92results about How to "Enhanced implementation" patented technology

The invention discloses a microphone array voice enhancement method and system. The microphone array voice enhancement method comprise the steps that multi-channel voice signals are collected by usinga microphone array, voice energy obtained by an energy collector is received, sound source azimuth information is acquired, a time delay compensation value of each channel of each microphone array signal is calculated by using geometric relationships of the sound source azimuth information, and beam forming processing is performed according to the time delay compensation value of each channel ofeach microphone array signal to realize the enhancement of voice information. According to the microphone array voice enhancement method and system, the signal reflection focusing characteristic of aconical structure is fully utilized to collect the voice energy in a corresponding angle range, the azimuth information can be obtained in real time, a complex azimuth estimation algorithm is not required, and the system is simple and convenient to implement.

The invention provides a method for enhancing images, which comprises the following steps: processing original images to obtain detail images; calculating local enhancement coefficients of an area which takes each pixel point as a center in the original images; and enhancing each pixel point corresponding to the original images in accordance with the obtained detail images and each calculatedly obtained local enhancement coefficient so as to obtain enhanced detail images. The invention also discloses a device for enhancing the images. The method and device for enhancing the images, which are provided by the invention, are used, thereby being capable of enhancing detail information of the images, such as edges and the like, and improving the accuracy of intelligent analysis and processing.

The invention discloses a medical image holographic AR (augmented reality) display technology system in the medical field. The system comprises an image recognition device, a motion capture device, a man-machine interaction interface and a three-dimensional virtual image superposing system and is characterized in that the image recognition device and the motion capture device acquire a real medical image of a human body, the real medical image acquired after digital conversion is superposed with a three-dimensional virtual human body model and displayed on the man-machine interaction interface, and by means of the human-computer interface which is an interactive advanced human-computer interface of a computer, medical technicians can sense objective existence of organs and tissue morphology in the body of a real patient by combining a virtual reality system with medical images.

The invention relates to a multi-scale multi-phase dispersion strengthening iron-based alloy and a preparing and representation method thereof. The alloy comprises a matrix and a strengthening phase.The strengthening phase comprises at least two kinds of strengthening phase particles different in size. The strengthening phase particles, with the grain size smaller than or equal to 50 nm, in the two kinds of strengthening phase particles different in size account for 85%-95% of the total volume of all the strengthening phase particles. The matrix is a Fe-Cr-W-Ti alloy. The strengthening phasecomprises Y2O3 crystals, a Y-Ti-O phase, a Y-Cr-O phase and a Y-W-O phase. A representation method of the multi-scale multi-phase dispersion strengthening iron-based alloy comprises the steps that thestrengthening phase in the alloy is separated through electrolysis, and then representation is conducted through an electron microscope. The room-temperature tensile strength of the prepared alloy ishigher than 1,600 MPa, the tensile strength of the alloy is larger than 600 MPa when the temperature is 700 DEG C, and the comprehensive mechanical performance is obviously superior to that of same-mark and same-type alloys.

A life prolonging method for quickly repairing microcracks of pump parts and a life prolonging device for quickly repairing the microcracks of the pump parts relate to the technical fields of metal surface protection and laser machining and application. The microcracks of pumps, particularly the microcracks of centrifugal pump shells are strengthened in a cold shock manner by the aid of power effects of laser shock waves; and coatings sprayed by the aid of the power effects of the laser shock waves contain 10-60% iron-titanium compound powder or flake zinc powder. Heating cracks are avoided during repair, the microcracks can be effectively repaired, the mechanical properties of the pump parts can be improved, and the problem of difficulty in practically applying the pump parts in acid and alkali liquid environments is solved. After the method and the device are used for repairing the damaged pump parts, the purposes of life prolonging and improvement of reliability and safety of the pump parts are achieved.

The invention discloses a multi-picture combining and recognizing cartoon education application system for an augmented reality technology, and belongs to the technical field of cartoons. Through eleven steps of beginning or powering on a computer, initializing pictures, selecting types, opening a camcorder lens, putting recognition pictures, confirming the recognition pictures and the like, virtual objects, virtual scenes or system prompt information which are generated by the computer are superimposed into real scenes, so that reality augmentation is realized. Besides, the relation of the system with the real world is not cut off, so that the interactive manner is more natural. When any one recognition picture is taken out alone, the content of one card can be shown on a flat panel display alone. The system disclosed by the invention can be applied to scene simulation of studying, exhibition and presentation, advertisements, games, military trainings and the like.

