Patents
Literature
Hiro is an intelligent assistant for R&D personnel, combined with Patent DNA, to facilitate innovative research.
89results about How to "Improve target detection efficiency" 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
Sea target detecting system of adaptive radar and method thereof
The invention discloses a sea target detecting system of an adaptive radar, which comprises a radar, a database, and an upper computer connected in order; the radar is used for irradiating the detected sea area and storing radar sea clutter data in the database; the upper computer comprises a data pre-processing module, a Robust prediction model modeling module, an adaptive swarm intelligent optimizing module, a target detecting module, a model updating module and a result displaying module. The invention further provides a sea target detecting method of the adaptive radar. The invention provides a sea target detecting system of the adaptive radar and a method thereof with adaptive parameter, good optimization effect and high prediction efficiency.
Owner:ZHEJIANG UNIV
Image target detection method and device and storage medium
ActiveCN109815868AFast target detectionImprove target detection efficiencyImage enhancementGeometric image transformationImage resolutionOverlap ratio
The embodiment of the invention discloses an image target detection method and device and a storage medium, and the method comprises the steps: obtaining a to-be-detected image, carrying out the zooming of the to-be-detected image according to different resolutions, and obtaining a plurality of zoomed images; respectively screening an area conforming to the type of the target area from each zoomedimage to obtain a plurality of initial image blocks; dividing each initial image block into a plurality of regions, and obtaining the probability that each region belongs to a target region; extracting an area of which the probability is greater than a preset threshold from each initial image block to obtain a plurality of candidate image blocks; and mapping the plurality of candidate image blocks to the to-be-detected image, and screening out an area where the candidate image blocks conforming to a preset condition are located according to the overlap ratio of the plurality of candidate image blocks to obtain a target area. According to the scheme, the requirement for computing resources is low, the detection speed is high, and the image target detection efficiency and the target detection accuracy are improved.
Owner:TENCENT TECH (SHENZHEN) CO LTD
Optical remote sensing image target detection method based on integrated deep convolutional network
ActiveCN108460341AMeet the characteristicsEasy to distinguishScene recognitionNeural architecturesData setComputer science
The invention discloses an optical remote sensing image target detection method based on an integrated deep convolutional network, and mainly solves problems in the prior art that the number of targets which are wrongly detected and a testing process is complex and tedious. The method comprises the following specific steps: (1), constructing a multi-branch deep network; (2), generating a trainingdata set containing a target area; (3), carrying out the first training of the integrated deep convolutional network; (4), generating training data sets of all regions; (5), carrying out the second training of the integrated deep convolutional network; (6), generating a test data set; (7), obtaining a test result map; (8), calculating an average precision. The method can achieve the extraction oftarget candidate frames of all no-target regions as negative samples, makes the most of the information of an optical remote sensing image, achieves the better discrimination of the target in the optical remote sensing image and a complex background, is simple in testing process, and is small in number of targets which are wrongly detected in a detection result.
Owner:XIDIAN UNIV
Multi-mode adaptive fusion three-dimensional target detection method
InactiveCN110543858AReduce inputReduce computational complexityThree-dimensional object recognitionComputation complexityRgb image
The invention discloses a multi-mode adaptive fusion three-dimensional target detection method, which is used for solving the technical problem of low detection efficiency of an existing three-dimensional target detection method. According to the technical scheme, the method comprise: inputting an RGB image and BEV Map; firstly, using an FPN network structure, comprising an encoder structure and adecoder structure, obtaining and using full-resolution feature maps of the FPN network structure and the encoder structure for being combined with bottom-layer detail information and high-layer semantic information, then extracting features corresponding to the two feature maps through feature clipping to be clipped and fused in a self-adaptive mode, and finally selecting 3D suggestions to achieve 3D object detection. The whole process is two-stage detection. In addition, the RGB image and the point cloud are used as original input, LIDAR FV input is reduced, the calculation amount is reduced, the calculation complexity of the algorithm is reduced, and the efficiency of three-dimensional space vehicle target detection is improved. According to the algorithm, the detection effect on smallobjects and the detection rate of shielded vehicles and intercepted vehicles are effectively improved.
