116 results about "Image perception" patented technology

A handheld image projecting device that modifies a visible image being projected based upon the position, orientation, and shape of remote surfaces, remote objects like a user's hand making a gesture, and / or images projected by other image projecting devices. The handheld projecting device utilizes at least one illuminated position indicator for 3D depth sensing of remote surfaces and optically indicating the location of its projected visible image. In some embodiments, a handheld projecting device enables a plurality of projected visible images to interact, often combining the visible images, reducing image distortion on multi-planar surfaces, and creating life-like graphic effects for a uniquely interactive, multimedia experience.

The invention relates to a method for evaluating the perception sharpness of a fused image based on human visual characteristics, and belongs to the technical field of image fusion in image processing. A human perception contrast module is constructed on the basis of two main human visual characteristics, namely a contrast sensitivity characteristic and a brightness mask characteristic, a novel image perception contrast algorithm is disclosed through an improved Peli contrast model, and the perception sharpness of the image is evaluated by calculating the human perception contrast of a detailed edge area in the image to obtain an image sharpness objective evaluation model consistent with human subjective evaluation. The method is mainly used for judging whether fused images in different bands can meet specific application requirements or not, namely judging whether the fused images are favorable for an observer to understand image scene contents or not.

The invention provides a method for finding a landing page from a phishing web page. The method comprises the followings: firstly, keywords are extracted from web page text and web graphics, so as to form the lexical signature of the phishing web page, then searching for the lexical signature is performed on a plurality of search engines, the front K most relevant web pages are found out by synthesizing the results of those search engines, the K web pages and the phishing web pages are kept in a picture form, an image perception hash sequence is extracted, finally, the Hamming distances between the K web graphics and the phishing web graphics can be respectively calculated, and one or more lawful web pages simulated by the phishing web page can be selected according to the sizes of the distances.

The invention discloses a CMOS (complementary metal-oxide-semiconductor transistor) imaging measured value obtaining system based on compressed sensing and a method thereof. The system provides an analogue pixel matrix to the system by a CMOS image sensor and is characterized by comprising a linear feedback displacement register, a displacement register, a line selector, a multi-path selector and an analogue / digital converter. Compared with the prior art, the system has the characteristics of generality, encryption, robustness, scalability and the like of the imaging system based on compressed sensing, can shorten the measured value obtaining time and effectively reduce the power consumption of the sensor in comparison with a CMOS imaging system based on random convolution, and has simple structure and is easy to implement in comparison with other imaging systems based on compressed sensing; moreover, the system effectively shortens the measured value obtaining time on the basis of a parallel processing idea, and remarkably reduces the power consumption of the CMOS image sensor.

The invention provides an industrial Internet intrusion detection method based on a flow feature map and perception hash for mainly solving the problems of low detection performance and poor adaptability of the existing industrial Internet intrusion detection method. The industrial Internet intrusion detection method draws lessons from an image processing method and comprises the following steps: firstly obtaining a standard test bed experimental data set, performing feature selection by using an information entropy method to construct a flow feature vector, and performing a normalization operation on a part of attributes; then, converting the flow feature vector into a triangle area mapping matrix by using a multivariate correlation analysis method to construct the flow feature map; and finally, obtaining a hash abstract of the flow feature map by using an image perception hash algorithm based on discrete cosine transform SVD and singular value decomposition SVD, and generating an intrusion detection rule set in the form of a binary character string. Moreover, hash matching is performed by using an accurate matching method based on character strings, a similarity measurement method based on a normalized Hamming distance and a clustering analysis method based on a Euclidean distance so as to detect abnormal flow and malicious intrusion in the industrial Internet.