The invention discloses a retina image enhancement algorithm based on multiple scales and multiple directions. The method comprises the steps that input original RGB retina images are preprocessed, the original RGB retina images are broken up into images of red, green and blue three-channel images, and green channel images of the retina images are extracted; the green channel images of the retina images are subjected to multi-scale Hessian matrix filtering, and multi-scale enhancement images are obtained; the green channel images of the retina images are subjected to multi-direction two-dimension matching filtering, and multi-direction enhancement images are obtained; and the multi-scale enhancement images and the multi-direction enhancement images are subjected to fusion, and the final retina enhancement image results are obtained. Blood vessels in an appointed width range in the retina images are enhanced through a multi-scale Hessian matrix, meanwhile, multi-direction two-dimension matching filtering is used for carrying out enhancement on blood vessels in 12 different directions in the retina images, then the two enhancement results are fused, and accordingly final image enhancement is achieved.

The invention discloses a color eye ground image enhancement method based on an imaging model. The method can improve brightness and low contrast at the same time, and realize image data fidelity, and can effectively enhance the region of interest of clinical diagnosis; and comprises the steps of preprocessing the brightness; extracting foreground pixels and background pixels in a color eye ground image: estimating imaging model parameter-background brightness and an imaging model parameter-transmission coefficient; and finally performing image enhancement according to the imaging model: substituting the background brightness and the transmission coefficient into the imaging model for calculation to obtain an enhanced image.

The invention provides a method and a device for enhancing a CT image sequence. The method includes the following steps: converting each CT image in a to-be-enhanced CT image sequence into a gray image, scaling the gray images into images with a preset size, horizontally stacking and combining the images to get a 3D stereoscopic image, and vertically segmenting the 3D stereoscopic image to get a first preset number of to-be-enhanced 2D images with a first size; inputting the first preset number of to-be-enhanced 2D images with the first size to an image enhancement model in turn, and outputting a first preset number of enhanced 2D images with a second size; and vertically arranging and combining the first preset number of enhanced 2D images with the second size into an enhanced 3D stereoscopic image, horizontally segmenting the enhanced 3D stereoscopic image into a second preset number of target 2D images with the preset size, and taking the second preset number of target 2D images with the preset size as an enhanced CT image sequence corresponding to the to-be-enhanced CT image sequence. Enhancement of the to-be-enhanced CT image sequence is realized.

The invention discloses an image sharpening method based on an SSE2 (Streaming SIMD extensions 2) instruction set. The method comprises the following steps of separating a component A, a component G, a component R and a component B from loaded image data, and obtaining a channel differential component with the component G as a reference component; carrying out high-pass filtering and margin gain amplification processing, and adopting a threshold shrinkage way and a value adjustment way to control a flat region of an image and a high-frequency component in an enhanced component G; adding the high-frequency component in the enhanced component G to an original component G to obtain a sharpened component G; summing the sharpened component G and the channel differential component to obtain a sharpened component R and a sharpened component B; and carrying out staggered arrangement on the component A, the sharpened component R, the sharpened component G and the sharpened component B through register shifting, writing into a buffer zone and then outputting. According to the method, when the image is sharpened, color space conversion is not needed, so that the high-efficient color image sharpening processing is achieved.

The invention relates to a grayscale image enhancement method based on a stochastic resonance mechanism of a delayed self-feedback FHN (fitzhugh-nagumo) model. A traditional method has no good effect in absence of noise model prior knowledge. The grayscale image enhancement method includes steps of performing row scan for the noise-contained grayscale images at first, acquiring one dimensional sequence formed by various pixel gray values, and normalizing various factors of the one dimensional sequence; inputting the normalized one-dimensional sequence into the delayed self-feedback FHN neuron model so as to enable system output to achieve the best state of the stochastic resonance; restoring the outputted sequence as an image pixel value range in an inverted normalization manner, and restoring the image pixel value range as grayscale images; and then performing column scan for the restored grayscale images by repeating the process, and finally acquiring the reinforced grayscale images after row scan and column scan. Since dimension reduction of the images is realized in a row direction and a column direction, peer to peer relationship of two-dimensional images in row and column directions is effectively improved, and retention of space structure characteristics of the grayscale images which are processed by the stochastic resonance is facilitated.

The invention relates to an eye fundus image enhancement method and system. The method comprises the steps: constructing a blood vessel dictionary through an eye fundus learning image; carrying out the Frangi filtering of a to-be-enhanced eye fundus image, dividing blood vessels in a second subimage block into thick blood vessels and fine and fragile blood vessels through direction filtering, and setting the residual error weight and residual error threshold value of a blood vessel region according to the above; solving the inner products of the second subimage block with all first subimage blocks in the dictionary, selecting the first subimage block with the biggest inner product, and calculating the corresponding sparse coefficient of the first subimage block; calculating a residual error image through the selected first subimage block and the second subimage block, calculating the residual error of the blood vessel region in the second subimage block through the residual error weight of the blood vessel region, setting the residual error image as the second subimage block if the residual error is greater than the residual error threshold value, and repeatedly carrying out the process of selecting the first subimage block and calculating the residual error; reconstructing the second subimage blocks through the sparse coefficient, recombining the reconstructed second subimage blocks, and obtaining an enhanced eye fundus image. The method effectively inhibit the background noise, and keeps the fine and fragile blood vessels.

