Patents
Literature
104 results about "Feature detection" patented technology
Efficacy Topic
Property
Owner
Technical Advancement
Application Domain
Technology Topic
Technology Field Word
Patent Country/Region
Patent Type
Patent Status
Application Year
Inventor
Feature detection is a process by which specialized nerve cells in the brain respond to specific features of a visual stimulus, such as lines, edges, angle, or movement. The nerve cells fire selectively in response to stimuli that have specific characteristics E.G shape, angle, or movement. Feature detection was discovered by David Hubel and Torsten Wiesel of Harvard University, an accomplishment which won them the 1981 Nobel Prize. In the area of computer vision, feature detection usually refers to the computation of local image features as intermediate results of making local decisions about the local information contents in the image; see also the article on interest point detection. In the area of psychology, the feature detectors are neurons in the visual cortex that receive visual information and respond to certain features such as lines, angles, movements, etc. When the visual information changes, the feature detector neurons will quiet down, to be replaced with other more responsive neurons.
A method for quickly detecting geometric size of track slab based on binocular camera
The present application relates to a kind of track board geometry fast detection method based on binocular camera, comprising: using binocular camera to collect track board point cloud data, local geometric center point is obtained by carrying out voxel filtering processing to point cloud data, and the position of rail platform is detected using geometric feature;Construct GeoTransformer network, introduce geometric self-attention mechanism and cross attention mechanism, fuse the distance and angle information between point pair, generate hybrid feature, based on the feature, construct Gaussian correlation matrix, and combine double normalization constraint to inhibit fuzzy matching, realize coarse registration by RANSAC algorithm, with the initial value of coarse registration result, carry out fine registration using ICP algorithm, output final transformation matrix, obtain complete track board point cloud model;Finally, based on the track board point cloud model obtained, extract track board length, width, height and other geometric dimensions, and compare with standard value, realize track board geometry fast detection.
Owner:EAST CHINA JIAOTONG UNIVERSITY
Face recognition methods, devices, systems, and storage media based on element detection.
ActiveCN116189270BCharacter and pattern recognitionEnergy efficient computingFeature setSingle image
This invention provides a face recognition method, device, system, and storage medium based on feature detection. The method includes: acquiring image data; inputting the image data into a face recognition model to identify several facial feature information; the face recognition model includes an encoding / decoding deep learning neural network and a predictive feature neural network; filtering identical facial feature information through an IOU module; classifying target features through a candidate box classification module, aggregating each facial feature information element and its related target elements into a feature set and assigning a unique code to each feature set; filtering each feature set through the IOU module; inputting each feature set into a logic judgment module for occlusion judgment, obtaining a mapping between the code and the occlusion judgment result; and outputting the face recognition result. This invention supports simultaneous recognition of multiple faces occluded in a single image, meets the real-time requirements of face recognition, and offers high accuracy and stronger robustness.
Owner:CHINA GUANGFA BANK
Method and apparatus for generating face image based on genetic rule
The application relates to a method and device for generating a human face image based on genetic rules, which comprises the following steps: obtaining a target kinship image and a target description text, the target kinship image being an image of an object having a blood relationship with a target object; extracting a target kinship feature from the target kinship image; detecting whether a target description feature is extracted from the target description text, the target description feature being used for describing face feature information of the target object; when the target description feature is not extracted from the target description text, calling a face generation model to process the target kinship feature to obtain a human face image of the target object; when the target description feature is extracted from the target description text, detecting whether the target description feature comprises a genetic description feature and a face description feature to obtain a detection result, and calling the face generation model to process the target kinship feature and the feature in the detection result to obtain the human face image of the target object.
