10results about How to "Remove distortion effects" patented technology

The present invention provides a calibration device and method for a 3D camera, and relates to the technical field of camera calibration. The calibration device for a 3D camera includes a calibration plate, a translation stage, and a linear structured light laser mounted on the translation stage, wherein the translation stage is used to carry a 3D camera, and the translation stage can be moved closer to or farther from the calibration plate, and a detachable groove structure is provided on the surface of the calibration plate facing the translation stage, and a groove is provided on the groove structure. The cross-sectional contour of the groove on the calibration plate is measured by triangulation using a laser line, the coordinates of the points on the cross-sectional contour are extracted, and a set of equations is established with the coordinates of the known edge points of the groove, so that the coordinate transformation matrix coefficients can be solved. According to the coordinate transformation matrix coefficients, the corresponding object coordinates can be found according to the image coordinates, which not only has high solution efficiency, but also reduces the influence of camera lens distortion and has high calibration accuracy.

The invention provides a noise robustness streaming speech recognition method and model fused with a speech enhancement front end. The training method of the model comprises the following steps: acquiring a speech recognition training set; processing the speech training sample by using a causal speech enhancement network to obtain target filter bank features containing global information; processing the voice training sample by using a feature extraction method based on a filter bank to obtain original filter bank features; performing dynamic weighted fusion processing on the target filter bank features and the original filter bank features by using an attention mechanism network to obtain target fusion features containing supplementary information in the voice training sample; processing the target fusion feature by using an initial speech recognition network to obtain training text recognition information; generating a target loss result according to the target filter bank features, the training text identification information and the identification text label; and iteratively adjusting network parameters of the initial speech recognition model according to the target loss result to obtain a speech enhancement recognition model.

The invention provides a millimeter wave phased-array antenna of a pure metal dipole and a packaging process of the millimeter wave phased-array antenna. The millimeter wave phased array antenna mainly solves the problems that a traditional millimeter wave phased array antenna is high in complexity, angle scanning of the antenna is limited, antenna packaging efficiency is low, and machining difficulty is large. According to the invention, a pure metal dipole antenna structure is adopted to complete an antenna radiation function, and metal walls are arranged at the periphery of the antenna, so that the influence of mutual radiation between antenna dipoles is avoided. The millimeter wave antenna array is integrally packaged and integrated on an Al2O3 ceramic packaging structure, and connection and packaging of the millimeter wave phased array antenna and the chip assembly are realized through a single-layer ceramic packaging layer. The antenna provided by the invention has the advantages of simpler structure, wide-angle scanning, low loss of a packaging process and high efficiency. The method can be used for realizing a high-integration and miniaturized millimeter wave packaging phased-array antenna.

The invention discloses an optical distortion test method, apparatus and device, a storage medium and a system, and relates to the technical field of virtual reality. According to the method, the intelligent glasses are controlled to display the test image containing the plurality of test points, the test points displayed by the intelligent glasses are sequentially shot through the image acquisition equipment, and in the obtained local images, the test points are located in the central areas of the corresponding local images, namely in the central area (near the optical axis) of the view field of the image acquisition equipment; the distortion influence of the lens of the image acquisition device is minimum, and then the test point of the central area of each local image is utilized to carry out optical distortion test on the intelligent glasses, so that the optical distortion of the intelligent glasses can be effectively separated from the lens distortion of the image acquisition device, the influence of the lens distortion of the image acquisition device is reduced, and the detection accuracy of the intelligent glasses is improved. And the real distortion level of the intelligent glasses can be accurately evaluated.

The present application discloses a camera detection device and a detection method for contrast detection probability. The camera detection device at least includes: a box body, a telescopic component, two first support components, two second support components, two first light box components and two first detection components; the two first light box components are respectively arranged in the box body through a first support component; one end of the two second support components are respectively connected to the two first light box components, and the other end is connected to the telescopic component; the two first detection components are respectively arranged on the light-emitting surfaces of the two first light box components. The camera detection device proposed in the present application can acquire data for contrast detection probability test, blur test, flicker test and high dynamic range test, integrates the four tests into one, and simplifies the test space and test process. The contrast detection probability detection method proposed in the present application can effectively reduce the interference caused by distortion for wide-angle lenses.

