36results about How to "Reduce blurriness" patented technology

The invention discloses an image semantic segmentation method of guiding feature fusion based on attention mechanism. The image semantic segmentation method comprises the following steps: (10) constructing an encoder basic network: using an improved ResNet-101 to generate a series of features changing from high resolution low semantics to low resolution high semantics; (20) constructing a decoderfeature fusion module: extracting high-level semantics with strong consistency constraint by adopting a pyramid structure module based on three-layer convolution operation, and performing weighted fusion on low-level stage features layer by layer to obtain a preliminary segmentation heat map; and (30) constructing an auxiliary loss function: adding auxiliary supervision to each fusion output in the decoding stage, superposing the auxiliary supervision with main supervision loss after the up-sampling of the heat map, and strengthening the hierarchical training of the model to obtain a semanticsegmentation map. The image semantic segmentation method of guiding feature fusion based on attention mechanism is high in accuracy and clear in boundary contour.

An image-taking apparatus includes a LED emission section that irradiates a subject with light by causing a LED to emit the light and a xenon-tube emission section that irradiates a subject with light by causing a xenon tube to emit the light. The light from the xenon-tube emission section has higher intensity than the light from the LED emission section. The image-taking apparatus also includes a shake detection section that predicts a shake that will occur at shooting by detecting a shake before shooting. The image-taking apparatus also includes an emission control section that causes the xenon-tube emission section to irradiate a subject with light when the detected shake is above a predetermined level, and causes the LED emission section to irradiate a subject with light when the detected shake is equal to or below the predetermined level.

The invention provides an anisotropic optical flow field and deskew field-based brain MR (magnetic resonance) image registration method, and discloses an optical flow field model-based image registration and biased field recovery coupling model. Aiming at the defect that a biased field can easily cause low registration accuracy of the traditional registration model, the anisotropic optical flow field and the deskew field are combined, and are brought into a unified variational framework, thus the two can supplement each other, and in view of the fact that global regular terms of a Horn model are incapable of guiding the optical flow precisely due to the lack of image information and the movement is insufficient, the image structure information is introduced to regulate the optical flow field, so as to get smooth and accurate optical flow information. According to the anisotropic optical flow field and deskew field-based brain MR image registration method, the image gradation non-uniform field can be restored while registration is performed, and the influence of the biased field is reduced; the image structure information is introduced to reduce the registration results fuzzy degree and retain the image structure information, the integrity of boundary structure information is guaranteed, and further recovery of real images is facilitated. The registration precision is greatly improved, and the robustness is good.

The invention provides a knowledge graph representation learning method fusing entity description and types. The method comprises the following steps of S1, obtaining embedding of a triple entity by utilizing a translation model, and taking a relationship in a triple as a translation operation between a head entity and a tail entity to obtain numerical vector representation of each triple entity and the relationship; s2, adopting a Doc2Vec model to embed the text information described by the entity; s3, combining entity embedding obtained through a Transform model with the entity hierarchicaltype mapping matrix to obtain embedding of a triple entity type; and S4, connecting all the representation vectors to obtain a final triad entity vector, optimizing the training model by adopting a random gradient descent method, and carrying out effect evaluation. According to the method provided by the invention, through entity description and entity type embedding, semantic information expressed by knowledge graph triad entities is improved.

An embodiment of the invention discloses an image acquisition control method and an image acquisition control device. The image acquisition control method includes acquiring movement information of to-be-photographed objects; determining relative movement distances of images on images sensors within first exposure durations according to the movement information; determining imaging regions in the image sensors; adjusting image acquisition control parameters of the imaging regions to shortened exposure durations of the imaging regions; acquiring the images of the to-be-photographed objects by the aid of the adjusted image sensors. The images correspond to the to-be-photographed objects. The imaging regions correspond to the relative movement distances. The image acquisition control method and the image acquisition control device in the embodiment of the invention have the advantages that degrees of image blurring of the imaging regions due to movement of the to-be-photographed objects can be reduced to the greatest extend on the basis that the quality of images acquired in other regions of the image sensors is unaffected, and accordingly the integral quality of the images acquired by the image sensors can be improved.

A camera includes: a photographic lens; an image sensor that captures a subject image; a motion detection unit that detects motion of the camera and outputs a detection signal corresponding to an extent of the motion; a position measurement unit that measures a position assumed by the camera by receiving a radio wave originating from an external source; a blur reducing unit that reduces an extent of blurring in a photographic image attributable to motion of the camera in correspondence to the output value indicated in the detection signal output from the motion detection unit; and a position measurement engagement control unit that disallows engagement of the position measurement unit in operation based upon the output value indicated in the detection signal output from the motion detection unit.