Owner:BEIJING DEEPGLINT INFORMATION TECH
An Eye Feature Detection and Light Source Matching Method Based on Event Camera
This invention discloses an eye feature detection and light source matching method based on an event camera, belonging to the field of eye-tracking detection. The method includes first placing an event camera and a ring-shaped light source assembly in front of the eye being tested, controlling the ring-shaped light source assembly to switch from off to on, and simultaneously acquiring event stream data generated by brightness changes in the eye area through the event camera; constructing a global grayscale image and a light spot grayscale image based on the event stream data, then coarsely locating the pupil contour and extracting candidate corneal reflective spots; then obtaining the precise pupil contour and pupil center position based on the candidate corneal reflective spots; finally, based on a preset light spot geometric distribution model, eliminating abnormal spots among the candidate corneal reflective spots, and identifying the correspondence between each remaining candidate corneal reflective spot and the light source in the ring-shaped light source assembly. This invention achieves pupil and corneal reflective spot detection and light source matching based on event data, providing structured feature input for subsequent eye tracking and gaze direction estimation.
Owner:ZHEJIANG UNIV +1
A device for detecting micro-features of a pyrotechnic product with an adaptive light source
This invention provides a device for detecting the microscopic features of pyrotechnic items with an adaptive light source, relating to the field of pyrotechnic item detection technology. It includes a detection platform housing, a vibration-adaptive platform assembly on the upper end of the housing, a camera support arm integrated into the rear side of a visual camera support platform, a visual inspection camera at the lower end of the camera mount, a camera supplementary light at the lower end of the camera mount, and several item clamping assemblies fixedly mounted on the upper end of the vibration-adaptive platform assembly. This invention places pyrotechnic items on the vibration-adaptive platform assembly and clamps them with claw-shaped rubber heads at the front of the item clamping assemblies on its surface. Vibration monitoring sensors on the vibration-adaptive platform assembly detect external vibrations, enabling the assembly to adaptively absorb the generated vibration frequency and reduce the impact of vibration on pyrotechnic item detection.
Owner:LUZHOU VOCATIONAL & TECHN COLLEGE
Night road pit identification method and system based on low-illumination enhancement and parallax calculation
The application discloses a night highway pit and pond recognition method and system based on low-illumination enhancement and parallax calculation, and relates to the technical field of intelligent traffic monitoring and road maintenance. The method comprises the following steps: S1: collecting environmental data, calculating a meteorological index to determine a laser compensation power, and obtaining a stereo image pair to generate environmental metadata; S2: determining an enhancement strategy based on the environmental metadata, and obtaining an enhanced image pair by using an improved Retinex algorithm; S3: performing stereo matching on the enhanced image pair, and generating a scene depth map by fusing pose data; S4: based on a double-branch recognition model, fusing image and depth features, detecting and verifying a pit and pond area, and obtaining a pit and pond area mask; and S5: based on the pit and pond area mask, the depth map and the pose data, calculating three-dimensional parameters and a position of the pit and pond, and generating a detection report. Through adaptive image enhancement and multi-source data fusion technology, the accuracy of pit and pond recognition and the three-dimensional measurement precision in a night low-illumination environment are improved, and automatic and efficient inspection without affecting traffic is realized.
Owner:JIANGSU YANNING HIGHWAY PROJECT TECH CO LTD
Robot inspection method, device, equipment and storage medium
The application provides a robot inspection method, device, equipment and storage medium. The method comprises: acquiring multi-modal data of a scene to be inspected, including visual data and positioning data; performing event identification processing on the multi-modal data according to an end-side multi-modal large model configured for the robot to obtain an event identification result; the event identification processing refers to a processing process of performing abnormal feature detection on the visual data, performing position correlation verification on the positioning data, combining a preset inspection event rule to identify an event in the scene to be inspected, and obtaining the event identification result; the event identification result includes an event identifier, an event type and an event-associated multi-modal data index; generating a structured event report according to the event identification result through the end-side multi-modal large model; wherein the structured event report includes an event time, positioning data corresponding to the event, visual data corresponding to the event, an event type, an event record and text data corresponding to the event. The real-time inspection is improved.