The present invention discloses an anti-interference evaluation method for rooftop photovoltaic potential based on satellite images. In order to overcome the problem that the prior art is difficult to distinguish whether there is interference and error in the roof recognition result; the present invention includes the following steps: obtaining a satellite image from the top of the building; dividing the satellite image at equal distances and identifying the boundary lines; restoring and combining the matrix image according to the original position and connecting it into a number of closed areas; matching the shape of the closed area with the database to determine whether it is a building roof; evaluating the accuracy of the building roof recognition result obtained according to the closed area through the roof recognition result output by the trained recognition model, and optimizing the building roof recognition result; calculating the building roof area, and evaluating the rooftop photovoltaic potential in combination with the energy consumption and environmental information of the station area where the building is located. The recognition results are verified and evaluated by different roof recognition methods, the recognition results and the recognition process are optimized, and the accuracy of the recognition results is improved.

The application discloses a cross-scene adaptive nutritional meal plan intelligent generation and execution management system, and relates to the technical field of computer information processing.The system builds a scene capability portrait of different application scenes, uniformly models and constraint-compiles nutritional targets, individual constraints and scene execution conditions, and generates executable nutritional meal plans matched with scenes such as hospitals, community medical institutions, enterprises and non-medical institutions.The application adopts a double-engine generation mechanism to improve the executability of the plans while meeting the nutritional constraints, and decomposes the generated plans into executable tasks for process management;through multi-source execution evidence collection and consistency evaluation, the application realizes objective evaluation of the plan execution, and automatically adjusts the plan when execution deviation or risk is triggered, forming a closed-loop adaptive management.The system can improve the adaptability, landability and management continuity of the nutritional meal plan in different scenes.

The disclosed embodiment provides a method for monitoring the depth of anesthesia based on a neural network, including: collecting physiological signal data of a patient during surgery; preprocessing the physiological signal data to reduce the influence of noise and distortion; constructing a neural network model adapted to a non-stationary state based on the preprocessed physiological signal data; using the neural network model to predict the physiological signal data in real time to determine the patient's depth of anesthesia; and adjusting the dosage of anesthetic drugs according to the predicted result of the depth of anesthesia. The solution of the disclosed embodiment can solve the problem of inaccurate model prediction caused by complex changes in physiological signals during surgery under a non-stationary state.

A method, system, storage medium, and program product for detecting non-methane total hydrocarbons (NMRs) are disclosed, relating to the field of testing or analyzing materials by measuring their chemical or physical properties. In this method, gas chromatographic data and component data are acquired, and the initial NMR concentration is calculated. CAS numbers are matched to distinguish known interfering components from dynamically added components. Preset response factors for known components and theoretical response factors for added components are obtained, and their accurate concentrations are calculated. The baseline NMR concentration deviation caused by interfering components calculated using the basic factors is calculated. This deviation is subtracted from the initial NMR concentration to compensate for the accurate concentrations of each interfering component, resulting in a corrected target NMR concentration. The target methane concentration is calculated and subtracted from the target NMR concentration to obtain the NMR concentration. This application aims to reduce the interference of dynamically changing added organic components on NMR detection, thereby improving the accuracy and intelligent processing efficiency of continuous NMR detection in complex waste gas scenarios.

The invention provides a field scene non-cooperative target measurement method, which comprises the following steps that: step 1, an unmanned aerial vehicle shoots multiple frames of field scene images through hovering, obtains high-precision self-positioning geographic coordinates of pixels corresponding to a cooperative target at corresponding moments in each frame of image, and establishes a control point set among the coordinates; step 2, fitting the control point set by using a quadric surface function, and obtaining a mapping relation of mutual mapping between coordinates of any pixel in the image based on fitting of each control point; and step 3, according to the pixel coordinates of the non-cooperative target in the aerial image, calculating to obtain the geographic coordinates of the non-cooperative target according to the mapping relation, and realizing the positioning measurement of the non-cooperative target in the field scene. According to the invention, the real-time position of the non-cooperative target such as the unmanned vehicle is used as the control point of the aerial image of the unmanned aerial vehicle in the field scene, and the positioning detection of the non-cooperative target on the flat ground in the image in the field scene is realized.