The invention relates to a typical scene color fusion image quality evaluation method based on a vision task and belongs to the image fusion technology field in image processing. According to the invention, through a subjective evaluation experiment, a regression analysis method is used to establish a fusion image integration quality prediction model based on the vision task. Object background perception contrast and image definition can be used to effectively predict an image perception quality based on object detection. Color compatibility and the definition are used to effectively predict the image perception quality based on scene understanding. Compared to the traditional image quality evaluation method, an image quality evaluation index based on the vision task can carry out comprehensive evaluation of the image quality aiming at an application purpose of the color fusion image. Although it is difficult to perform objective quantification, three basic indexes of the object background contrast, definition and the compatibility comprised in a prediction model are easy to carry out quantification calculating. An effective solution is provided for a problem of the fusion image integration quality objectivity evaluation.

An apparatus for projecting modulated light onto a retina of a viewer, to thereby allow the viewer to perceive a display object via a virtual image is disclosed. The apparatus is configured to include: an emitter emitting light; a modulator modulating the light; and a controller controlling the emitter and the modulator for generating virtual images to be perceived by the viewer in an image display field. The controller is operated, such that a principal virtual image defining the display object is perceived by the viewer in a principal display region of the image display field, and such that an auxiliary virtual image is perceived by the viewer together with the principal virtual image, in an auxiliary display region which is located in the image display field in a predetermined positional relationship with the principal display region, as viewed from the viewer.

The invention belongs to the technical field of electronic equipment, and discloses a full screen display mobile phone and a method for realizing full screen display. The full screen display mobile phone comprises a full screen and an outer shell connected with the full screen, wherein the full screen is a full transparent display screen; a front camera and other light or image perception functionmodules of the mobile phone are arranged under the full transparent display screen; the front camera and the other light or image perception function modules can directly sense light or images through the full screen. According to the full screen display mobile phone, positions do not need to be reserved for the front camera and the other light or image perception function modules, and a frameless technology and the like are combined for use, so that the screen-to-body ratio of the full screen device can be greatly improved and can reach about 100%; the full screen display mobile phone is free from bang, is attractive in appearance, can realize functions such as taking pictures at the front without needing a telescopic support piece, and is convenient to use.

The invention discloses a no-reference image quality evaluation method based on a multi-task convolutional neural network, and belongs to the field of image perception. The method specifically comprises the following steps: step 1, extracting a plurality of image blocks with fixed sizes from each image of a manually labeled image quality data set, and each image block corresponding to two labels,which are respectively a degradation category of the image and a degradation degree of the image, so as to form a training set; step 2, constructing a convolutional neural network model based on dictionary learning; step 3, training the constructed convolutional neural network model by using the training set, and determining parameters of the convolutional neural network model after the training is finished; and step 4, during application, inputting the to-be-scored pollution image into the trained convolutional neural network model to obtain a corresponding image quality score. Compared witha traditional method, the method is higher in consistency with subjective evaluation in the field of non-reference image quality evaluation, and key indexes such as the Spearman rank correlation coefficient and the Pearson linear correlation coefficient are obviously improved.

The invention discloses a ship feature re-identification method, application method and system based on deep learning, and According to the method, a deep feature extraction network with the efficientprocessing capability in the field of image perception is used to extract the deep features with the distinction degree, the deep features with the high-level perception semantics is autonomously extracted, the change of pixel levels is not depended on, and the overall features are considered, so that the problems of overwater supervision, low ship information query efficiency and the like causedby inaccurate matching and high detection error rate are solved, and the target ship discrimination can be effectively carried out. The PCB partitioning and the matrix type operation matching have the good recognition effect for the ships only parts of which occur, the matching speed is high, the problems of ship transformation, hidden identity escape supervision and the like are solved, and theintelligent auxiliary effects on the maritime traffic management, accident investigation, water conservancy attack illegal sand mining, navigation channel ship gate passing charging, customs attack private activities and the like are achieved. Compared with a ship identification method in the prior art, the ship re-identification method has obvious advantages in efficiency, cost and accuracy.