Owner:CHINA UNITED NETWORK COMM GRP CO LTD +1
A VOCs concentration distribution monitoring system in a spray booth
This invention relates to the field of substance concentration measurement technology and discloses a VOCs concentration distribution monitoring system in a spray booth, including a feature detection module, a signal receiving module, a pneumatic protection module, a concentration inversion module, and a differential pressure adjustment module. The signal receiving module receives the transmitted signal through a central detector and captures the forward scattered signal using a ring detection array. The concentration inversion module determines the particle size shift of paint mist particles during the phase transition process based on the ratio of transmitted to scattered light intensity, thereby correcting the extinction section parameter in real time and outputting concentration data. The differential pressure adjustment module maintains a slightly positive pressure inside the pneumatic protection module that is 5 Pa to 15 Pa higher than the ambient static pressure. This invention utilizes the scattering physical characteristics to offset the measurement error caused by the nonlinear attenuation of paint mist particle size in situ, effectively eliminating the optical path signal interference caused by droplet evaporation in a multiphase flow field.
Owner:江苏锐达机械制造有限公司
Image forming apparatus
Provided is an image forming apparatus including: a transferor that includes a transfer body and a pressing roller which hold a recording material between the transfer body and the pressing roller, where the transfer body transfers an image onto the recording material, the pressing roller presses the transfer body, and the recording material is conveyed; a characteristic detector that detects characteristic information corresponding to a recording material characteristic of the recording material that is conveyed to the transferor; and a controller that performs adjustment, in which a pressing force of the pressing roller against the transfer body or an interval between the transfer body and the pressing roller is adjusted, at the transferor based on the detected characteristic information.
Owner:KONICA MINOLTA INC
Image automatic labeling method and device, electronic equipment and storage medium
This application provides an automatic image annotation method and apparatus. The method includes acquiring an image to be annotated; identifying a first position of a target feature in the image by performing target feature detection on the image to be annotated; determining a bounding box corresponding to the target feature at the first position according to a preset annotation rule; determining an annotation strategy based on the distribution type of the target feature in the image to be annotated; selecting a target bounding box from the bounding boxes determined according to the preset annotation rule according to the annotation strategy; and annotating the target feature in the image to be annotated using the target bounding box. This method can accurately and automatically annotate target features in an image, with low cost, high speed, and high accuracy.
Owner:TCL TECHNOLOGY GROUP CORPORATION
A deep learning-based analog front-end sensor noise filtering system and method
This invention discloses a deep learning-based analog front-end sensor noise filtering system and method, belonging to the field of sensor signal processing technology. It includes an analog front-end sensor, a signal preprocessing module, a differentiable noise coupling feature detection module, a programmable RC filter network, an ADC module, a deep learning fine filtering module, and a signal output module. By first preprocessing the sensor output signal, detecting noise coupling features, and then driving the programmable RC filter network to achieve analog domain coupling noise pre-suppression, followed by analog-to-digital conversion and digital domain fine filtering by a lightweight residual CNN model, a closed-loop linkage mechanism between the analog and digital domains is constructed. This invention achieves precise end-to-end suppression of coupling noise, solves the problem of low-frequency noise residue, adapts to a wide temperature range environment, improves system filtering stability and operational reliability, and is suitable for high-precision sensor signal acquisition scenarios.
Owner:深圳市鲸昕科技有限公司
Prototype-enhanced multi-instance learning-based brain disease ai prediction method and system
The application provides a brain disease AI prediction method and system based on prototype enhanced multi-instance learning, comprising: regarding the electroencephalogram signal of each subject as a bag, regarding each slice segment in the bag as an electroencephalogram instance, inputting the electroencephalogram instance into an instance coding module, performing electroencephalogram instance microstate multi-scale time sequence feature detection and space-time domain joint representation, obtaining space-time domain joint representation features of the electroencephalogram instance, inputting into a cross-instance self-attention modulation module to obtain modulated instance representation, inputting into a prototype calibrated enhanced multi-instance aggregation module, using the constructed and updated prototype as a representative mode of the instance in the bag, calibrating and utilizing prototype guided aggregation to obtain aggregation features, inputting into a classification head and outputting subject level brain disease prediction probability. The application can predict brain diseases.