Owner:NORTHWESTERN POLYTECHNICAL UNIV
Lightweight target detection method
PendingCN114120019AShorten forecast timeEasy to detectCharacter and pattern recognitionNeural architecturesFeature extractionAlgorithm
The invention discloses a lightweight target detection method. The lightweight target detection method comprises the steps of 1, performing data enhancement processing on a sample image; obtaining a prior bounding box size of the network model; step 2, constructing a target detection network model: the target detection network model is based on YOLOv4, a MobileNetv3 network is introduced to reconstruct a feature extraction network, standard convolution is replaced by deep separable convolution in PANet, and model parameter quantity and operand are reduced; after convolution operation is carried out on the feature layer with the same channel number, an improved CBAM attention mechanism is integrated, and the network detection performance is further improved; 3, training a target detection network model; and 4, detecting by using the target detection network model to obtain a detection result. The method has the characteristics of improving the target detection efficiency and reducing the network prediction time.
Owner:GUIZHOU UNIV
Iterative optimization distance categorization-based space weak and small target detection method
InactiveCN102096829AReduce computational complexityImprove target detection efficiencyCharacter and pattern recognitionFixed starsComputation complexity
The invention discloses an iterative optimization distance categorization-based space weak and small target detection method, which is used for solving the technical problem that the existing space weak motion target detection method is low in detection efficiency. The technical scheme of the method comprises the steps: extracting a plurality of candidate targets by a method based on the iterative optimization distance categorization; constructing an error square and a criterion function; categorizing all star points into fixed stars and non-fixed stars; iteratively computing a class mean andan error square function to obtain an optimization distance categorization threshold value; and filtering out the great mass of fixed star backgrounds and noise points, so that the complexity of the successor operation is reduced. A correlation method of a target track is used in the process of filtering the candidate targets, so that the computing complexity of the algorithm is reduced, and the detection efficiency of the space weak motion target is improved.
Owner:NORTHWESTERN POLYTECHNICAL UNIV
Image target detection method and system, electronic equipment and storage medium
InactiveCN110781819AImprove target detection efficiencyReduce communication data volumeCharacter and pattern recognitionNeural architecturesRandom access memoryEngineering
The invention discloses an image target detection method, which is applied to a parallel processing chip of a heterogeneous platform, and comprises the steps of receiving bounding box information of aplurality of target object bounding boxes obtained by performing target identification on an image through a convolutional neural network; storing the bounding box information to a random access memory RAM; executing a local maximum search operation based on a non-maximum suppression algorithm on the read bounding box information in a multi-path parallel manner to obtain a preset number of to-be-output bounding boxes; and transmitting the bounding box information of the bounding box to be output to a central processing unit (CPU) of the heterogeneous platform, so that the CPU outputs an imagetarget detection result. According to the invention, the communication data volume of the parallel processing chip and the CPU in the image target detection process can be reduced, and the image target detection efficiency is improved. The invention further discloses an image target detection system, electronic equipment and a storage medium, which have the above beneficial effects.
Owner:LANGCHAO ELECTRONIC INFORMATION IND CO LTD
End-to-end weak supervision target detection method based on frame regression of deep learning
InactiveCN110533067AImprove object detection accuracySimple network structureCharacter and pattern recognitionNeural architecturesFeature vectorObject Class
The invention discloses an end-to-end weak supervision target detection method based on frame regression of deep learning. According to the method, in a weak supervision convolutional neural network,a frame subjected to convolution layer and selective search is subjected to pyramid pooling layer and two full connection layers through a feature map, then a feature vector of a prediction frame is output, and then a full connection layer and a softmax layer on the category are connected; and finally, a prediction score corresponding to each object class in the selective search box is output. A box with the highest score of each class is selected as a pseudo annotation frame of the class; and the object frame predicted by the weak supervision model is regressed by using the frame with the highest score detected by each category as a pseudo labeling frame to generate a regression loss function, and a new loss function supervision weak supervision detection model is formed by the classification of the regression loss function and the weak supervision model and the positioning loss function. According to the invention, the detection time is reduced, and the target detection efficiency isimproved.