The invention provides a blind image quality evaluation method based on complementarity combination characteristics and multiphase regression. On the aspect of characteristic extraction, image perception related information can be more accurately captured through overall frequency domain image characteristics and local empty domain image characteristics, wherein the overall frequency domain image characteristics and the local empty domain image characteristics have complementarity. On the aspect of prediction model construction, multiple supporting vector regression schemes are introduced, and the independent training sample set of each test image is established by searching for K pairs of neighbors of the test image. Through the segmented regression operation, the prediction accuracy of a perception quality prediction model can be effectively improved. Compared with an existing representative blind image quality evaluation method, the method has higher robustness, and the prediction quality grade more consistent with the grade obtained through manual work can be obtained.

A system for a 3 dimensional (3-D) user interface comprises: one or more 3-D projectors configured to display an image at a first location in a 3-D coordinate system; one or more sensors configured to sense user interaction with the image and to provide user interaction information; and a processor configured (i) to receive the user interaction information from the one or more sensors; (ii) to correlate the user interaction with the image; and (iii) to provide one or more indications responsive to a correlation of the user interaction with the image, wherein the one or more indications comprise displaying the image at a second location in the 3-D coordinate system. A method for providing a 3-D user interface comprises: generating an image at a first location in a 3-D coordinate system; sensing user interaction with the image; correlating the user interaction with the image; and providing one or more indications responsive to a correlation of the user interaction with the image, wherein the one or more indications comprise displaying the image at a second location in the 3-D coordinate system. Computer readable program codes related to the system and the method of the present invention are also described herein.

The embodiment of the invention provides a short-range substitute automatic parking method and device and an automobile. According to the method, the automobile can be automatically parked into a common parking space of an automobile owner. The method comprises the steps of when a short-range substitute parking function is activated, recognizing a target garage where a vehicle is located based onvehicle positioning; judging whether parking information associated with the target garage exists in a parking information base or not; if so, reading all parking information associated with the target garage from the parking information base; pushing all the parking information, and receiving a piece of target parking information fed back by the user after selection; judging whether a parking path exists in the target parking information or not; if so, controlling the vehicle to run according to the parking path in the target parking information; comparing the parking space information acquired by the image sensing module with the parking space number in the target parking information to determine the specific position of the target parking space when the vehicle runs to the vicinity of the end point position of the parking path in the target parking information; and controlling the vehicle to park in the target parking space.

The invention discloses an anti-counterfeit halftone intelligent digital watermark making method of paper medium output. The specific steps are as follows: the host image is binarized, at the same time, the embedded watermark image is binarized, forming a binary digit string, and as the embeddable watermark coding information, a series of two-dimensional random number pairs are generated by usingthe random number seed as the key. The watermark coding information formed in S1 is embedded into the selected embedding position as the embedding position of the embeddable watermark coding information, and the host image embedding the watermark coding information is optimized by using the visual iterative algorithm. The trained neural network model is used to extract the watermark image. The invention can produce printing watermark with anti-counterfeiting function, has wide application prospect in copyright identification, anti-counterfeiting packaging and intelligent packaging, has high technical content, has large printing watermark embedding quantity, does not obviously affect image perception quality, and has hiding property and does not affect consumers to use printed matter.

The super-resolution image reconstruction method comprises the steps of generating images with different perception qualities through different super-resolution image generation methods; Calculating perception quality evaluation scores of the images with different perception qualities through the perception quality evaluation indexes; Training the sorting estimation network according to the calculated perception quality evaluation score; Calculating the sorting content loss of the image generated by the generator of the generative adversarial network according to the trained sorting estimationnetwork, and guiding the generator of the generative adversarial network to train according to the sorting content loss; Carrying out super-resolution image reconstruction on the low-resolution imageaccording to the trained generator. The better image perception quality is obtained by directly optimizing the perception quality evaluation indexes by using the sorting estimation network, expansioncan be flexibly carried out according to requirements, and a generator can be restrained to generate super-resolution reconstructed images with different characteristics.