Owner:HARBIN INSTITUTE OF TECHNOLOGY (SHENZHEN) (INSTITUTE OF SCIENCE AND TECHNOLOGY INNOVATION HARBIN INSTITUTE OF TECHNOLOGY SHENZHEN)
A ship radiated noise line spectrum estimation method based on multi-snapshot sparse Bayesian learning
The application discloses a ship radiation noise line spectrum estimation method based on multi-snapshot sparse Bayesian learning, and belongs to the field of underwater acoustic signal processing; the method utilizes the sparse characteristics of the ship radiation noise line spectrum, that is, the line spectrum is not uniformly distributed in the whole frequency range, but is discrete and sparse, converts the line spectrum estimation problem into a sparse signal recovery problem, so as to solve the problems in the background technology, under the condition that the data sample length is limited and the sparsity is unknown, the Bayesian framework utilizes probability modeling to adaptively adjust the hyperparameters from the global optimization angle, can effectively reduce the sidelobe, realizes high resolution, reduces the background noise fluctuation variance through multiple observation snapshots statistical average, enhances the signal-to-noise ratio of the line spectrum output, and improves the line spectrum feature detection performance.
Owner:THE 92899TH UNIT OF THE PEOPLES LIBERATION ARMY OF CHINA
An image tamper detection method for claim settlement materials
The present application relates to the technical field of image detection analysis, and particularly relates to an image tampering detection method for claim settlement materials, comprising the following steps: performing connected aggregation on the edge pixel set to obtain a soft edge suspicious area with edge feathering traces; generating a local autocorrelation anomaly map corresponding to a digital image of the claim settlement material to be detected; determining the target connected area as a splicing tampering area after edge post-processing and local geometric transformation, to output a tampering detection result of the digital image of the claim settlement material to be detected. The present application effectively reduces the false detection risk caused by single feature detection, improves the accuracy of tampering area recognition, and can intuitively mark the position and contour of the tampering area on the original image, so that the detection result is clearer and more intuitive, and is convenient for subsequent claim settlement audit and evidence analysis.
Owner:广州美保科技有限公司
Display device and button switching method
This application provides a display device and a button switching method in some embodiments. The method can detect the interface type of the current user interface in response to a button command input based on a third button. The interface type is obtained by performing feature detection on the recorded audio of the display device responding to the button command and on point images captured in the user interface. If the interface type is a non-live interface, the button function of the third button is set according to the button function of the first button to perform focus movement in the non-live interface. If the interface type is a live interface, the button function of the third button is set according to the button function of the second button to perform channel switching in the live interface. By detecting the interface type of the user interface, when the interface type is a live interface, the third button can directly control channel switching in the live interface, thus solving the problem that set-top boxes cannot use the display device's remote control to switch channels.
Owner:HISENSE VISUAL TECH CO LTD
An indoor dynamic scene unmanned aerial vehicle vision SLAM method, device and medium
This application discloses a visual SLAM method, device, and medium for UAVs in indoor dynamic scenes. The method includes: acquiring a color image; inputting the color image into a target detection network to obtain an initial detection box; simultaneously extracting ORB feature points from the color image; inputting the initial detection box into a target tracking module for target tracking and compensation to obtain a dynamic target tracking box; determining whether each ORB feature point falls within the dynamic target tracking box, performing dynamic feature detection on the feature points within the box, and obtaining a static feature point set by backprojection from the depth map; performing inter-frame pose calculation using the pure static feature point set, and determining whether the current frame is a candidate keyframe based on the pose change; if so, performing a two-stage visual similarity screening on the current frame to determine whether it is a redundant keyframe; if not, performing local BA optimization together with the current keyframe, keyframes within a preset range, and their co-existing static map points, while the keyframe participates in loop closure detection, outputting the accurate pose of the UAV.
Owner:QINGDAO UNIV OF SCI & TECH
Image forming apparatus
This reduces shock noise generated during transfer, allowing for optimal printing. [Solution] The transfer unit includes a transfer body that sandwiches a recording material being transported and transfers an image onto the recording material, and a pressing roller that presses the transfer body; a characteristic detection unit that detects characteristic information corresponding to the characteristics of the recording material being transported to the transfer unit; and a control unit that adjusts the pressing force of the roller against the transfer body or the distance between the roller and the transfer body in the transfer unit based on the detected characteristic information.