The invention provides a focusing curve establishing method and device, and relates to the field of image processing. The method comprises the following steps: establishing a coordinate system by taking a zoom parameter as an abscissa and a focusing parameter as an ordinate, wherein an object distance in the process of capturing a to-be-processed image by a lens is a fixed value; adopting a monotonous hill climbing algorithm to determine preset zoom parameter key points in the coordinate system and focusing parameter points meeting preset clear conditions corresponding to the preset zoom parameter key points; and generating a focusing curve according to the focusing parameter points meeting the preset clear conditions and corresponding to the preset zooming parameter key points. Compared with the prior art, the efficiency of generating the focusing curve can be improved, the focusing curve capable of being followed actually is provided for zoom following of the simple electric lens, the focusing time of the simple electric lens is shortened and focusing jitter floating of the simple electric lens is weakened.

The invention discloses a high-quality face generation method based on a multi-scale residual network, and the method comprises the following steps: a data set design step: integrating a current mainstream face data set, and generating a data pair of a low-quality face and a high-quality face through software; A model design and training step: designing a multi-scale residual network, and performing model training by using data pairs to obtain a network model weight; And a model prediction step: carrying out model processing on the face image obtained in reality to obtain a prediction result.According to the invention, a deep learning network technology is applied to a generation task of a high-quality face to generate a color face image with high resolution, low fuzzy degree and low noise; And by using a deep learning network method, the time and cost of manual drawing can be reduced, and a solid foundation is laid for subsequent face feature extraction and recognition.

The invention relates to a soil moisture fusion method based on Bayesian maximum entropy, which makes full use of soil moisture obtained by artificial soil sampling, real-time monitoring soil moistureby a sensor (EC-5) and the information of soil moisture retrieved from hyperspectral data, Applied Data Fusion Technology, the data of soil moisture from three different sources were fused, so as toobtain the distribution of soil moisture at the regional scale, Eliminate redundancy and contradiction between information, reduce uncertainty of irrigation decision-making information and fuzzy degree of decision-making reasoning, improve the accuracy of irrigation decision-making, provide irrigation decision-making opinions and suggestions for government drought relief work, guide irrigation district water management departments to carry out scientific allocation of water resources.

The invention relates to the technical field of projection display, and provides a focal length adjusting method and a focal length adjusting device. The method includes that a projection device projects a to-be-projected image according to each focal length in a preset group of focal lengths to acquire a projected image corresponding to each focal length, the area of a region of a third color in the projected image corresponding to each focal length is determined, the third color refers to any color different from three primary colors, the to-be-projected image at least comprises at least two regions, the image color of each region refers to any of the three primary colors, and the image colors of at least two regions are different; the projection device determines a focal length as a projection focal length from the preset group of focal lengths according to the area of the region of the third color in the projected image corresponding to each focal length and projects the image according to the projection focal length.

The invention discloses a microstructure substrate and a manufacturing method thereof and a display device. The method comprises the steps that a fingerprint-resistant film layer is coated on an underlying substrate; a silicon dioxide layer is sputtered on the fingerprint-resistant film layer; and the silicon dioxide layer the etched so as to prepare the silicon dioxide layer having multiple arc concave microstructures. With application of the mode, the influence of the microstructures on the high-resolution panel picture quality can be improved, the blur degree of the picture quality can be reduced and the substrate is enabled to be applied to the high-resolution panel.

An interactive projector includes a projection section adapted to project a projected screen on a screen surface, an imaging section adapted to take an image of an area of the projected screen, and a pointing element detection section adapted to perform a detection process for detecting the pointing element based on the taken image, which is taken by the imaging section, and includes the pointing element, and the pointing element detection section sections the taken image into a plurality of small areas, and performs the detection process in at least one of the small areas using a parameter different from parameters in the rest of the small areas.

The invention discloses blood pressure lowering health care green tea and its preparation method. The method comprises: mixing 15-30 parts by weight of white tea, 20-25 parts by weight of zinc-selenium tea, 20-25 parts by weight of green tea, 20-30 parts by weight of baked green tea or Maofeng tea, 5 parts by weight of folium llicis latifoliae, and 5 parts by weight of chrysanthemums, wherein, the free amino acid content of the white tea is over 7%, the zinc-selenium tea has Zn content of 80-120mg / kg and Se content of 0.25-0.35mg / kg, the green tea has free amino acid content of more than 5%, and the baked green tea or Maofeng tea has tea polyphenol content of over 30%; then mixing the well mixed tea with crushed kenaf accounting for 20-30 wt% of the mixed tea uniformly, and storing the mixture for 30-60d, then conducting screening to filter out kenaf, and carrying out fragrance increasing at a temperature of 150DEG C for 10-12min, and performing depilation at a temperature of 110DEG Cfor 12-15min, thus obtaining the product. With high content of water extract, tea polyphenol and amino acid as well as low caffeine, the health care green tea of the invention plays an active role inprotecting a normal ultrastructure of the cardiac muscle.