The invention discloses a compressed noise removing method based on image perception quality. Aiming at the characteristics of diversity, instability and the like of noise caused by video compression,a data set is constructed, and a deep residual denoising model is provided; meanwhile, in order to solve the problem that high-frequency information is lost in the denoising process of an existing algorithm, loss calculation based on visual perception characteristics is provided, a denoising model is assisted in learning noise residual errors, and the deep learning technology is utilized to learnactual compression noise to obtain a denoised image. The beneficial effects of the invention are that the method achieves the learning of the actual compression noise through the deep learning technology, and obtains more accurate residual noise; characterization image perception quality characteristics are extracted by using an image quality evaluation model and are used for loss calculation ofa denoising model, so that the denoised image better conforms to visual perception of people.

The invention relates to the field of image detection in artificial intelligence, and provides an artificial intelligence-based confrontation sample generation method, which comprises the steps of obtaining a target image and a confrontation sample generation model; inputting the target image into an encoder network for encoding to obtain a hidden variable, and inputting the hidden variable into a generator network for image generation to obtain a candidate adversarial sample; determining an image perception similarity between the target image and the candidate adversarial sample; when the image perception similarity is smaller than or equal to a preset threshold value, inputting the candidate adversarial samples into a preset image classification model for classification to obtain an output probability; according to the output probability and a real classification label of the target image, determining whether the candidate adversarial sample meets a preset adversarial sample condition; and if yes, determining the candidate adversarial sample as a target adversarial sample of the target image. The accuracy of the adversarial sample is improved. The invention further relates to the field of block chains and medical treatment. The adversarial sample generation model can be stored in the block chain.

The purpose of the invention is to realize efficient identification of warehouse large cargos by using a video means and complete cargo category analysis and quantity statistics. The invention provides a storage counting method, which comprises the following steps of: 1, adding a warehouse position sensing module in a monitoring system, and finishing data interaction between the sensing module andacquisition equipment; 2, taking the video acquisition device as a controlled device and responding to an instruction of a sensing module in real time to capture images of goods in the shipping spaces, and enabling the sensing module to collect the images of all the shipping spaces; step 3, analyzing, identifying and counting the cargos in the warehouse spaces; 4, preprocessing the images by adopting an image processing and computer vision analysis algorithm, and establishing a warehouse position cargo data set; 5, training the data set to obtain a target recognition model; step 6, convertingthe snapshot shipping space image signal into a digital signal according to pixel distribution, brightness, color and other information; and step 7, performing operation on the signals in the step 6to extract features of a target, and performing contour detection and classification calculation on the large cargos in the shipping spaces.

An eyesight training method considering binocular eyesight development comprises two parts of a binocular object image perception detection module and a eyesight training module, wherein the binocularobject image perception detection module comprises a binocular simultaneous visual brightness-contrast relative threshold detection module, a binocular object image displacement detection module anda binocular object image unequal detection module; and the visual training module generates a visual training sighting mark suitable for training of the patient at the current illness state stage according to the detection result detected by the binocular object image perception detection module. The monocular fine training under the visual field of the two eyes can obviously improve the eyesightand improve the efficiency; obstacles of two-eye competition, object image displacement, object image inequality and the like are cleared for establishing and perfecting sensory fusion, and the efficiency of establishing and perfecting two-eye eyesight can be obviously improved.

A projection screen for forming an image by converting light pixel pulses from a digital projector comprises a three-dimensional sheet matrix made of a transparent composite. Functional inclusions for light-scattering, light-absorbing and luminescence of the light from the projector are distributed through the matrix thickness to thereby enable that the conversion of the light pulses into the image for direct perception by eyesight be performed throughout the volume of the matrix. The matrix thickness between the frontal and rear surfaces of the matrix is selected for digital image sources between an inter-pixel grid width and tenfold diagonal size of a pixel of a digitized image on the screen. The object of the invention is to reproduce identifying features of informational models of real objects in a wide angle of image perception under side lighting.