Owner:KONICA MINOLTA INC
A CAD construction drawing intelligent segmentation method and related equipment
The application discloses a CAD construction drawing intelligent segmentation method and related equipment, which realizes intelligent segmentation through steps such as graphic element recursive scattering analysis, double frame positioning identification and the like, acquires a CAD file and analyzes a basic graphic element set with geometric properties, identifies a frame, allocates graphic elements, and finally generates and outputs an independent sub-file containing corresponding graphic elements and attributes. Through recursive scattering and matrix transformation technology, the application deeply analyzes complex block entities and restores attributes, ensuring that graphic element information is not missed; in combination with the double mechanism of geometric feature detection and professional keyword verification, the application efficiently identifies non-standard and nested frames, breaking the limitations of traditional tools. A differentiated judgment system is established for different graphic element types, exclusive strategies are adopted to realize accurate classification of cross-frame elements, and information errors and omissions are avoided. The segmented sub-file is small in size and can be accurately distributed according to the profession, which not only optimizes multi-professional collaboration and reduces design conflicts, but also facilitates storage, retrieval and printing, and provides core support for communication drawing digitization.
Owner:GUANGDONG KAITONG SOFTWARE DEV
Feature and action based video behavior detection artificial intelligence method and robot
The application discloses a feature and action based video behavior detection artificial intelligence method and robot, which comprises a typical feature selection step, a typical feature detection step, a to-be-detected frame marking step, a to-be-detected video segment cutting step, a preset type of behavior action detection step and a to-be-detected video marking step. The method, system and robot can obtain typical feature categories of behavior actions by selecting typical features first, then detecting the typical features, and then detecting the preset type of behavior actions in those video segments with the typical features, so that the detection of many video segments without the typical features is reduced, and the detection speed and efficiency are greatly improved. The method is also applicable to real-time video acquisition, and the preset type of behavior action detection is not needed when there is no typical feature, so that the calculation cost is greatly saved, the detection speed is improved, and the real-time detection effect is achieved.
Owner:SOUTH CHINA NORMAL UNIV
Real-time detection and diagnosis system and method for intraoperative malignant biliary stricture based on DSOC
The present application relates to the detection and diagnosis of malignant biliary strictures, in order to solve the problem that the prior art cannot accurately locate and sample in real time during DSOC operation, the present application provides a real-time detection and diagnosis system and method based on malignant biliary strictures during DSOC operation, including data acquisition and quality control module for collecting video data during DSOC operation, output quality control frame sequence; The first stage morphological feature detection module is used for detecting and positioning a plurality of malignant related morphological features, and outputting the boundary box coordinates and the confidence; The second stage benign and malignant diagnosis classification module is used for carrying out benign and malignant two classification on the quality control frame sequence, and outputs the malignant probability and the two classification results; The explainability module is used for generating a heat map and spatial consistency analysis, and obtaining a visual video frame sequence; The deployment demonstration module is used for real-time receiving, displaying and storing the intraoperative DSOC video data, the boundary box coordinates and the confidence, the video level diagnosis result and the visual video frame sequence.
Owner:THE FIRST AFFILIATED HOSPITAL OF MEDICAL COLLEGE OF XIAN JIAOTONG UNIV
Civil aviation employee fatigue risk prediction method and system based on physiological signal monitoring
The application discloses a civil aviation employee fatigue risk prediction method and system based on physiological signal monitoring, relates to the technical field of fatigue risk detection, and comprises the following steps: target sign information is subjected to separate and associated reflection fatigue state variability index feature extraction, and first pretreatment signal features are comprehensively obtained; if the initial to-be-detected position of a civil aviation maintenance employee does not move in a space position area, a first feature difference value between environment data to which the initial to-be-detected position belongs and a preset environment data judgment threshold value is counted; if the space position area moves, a second feature difference value between environment feature data respectively belonging to a moving position area and a final landing position of the civil aviation maintenance employee is counted; and according to the first pretreatment signal features, the first and second feature difference values, and the correlation influence coefficient between the environment feature data and the sign index error, a real-time fatigue risk prediction result is obtained. The application can improve fatigue risk prediction efficiency.