The invention discloses an image noise reduction method and a storage medium, and the method comprises the following steps: S1, judging whether an input image contains noise pixel points or not, and further judging the type of each noise pixel point if the input image contains the noise pixel points; s2, sequentially processing the noise pixel points by adopting a corresponding noise reduction method according to the types of the noise pixel points; and S3, outputting the image after noise reduction processing. Since different types of noise pixel points in the input image are respectively processed by adopting corresponding noise reduction methods, detail information of the image can be reserved as much as possible, the blurring degree of the processed image is greatly reduced, the quality of the image after noise reduction is improved, and the method is relatively low in algorithm complexity and relatively high in applicability.

The invention discloses a bridge navigation aid indicator lamp. The bridge navigation aid indicator lamp structurally comprises a lamp holder, a switch, a lamplight plate, an impact ring and a base, wherein the top end of the lamp holder is provided with the switch and is connected with the switch, the lamplight plate is located on the side edge of the lamp holder and adjacent to the impact ring,and the lamplight plate and the impact ring are connected to the exterior of the lamp holder in an annular mode, and the base is arranged below the lamp holder. The bridge navigation aid indicator lamp has the beneficial effects that the bridge navigation aid indicator lamp is arranged at the position of a bearing platform on the water surface by using the impact ring, so that the bridge navigation aid indicator lamp is half-exposed out of the water surface, and meanwhile vibration waves generated when the structure makes contact with water flow serve as an energy supplementing mode generatedby the combined action of the lamplight plate and the water flow, and the water flow and water waves impact a beating bag structure so that the fuzzy degree of the water waves covering the lamplight plate can be effectively reduced, and the marked identification degree of the bearing platform on the water surface is enhanced.

A display panel comprises a plurality of first pixel electrodes and a plurality of second pixel electrodes, the first pixel electrodes and the second pixel electrodes are distributed alternately, each first pixel electrode and each second pixel electrode comprise a main part and a plurality of branch parts connected with the main part, and the main parts divide the first pixel electrodes or the second pixel electrodes into at least four domains; the extension directions of the branch parts in adjacent domains are different, and the extension directions of the plurality of branch parts in the same domain are the same; in one first pixel electrode and one second pixel electrode which are adjacent to each other, the extension direction of the branch part of the first domain of the first pixel electrode is the same as the extension direction of the branch part of the second domain of the second pixel electrode, and the first domain of the first pixel electrode is adjacent to the second domain of the second pixel electrode; and according to the display panel provided by the embodiment of the invention, the period of a liquid crystal grating can be prolonged, the diffraction phenomenon is improved, and the blurring degree of images shot by a camera through a screen is reduced.

The invention relates to a video processing method and a video processing device, and the method comprises the steps of obtaining an input feature of a current moment for an image frame of each moment in a video based on the image frame of the current moment, hidden layer features of m moments before the current moment, and super-division features of n moments before the current moment; inputting the input feature of the current moment into a video super-resolution model to obtain a super-resolution feature of the current moment output by an output layer of the video super-resolution model and a hidden layer feature of the current moment output by a hidden layer of the video super-resolution model, wherein the hidden layer feature of the current moment is used for obtaining input features of m moments after the current moment; and obtaining a super-resolution image at the current moment based on the super-resolution feature at the current moment and the image frame at the current moment, wherein m is an integer greater than or equal to 1, n is an integer greater than or equal to 1, and m and n are not 1 at the same time.

The invention provides an automatic focusing method and system, a storage medium and equipment, and the method comprises the steps: in response to the starting of automatic focusing of a focusing lens, controlling a focusing motor to start rotation, and sequentially collecting a focusing evaluation value and a corresponding focusing position value of each frame of image; judging whether n-1 is an integral multiple of a preset frame interval M based on the nth frame of the current image; if yes, searching the maximum focusing evaluation value for in the focusing evaluation values of the current image and the previous image, using the focusing position value corresponding to the searched maximum focusing evaluation value as the maximum focusing position value, and judging whether the focusing evaluation value of the current image is larger than the focusing evaluation value of the (n-M) th frame image; if not, judging whether the difference value exceeds a preset threshold value; if yes, judging whether a continuous rising mark exists; if yes, controlling the focusing motor to stop rotating and rotate reversely, and making the focusing lens to move to the focusing position corresponding to the maximum focusing position value. According to the method, the reliability and efficiency of automatic focusing are improved.