The invention discloses a multi-sensor fusion vehicle-road collaborative sensing method for automatic driving. The method comprises a data enhancement module, a point cloud sensing module, an image sensing module, a multi-sensor fusion module, a V2X real-time communication module, a selective compensation module and a positioning module based on SLAM and GPS / INS fusion. Firstly, a public data set is processed through a data enhancement module; the three-dimensional information obtained in the point cloud sensing module and the two-dimensional information obtained in the image sensing module are fused through a multi-sensor fusion module; the position information of the vehicle is obtained by means of a positioning module based on SLAM and GPS / INS fusion, and the automatic driving vehicle is helped to make accurate judgment in a complex environment; meanwhile, perception information is shared through the V2X real-time communication module and vehicles or roads in the surrounding environment, shielding missing information is effectively compensated through the selective compensation module, and the real-time communication efficiency is improved; the method is high in accuracy and reliability, and can effectively solve the problems of information loss and shielding under a complex road.

The present invention is directed to an apparatus which can acquire, readout and perceive a scene based on the insertion, or embedding of photosensitive elements into or on a transparent or semi-transparent substrate such as glass or plastic. The substrate itself may act as the optical device which deflects the photons of an incident image into the photosensitive elements. A digital neural memory can be trained to recognize patterns in the incident photons. The photosensitive elements and digital neural memory elements may be arranged with light elements controlled in accordance with the patterns detected. In one application, intelligent lighting units provide light while monitoring surroundings and / or adjusting light according to such surroundings. In another application, intelligent displays display images and / or video while monitoring surroundings and / or adjusting the displayed images and / or video in accordance with such surroundings.

An apparatus which can acquire, readout and perceive a scene based on the insertion, or etching of photosensitive elements into or on a transparent or semi-transparent substrate such as glass. The substrate itself acts as the optical device which deflects the photons incident to the reflected image into the photosensitive elements. Photosensitive elements are interconnected together by a transparent or opaque wiring. A digital neural memory can be trained to recognize specific scenery such as a human face, an incoming object, a surface defect, rain drops on a windshield and more. Other applications include image-perceptive car headlight and flat panel display detecting and identifying the viewer's behavior (gaze tracking, face recognition, facial expression recognition and more). Yet another application includes sliding doors perceiving the direction and speed of an individual coming towards that door. Yet another application includes permanent damage detection (texture change) in dam, bridge or other manmade construction.

The invention provides a perception-oriented image super-resolution reconstruction method and system with a large receptive field, and relates to the field of image processing.The method comprises the steps that firstly, original data in a super-resolution reconstruction data set is preprocessed, and paired LR-HR training data is constructed; secondly, inputting the preprocessed data set into a PSNR-oriented super-resolution reconstruction network with a receptive field, and only adopting L1 as training loss for training; then, the trained PSNR-oriented model is used as initialization of a generator, and the discriminator and the generator are alternately trained to obtain a final super-resolution reconstruction model; and finally, loading the model, and inputting a picture needing super-resolution into the trained super-resolution reconstruction network model to obtain a high-resolution image corresponding to the low-resolution image. According to the method, the multi-scale information of the image is effectively extracted, more high-frequency information, complex texture details and the like can be reconstructed, and the image perception index is remarkably improved.

The invention relates to a performance testing system of a visual perception system and a performance testing method thereof. A performance test system of the present invention includes a storage module for providing a video data source with original annotation information; A video injection unit for injecting the video data source offline into the visual perception system to be tested; And a comparative analysis unit, which is used for comparing and analyzing the perception result of the visual perception system and the original annotation information in the video data source to obtain the performance evaluation result of the visual perception system, wherein the perception result is the perception result output by the visual perception system after image perception processing is performed on the video data source. The invention can realize the offline performance test of the visual perception system, and the consistency of the performance test of different visual perception systems is good, and the quantitative performance evaluation result can be output.