Owner:CIVIL AVIATION FLIGHT UNIV OF CHINA
Computer vision-based structural vibration displacement monitoring method and system, and storage medium
The application provides a computer vision-based structural vibration displacement monitoring method and system and a storage medium, and the method comprises the following steps: step 1, structural vibration video preprocessing; basic information of video data is automatically acquired by using a video basic information automatic acquisition algorithm, and an image gray scale and local histogram equalization processing method is used to simplify the image calculation complexity and enhance the image features; step 2, structural displacement monitoring based on improved FAST corner detection; an automatic threshold calculation method based on the maximum inter-class variance is introduced to form an improved FAST corner detection algorithm, the image features obtained by the improved FAST corner detection algorithm are combined with the pyramid Lucas-Kanade optical flow method, and a whole-process automatic image feature detection and motion tracking scheme is established. The method has the advantages that the overall monitoring process is optimized, and the method is non-contact, simple to operate and high in automation.
Owner:HARBIN INSTITUTE OF TECHNOLOGY (SHENZHEN) (INSTITUTE OF SCIENCE AND TECHNOLOGY INNOVATION HARBIN INSTITUTE OF TECHNOLOGY SHENZHEN) +1
Utilization of sandboxed feature detection process to ensure security of captured audio and / or other sensor data
PendingUS20260154399A1Input/output for user-computer interactionCharacter and pattern recognitionOperational systemInteractor
Apparatus and methods for restricting egress of sensor data from a feature detection process to an interactor process. The sensor data can include audio data, image data, location data, and / or other sensor-based data. The feature detection process is sandboxed to restrict the egress of data from the component. Once the feature detection process determines that a feature has been detected in sensor data, the interactor process can be provided with the sensor data and / or additional sensor data. The sensor data and / or the additional sensor data can be provided directly by an operating system and not via the feature detection process. In some implementations, a notification can be rendered once data is sent to the interactor process. The notification can indicate that the sensor data is being accessed. Rendering of the notification can be suppressed when only the sandboxed feature detection process is accessing the sensor data.
Owner:GOOGLE LLC
Compact co-aperture three-waveband infrared optical system
This invention discloses a compact common-aperture three-band infrared optical system, comprising an infrared optical system body, a short-wave infrared imaging component, a mid-wave infrared imaging component, and a long-wave infrared imaging component. The infrared optical system body is sequentially arranged along the light incident direction with a Cassegrain afocal system, a secondary beam-shrinking structure, a first dichroic mirror, a second dichroic mirror, a short-wave infrared imaging component, a mid-wave infrared imaging component, and a long-wave infrared imaging component. All imaging components on the infrared optical system body are arranged in the axial space behind the primary mirror. The two dichroic separations achieve optical path separation, eliminating the need for complex folding and reflection structures, significantly reducing volume and weight, providing strong adaptability, simple processing and assembly, all-weather multi-feature detection, and simultaneously acquiring reflection features, high-temperature features, and room-temperature thermal radiation features, thereby improving target recognition rate.
Owner:SHANGHAI YANMU OPTOELECTRONIC TECH CO LTD
Video anomaly feature detection method and device, computer device, and storage medium
The application relates to the technical field of video anomaly detection, and discloses a video anomaly feature detection method and device, computer equipment and a storage medium, the video anomaly feature detection method comprises the following steps: obtaining target video data to be detected; performing first feature extraction on the target video data based on a text feature dimension to obtain semantic prior features; performing second feature extraction on the target video data based on a video feature dimension to obtain fusion visual features; wherein the fusion visual features are obtained by fusing normal visual features and abnormal visual features of the target video data; performing feature fusion based on the semantic prior features and the fusion visual features to obtain final features of the target video data; and performing abnormal feature detection based on the final features to obtain an abnormal feature detection result; the application can meet the demand of performing abnormal detection on large-scale video data, and can simultaneously detect multiple abnormal visual features during the abnormal feature detection.
