image bank, an infrared light target detection module, an image registration module, a video fusion module, a video signal emitting module, an RTSP transmission module, and a display module. The system uses an improved algorithm integrating a mixed Gaussian model and a symmetric difference method as well as an imaging concept of fusion of the infrared video and visible light background image. After processing by a data processing center, the system can output a clear color video image at night, in fog weather or other weather with poor visibility, so as to realize the color effect of the infrared light video truly. The system is easy to realize, is low in cost, is easy to install, and realizes long-distance real-time monitoring.

The invention provides a 2D and naked eye 3D dual-screen displaying endoscope system. The system comprises an image acquisition device, an image processing device and an image displaying device, the image acquisition device, the image processing device and the image displaying device are connected in sequence, the image processing device comprises an image acquisition module and a middle viewpointgeneration module, the image displaying device comprises a 2D monitor and a naked eye 3D monitor, the image acquisition module is connected with the 2D monitor, and the middle viewpoint generation module is connected with the 3D monitor. When a doctor conducts an operation, the dual-screen endoscope system can provide 2D pictures and 3D visions for the doctor simultaneously, and the doctor does not need to wear any auxiliary equipment. The invention further provides a displaying method of the 2D and naked eye 3D dual-screen displaying endoscope system.

The invention relates to an image registration, interpolation and denoising-based image enhancement method. The method includes the following steps that: S1, a main image and a plurality of auxiliary images of the same object are obtained, and the resolution of the main image is higher than that of the auxiliary images; S2, with the main image adopted as reference, each auxiliary image is registered with the main image; S3, color matching is performed on the main image and the registered auxiliary images; S4, interpolation operation is performed on the auxiliary images, and the auxiliary images are magnified; and S5, bilateral filter denoising is performed on the main image by using the magnified auxiliary images obtained in the step S4, so that an enhanced main image can be obtained. Compared with the prior art, and according to the image enhancement method of the invention, the image information of the plurality of images is utilized, and the image information is enriched by using image interpolation in an image enhancement process, so that high feasibility can be achieved; the obtained image has high accuracy and high resolution; and the method has a wide application range.

The invention discloses an improved YOLOv3 model side-scan sonar shipwreck target automatic identification method based on transfer learning, and belongs to the technical field of side-scan sonar image target identification and deep learning. The improved YOLOv3 model side-scan sonar image shipwreck target identification method based on transfer learning solves the problems in existing side-scan sonar image manual interpretation and manual feature extraction, and also solves the problems that a Faster R-CNN model is poor in small target recognition effect, high in missing alarm rate and low inrecognition speed. The identification and positioning precision of a shipwreck target is further improved, so that the model achieves a better convergence effect, and finally, the purposes of improving the overall performance of the model and achieving real-time detection are achieved.