The invention belongs to the field of image processing, and particularly relates to a parking lot parking space identification method based on an image similar degree. The method puts forward a parking space state discrimination index Y, and an image perception Hash technology is organically combined with a structure similarity method to finish parking space state identification. The method specifically comprises the specific processes that: carrying out parking space image collection, image enhancement, image correction and target region division, carrying out Gabor filtering processing on animage, extracting the textural feature map of the image, carrying out Hash processing on the image to obtain a Hamming distance h, carrying out structure similarity operation on the image to obtain an image structure similarity p, calculating the parking space state discrimination index Y, and finally, comparing the discrimination index Y obtained by calculation with a set threshold value w so asto finish identifying the parking space state. By use of the method, an adverse effect of an external condition (illumination changes and snowfall weather) for a parking space state identification result in the prior art is solved, and the method has high objectivity and universality.

The invention relates to the technical field of vehicle monitoring, in particular to a fusion method and device for vehicle visual image information, and the method comprises the steps: determining the current working condition of a vehicle; determining a detection range of the vehicle visual image according to the current working condition; in the detection range, judging whether a target vehicle is detected or not, wherein the target vehicle exists in other vehicles except the vehicle in the detection range; if the target vehicle is not detected, displaying an image of the vehicle; if the target vehicle is detected, receiving visual image information of the target vehicle transmitted by an Internet of Vehicles communication system; splicing images through an image perception fusion model according to the visual image information; storing the spliced image to a cloud system; and displaying according to the spliced image stored in the cloud system. The vehicle visual image information of the vehicle in multiple directions can be obtained through the wireless communication system, multi-vehicle cooperative control is achieved, and the driving safety is improved.

The embodiment of the invention discloses a vehicle error state updating method and device. The method comprises the steps: determining a target map element which has a matching relation with a targetimage element in a sensing image from a preset navigation map when the sensing image is received; based on a first target conversion relationship between a first position of the target map element inthe world coordinate system and a position of the target map element in the camera coordinate system, converting the first position into a second position based on the camera coordinate system at anexposure moment of the feature point of the target image element; projecting the converted target map element corresponding to the second position to a plane where the perception image is located, andupdating an error state matrix of the vehicle by using a re-projection residual error between the projected target map element and the target image element. By the adoption of the technical scheme, when the roller shutter camera is used for carrying out image sensing to position the vehicle pose, the positioning precision of the automatic driving vehicle is improved.

The invention provides a behavior detection method and system, a machine readable medium and equipment. The behavior detection method comprises the following steps: acquiring a to-be-detected continuous frame image containing a target behavior; representing the continuous frame images to be detected by using a neural network, so that each frame image can sense the target behavior characteristic information of the current frame image and the target behavior characteristic information of other frame images; classifying the represented continuous frame images to be detected according to a classifier to obtain at least two classification results; and combining the at least two classification results to obtain a target behavior detection result. According to the behavior detection method, eachframe of image can perceive all information contained in the video, and the scores are generated according to the image, and the scores are sorted, and the generation candidate time domain and regionwith high score values are selected, and missing detection and false detection are not caused. Meanwhile, boundary adjustment is performed on the classification result so that the starting time of thetarget behavior can be accurately positioned, and the behavior detection method is suitable for diversified human behaviors and human behaviors of different time scales.

The embodiment of the invention discloses a model optimization method and device and a computer program product, and the method comprises the steps: obtaining an image sample set, calling a to-be-optimized target perception model to carry out the image perception of each image sample in the image sample set according to a target image perception task, and obtaining a target perception result of each image sample; calling a reference perception model to perform image perception on each image sample according to the target image perception task to obtain a reference perception result of each image sample; performing differential processing on the target sensing result of each image sample and the corresponding reference sensing result to obtain a differential result of each image sample; according to the difference result of each image sample, carrying out difficult sample mining in the image sample set to obtain one or more difficult samples; and updating the target perception model according to optimization parameters of the target perception model determined through the one or more difficult sample. According to the embodiment of the invention, the sensing capability of the target sensing model can be improved.