The invention relates to the technical field of deep residual networks, and discloses a regional water resource classification evaluation method based on an improved deep residual network, which comprises four steps of image enhancement, image fusion, image marking and image segmentation. According to the regional water resource classification evaluation method based on the improved deep residual network, image enhancement is used for improving image information and quality, so that the features of the image are more obvious under human eye observation, the regional water resource identification effect is enhanced, visual analysis ignores the digital interference degree, and the method plays an important role in various aspects of remote sensing images; according to image fusion, a multi-spectral image with low spatial resolution and a single-band image with high spatial resolution are replayed and sampled to generate a new multi-spectral image with high resolution, so that the generated image has high spatial resolution and multi-spectral characteristics, a composite mode is mainly adopted, information provided by remote sensing image data sources of different sensors is integrated, and multi-spectral image fusion is realized. And high-quality image information is obtained.

The invention discloses a visual recognition device based on three-dimensional simulation, and relates to the technical field of recognition devices. The device comprises an inspection vehicle, the upper surface of the inspection vehicle is fixedly provided with a fixed seat, the upper surface of the fixed seat is rotatably connected with a digital holder, two sides of the digital holder are provided with recognition devices, and the upper ends of the recognition devices are provided with baffle plates. When the three-dimensional simulation visual recognition device is used and polled in rainy days, a threaded rod is rotated, so that a gear on the outer surface of the threaded rod drives a toothed plate to move up and down, the inclination of a baffle plate is adjusted, the periphery of identification equipment is shielded, and the inclined baffle plate prevents rainwater from being splashed on a lens of the identification equipment, therefore, the lens is prevented from being dropped by raindrops, images captured by the monitoring end are prevented from being blurred, and normal and safe inspection of the inspection vehicle is not affected.

A liquid crystal display device includes: a light source section; a liquid crystal display panel including pixels and performing video display; and a drive section. The drive section performs the driving on the light source section for turning ON and OFF in synchronization with the driving of the pixels for the light-sequential writing such that, the liquid crystal display panel is selectively illuminated with the light from a lower emission region in the light source section when the pixels in an upper display region are under the driving for line-sequential writing, and such that, the liquid crystal display panel is selectively illuminated with the light from an upper emission region when the pixels in a lower display region are under the driving for line-sequential writing. A phase difference between emission periods in the upper and lower emission regions falls within a range from 90 DEG to 150 DEG both inclusive.

The present invention proposes a knowledge map representation learning method that combines entity description and type, including the following steps: Step S1: use the translation model to obtain the embedding of triplet entities, and use the relationship in the triplet as the head entity and tail entity Inter-translation operation to obtain the numerical vector representation of each triplet entity and relationship; Step S2: Use the Doc2Vec model to embed the text information described by the entity; Step S3: The entity embedding obtained through the Trans model is mapped to the entity hierarchy type Combine the matrix to get the embedding of the triplet entity type; step S4: connect all the representation vectors to get the final triplet entity vector, optimize the training model by stochastic gradient descent, and evaluate the effect. The method proposed by the present invention improves the semantic information represented by the triplet entity in the knowledge map through the entity description and the embedding of the entity type.

A method, system and device for detecting damage to a windshield of a railway motor car, which belong to the technical field of fault detection for windshield damage of a railway motor car. The invention solves the problem that the accuracy of fault detection is low due to the fuzzy image obtained by preprocessing the image using the traditional median filtering method. The present invention adopts a balanced median filtering algorithm to retain the detail information of the image while removing most of the noise, reduce the fuzzy degree of the image, and further improve the accuracy of fault detection. In order to reduce the influence of the threshold selection in the Canny algorithm on the edge feature extraction effect, the present invention proposes a GA-Canny algorithm, which uses a genetic algorithm to find the optimal Canny threshold and improves the detection accuracy of the subsequent detection network. The invention can be applied to the fault detection of the windshield damage of the railway train.