The invention provides a Raw-to-Raw dark light image enhancement method, which comprises the following steps of: acquiring Raw data as a constructed training data set, and combining and expanding thetraining data set; building an image enhancement CNN network model according to the training data set, wherein the CNN network model comprises a multi-task training layer, an input layer, an encodingprocess, a decoding process, a jump connection layer and an output layer; performing iterative training on the CNN network model, updating each learnable parameter of the CNN network model to enable the learnable parameters to reach a certain convergence condition, and completing training; and performing model inference according to the trained CNN network model to finally obtain a dark light image enhancement network model, and performing image enhancement on the Raw data through the model. A dark light image enhancement model capable of intelligently processing a Raw format image is obtainedby constructing a data set and a model and training the data set and the model, so that the dark light image enhancement from Raw to Raw is realized.

Disclosed is a preparation method of liners for rolling mills. According to the preparation method, after the process of processing 45# steel plates according to drawings, size of working surfaces ofcontact moving parts of the liners is reduced by 1mm and then further processing is carried out; alloy powder is prepared from, by mass percent, 0.1-0.13% of C, 1.0-1.5% of Si, 14-16% of Cr, 1.6-1.7%of Co, 1.2-1.6% of Mn and the balance being Fe; every two processed liners of the same dimension are combined by the fixing way of keeping non-working surfaces of every two liners contacting; the alloy metal powder is spread in advance, and is subjected to metallurgical reaction with the basic metal of the liners through high-powder semi-conductor laser high-energy optical bean scanning, and thena reinforced alloy layer is formed on the working surface of each liner; finally, the reinforced alloy layer is subjected to finish machining through a surface grinder. The preparation method needs simple process and causes no pollution; the reinforced liners prepared by the method have the advantages of long service life and low cost.

The invention relates to an ADS-B monitoring system and method based on a Beidou foundation enhancement technology. The system comprises an enhanced CORS station, an airborne station and an ADS-B monitoring terminal. The enhanced CORS station is used to obtain the differential enhancement information of a Beidou satellite, and broadcast and send the differential enhancement information to the outside world. The airborne station is used to receive the positioning signal and the differential enhancement information of the Beidou satellite, and fuse the positioning signal and the differential enhancement information to obtain a high-precision positioning data source. The ADS-B monitoring terminal is used to acquire the positioning data source, display the monitoring information of the airborne station according to the positioning data source, and carry out real-time safety alarm on the airborne station according to the positioning data source and a preset autonomous interval maintaining standard. By using the system, the precision of the data source of the ADS-B monitoring system can be improved, and flight safety can be ensured through taking the autonomous interval maintaining standard as a ground ADS-B monitoring terminal alarm evaluation theoretical basis.

The invention discloses a light carbon nanotube reinforced aluminosilicate polymer foam material and a preparation method thereof, and belongs to the field of preparation of nano-phase reinforced composite foam materials. According to the invention, a carbon nanotube/aluminosilicate polymer material is directly foamed to obtain the light carbon nanotube reinforced aluminosilicate polymer foam material, so that the preparation problem of the aluminosilicate polymer composite foam material with high porosity and high strength is solved. The preparation method comprises the following steps: preparing an alkali-activated solution; preparing a carbon nanotube/alkali excited solution; foaming the mixed slurry; and carrying out curing molding. According to the invention, carbon nanotubes are introduced into the aluminosilicate polymer foam based on the characteristics of a high-strength and high-toughness nano structure of the carbon nanotubes, so that the pore structure and mechanical properties of the original aluminosilicate polymer foam material are improved, the strength of the aluminosilicate polymer foam material is improved, and the application field potential of the aluminosilicate polymer material is expanded. The method is convenient to operate, low in cost, environmentally friendly and suitable for large-scale production.

The invention discloses a mechanical pump small sample fault diagnosis method based on WGAN-GP-C and metric learning, which is based on one-dimensional convolution and time series data design and comprises four parts of data preprocessing, data enhancement, metric learning and model optimization. The data preprocessing realizes data self-adaptive denoising and standardization; in the data enhancement part, network and structure modification is carried out based on WGAN-GP to obtain WGAN-GP-C, so that data expansion according to categories is realized, and edge boundary information of the data is enhanced; a metric network combines a residual idea and a spatial adaptive structure to realize feature mapping, and then a KNN algorithm is utilized to realize state classification; in the model optimization part, the network is optimized by using a weight quantization idea, the KNN is realized by using a Ball-tree, and training data are deleted according to important factors, so that the algorithm performance is improved on the whole. The method has good performance under the condition of rare data, is high in practical value, and can provide thoughts for related workers maintaining mechanical pumps.