Owner:BEIJING NORMAL UNIVERSITY
Fit prediction based on feature detection in image data
ActiveUS12664730B2Spectales/gogglesDetails involving processing stepsData acquisitionDirection information
A system and method of predicting fit of a wearable device from image data obtained by a computing device together with position and orientation of the computing device is provided. The system and method may include capturing a series of frames of image data, and detecting one or more fixed features in the series of frames of image data. Position and orientation data associated with the capture of the image data is combined with the position data related to the one or more fixed features, to extract depth data from the series of frames of image data. A three-dimensional model of is generated based on the extracted depth data. The three-dimensional model is processed by a simulator and / or a machine learning model to predict fit of the wearable device for the user.
Owner:GOOGLE LLC
Dynamic white LED lamp power optimization control method
The application relates to the technical field of camera imaging control, and provides a dynamic white light LED lamp power optimization control method, which comprises the following steps: an ambient light detection and reflection compensation correction step; a feedforward coarse adjustment step based on a corrected photosensitive value; a picture overexposure and reflection feature detection step based on ROI partition weight statistics; a gain-PWM joint optimization step under SNR constraint; a PWM gradual output and persistent monitoring step; and a state machine anti-oscillation hysteresis control step. Through the construction of a complete control closed loop comprising ambient light correction, feedforward coarse adjustment, image feature feedback optimization and state machine anti-oscillation, the application realizes dynamic self-adaptive adjustment of light supplement power according to ambient light and scene features, solves the problems of power consumption redundancy, image overexposure and switch oscillation caused by fixed power light supplement, significantly reduces the whole machine power consumption and improves the system reliability.
Owner:70MAI CO LTD
Method for detecting a target of a vessel
The application provides a ship target detection method, and relates to the fields of target detection and remote sensing image processing. The method comprises the following steps: acquiring a remote sensing image containing a ship target, and pre-processing the remote sensing image; inputting the pre-processed remote sensing image into a trained target detection model to obtain the category and position information of the ship target; wherein the target detection model is improved based on a YOLOv5 network and comprises a feature extraction module, a feature fusion module and a feature detection module; in the feature extraction module, a two-dimensional convolution layer of the YOLOv5 network is replaced by a deep separable large kernel convolution; in the feature fusion module, an additional small target extraction layer is added, and a channel attention module is added before each fusion for the fusion of multiple feature maps after up-sampling and down-sampling; and in the feature detection module, an anchor-free target detection head is configured for each input feature map. The application can improve the accuracy and precision of lightweight remote sensing image target detection.
Owner:INST OF SEMICONDUCTORS - CHINESE ACAD OF SCI
Airport baggage detection method and device, storage medium and electronic equipment
Owner:INSPUR SUZHOU INTELLIGENT TECH CO LTD
A three-dimensional object texture mapping method based on semantic boundary perception
The application discloses a kind of texture mapping methods suitable for indoor furniture three-dimensional model based on semantic boundary perception, including edge enhanced point cloud sampling, semantic boundary detection, bidirectional texture parameterization mapping and structure feature supervision etc. steps.The method is first carried out structure sensitive sampling to three-dimensional grid by dihedral angle analysis, to improve the point cloud density of edge and high curvature area;Subsequently, the semantic segmentation result is calculated and the dihedral angle is screened according to the threshold value, and the semantic boundary point corresponding to the fabric piece area splicing seam, edge rolling bar etc. position is identified, and it is used as the basis for texture joint optimization.Adopt bidirectional network structure of forward texture mapping and inverse three-dimensional reconstruction to generate texture UV coordinate, and combined with conformal loss, stretching loss, boundary loss, smoothness loss, clustering loss and consistency loss are optimized jointly, to improve the parameterized geometric fidelity and structure continuity.The application further constructs columnar feature detection module.
Owner:ZHEJIANG SCI-TECH UNIV