The invention relates to a soil moisture fusion method based on Bayesian maximum entropy, which makes full use of the soil moisture obtained by manual soil extraction, the soil moisture (EC-5) monitored in real time by sensors and the soil moisture information retrieved from hyperspectrum, and the application data Fusion technology, which integrates soil moisture data from three different sources to obtain the distribution of soil moisture on a regional scale, eliminates possible redundancy and contradictions between information, and reduces the uncertainty of irrigation decision-making information and decision-making reasoning. The degree of fuzziness improves the accuracy of irrigation decision-making, provides irrigation decision-making opinions and suggestions for the government's drought relief work, and guides water management departments in irrigation areas to make scientific allocation of water resources.

The invention relates to the technical field of projection display, and provides a focal length adjusting method and a focal length adjusting device. The method includes that a projection device projects a to-be-projected image according to each focal length in a preset group of focal lengths to acquire a projected image corresponding to each focal length, the area of a region of a third color in the projected image corresponding to each focal length is determined, the third color refers to any color different from three primary colors, the to-be-projected image at least comprises at least two regions, the image color of each region refers to any of the three primary colors, and the image colors of at least two regions are different; the projection device determines a focal length as a projection focal length from the preset group of focal lengths according to the area of the region of the third color in the projected image corresponding to each focal length and projects the image according to the projection focal length.

The invention provides an anisotropic optical flow field and deskew field-based brain MR (magnetic resonance) image registration method, and discloses an optical flow field model-based image registration and biased field recovery coupling model. Aiming at the defect that a biased field can easily cause low registration accuracy of the traditional registration model, the anisotropic optical flow field and the deskew field are combined, and are brought into a unified variational framework, thus the two can supplement each other, and in view of the fact that global regular terms of a Horn model are incapable of guiding the optical flow precisely due to the lack of image information and the movement is insufficient, the image structure information is introduced to regulate the optical flow field, so as to get smooth and accurate optical flow information. According to the anisotropic optical flow field and deskew field-based brain MR image registration method, the image gradation non-uniform field can be restored while registration is performed, and the influence of the biased field is reduced; the image structure information is introduced to reduce the registration results fuzzy degree and retain the image structure information, the integrity of boundary structure information is guaranteed, and further recovery of real images is facilitated. The registration precision is greatly improved, and the robustness is good.

The present invention provides an automatic focusing method, system, storage medium and equipment. The method includes: in response to starting the automatic focusing of the focusing lens, controlling the focus motor to start and rotate, and sequentially collecting the focus evaluation value and the corresponding focus position of each frame of image value; based on the current image being the nth frame, judge whether n-1 is an integer multiple of the preset frame interval M; if so, find the maximum focus evaluation value in the focus evaluation values ​​of the current image and its previous images, and find the maximum focus value The focus position value corresponding to the evaluation value is used as the maximum focus position value, and it is judged whether the focus evaluation value of the current image is greater than the focus evaluation value of the n-M frame image; if not, it is judged whether the difference between the two exceeds the preset threshold; if , to determine whether there is a continuous rising sign; if there is, control the focus motor to stop rotating and reverse rotation, and move the focus lens to the focus position corresponding to the maximum focus position value. The invention improves the reliability and efficiency of automatic focusing.

The invention provides a method and device for establishing a focus curve, and relates to the field of image processing. The method includes: establishing a coordinate system with the zoom parameter as the abscissa and the focus parameter as the ordinate, wherein the object distance in the process of capturing the image to be processed by the lens is a fixed value; using a monotone hill climbing algorithm to determine the preset zoom parameters in the coordinate system The key points, and the focus parameter points corresponding to the preset zoom parameter key points that meet the preset sharpness conditions; according to the focus parameter points that meet the preset sharpness conditions corresponding to the preset zoom parameter key points, a focus curve is generated . Compared with the prior art, it can improve the efficiency of generating the focus curve, provide a focus curve that can actually be followed for the simple motorized lens zoom follow, reduce the focusing time of the simple motorized lens, and weaken the focus jitter and fluctuation of the simple motorized lens.

A denoising method for restraining a spectrum characteristic of a remote sensing image of a ground building group comprises: acquiring a reference image of a target of a ground building group from a remote sensing image library of the ground building group, and performing a Fourier transform on the reference image so as to obtain an amplitude spectrum; performing threshold segmentation, erosion and dilation operations on the result of the amplitude spectrum so as to obtain a binary template of the spectrum of the target of the ground building group; acquiring a real-time image of the target of the ground building group by means of the flight of a high-speed air vehicle, performing a Fourier transform on the real-time image so as to obtain the spectrum; performing frequency-domain filtering processing on the spectrum of the real-time image by using the obtained binary template of the spectrum of the target of the ground building group; and performing an inverse Fourier transform to generate the real-time image of the target of the ground building group after filtering.