Owner:HANGZHOU DIANZI UNIV
Target detection method, device, terminal device and computer-readable storage medium
ActiveCN109272016AHigh precisionImprove detection accuracyCharacter and pattern recognitionPattern recognitionTerminal equipment
Owner:GUANGZHOU SHIYUAN ELECTRONICS CO LTD
Target detection method and device
ActiveCN108010060AImprove target detection efficiencyImage enhancementImage analysisNetwork modelComputer vision
Owner:BEIJING XIAOMI MOBILE SOFTWARE CO LTD
Image local feature matching method and device based on non-geometrical constraint and terminal
ActiveCN106295710AAchieving Mismatch EliminationReduce the probability of mismatchCharacter and pattern recognitionComputer terminalMatching methods
The invention provides an image local feature matching method and device based on non-geometrical constraint and a terminal. The method comprises steps of performing local feature detection on a target image and an image to be matched, to obtain local feature points of the images; performing local feature matching of the local feature points of the target image and the image to be matched, to obtain a plurality of first candidate matching feature points and a plurality of second candidate matching feature points, wherein each first candidate matching feature point matches a corresponding second candidate matching feature point, the first candidate matching feature point is located in the target image, and the second candidate matching feature point is located in the image to be matched; and performing neighbor constraint detection on the plurality of first candidate matching feature points and the plurality of second candidate matching feature points, and adding the first candidate matching feature points in accordance with the neighborhood constraint and the correspondingly-matched second candidate matching feature points to a matching feature point set. The technical scheme of the invention is advantageous in that, elimination of mismatching in types of images is effectively achieved.
Owner:JINGZAN ADVERTISING SHANGHAI CO LTD
Multi-scale deep convolutional neural network-based satellite image target detection method
InactiveCN110991359ARobustImprove detection results and the ability to distinguish different objectsScene recognitionNeural architecturesComputer visionTraining data sets
The invention discloses a multi-scale deep convolutional neural network-based satellite image target detection method. The method comprises the steps of: collecting a satellite image training data set, and carrying out sample labeling; preprocessing the satellite image training data set; building a multi-scale deep convolutional neural network; inputting the preprocessed training data set into a target detection framework based on the multi-scale deep convolutional neural network to perform training so as to obtain a trained target detection neural network; and inputting a to-be-detected satellite image set, performing target detection by adopting the trained target detection neural network, and outputting an identification result. With the method adopted, the accuracy of the result of thedetection of fine-grained characteristics by the network is improved; the capability of the network in distinguishing different objects is improved; and the effect of the detection of small objects and dense object groups is improved. The method has higher robustness and can effectively improve target detection efficiency and lower hardware requirements.
Owner:CHONGQING GEOMATICS & REMOTE SENSING CENT
Target detection method, device and equipment and readable storage medium
ActiveCN109784290AReduce missed alarmsReduce false alarmsCharacter and pattern recognitionForeground detectionFalse alarm
The invention provides a target detection method, device and equipment and a readable storage medium, and the method comprises the steps: obtaining a to-be-detected image in an image sequence in a target scene, and obtaining the occurrence condition of each type of targets in the target scene at each corresponding position of the to-be-detected image; and determining a target detection result corresponding to the to-be-detected image based on a preset foreground detection algorithm and occurrence conditions of various targets in the target scene in the to-be-detected image. According to the target detection method provided by the invention, false alarms caused by non-ideal foreground detection effects can be greatly reduced, missed alarms of static targets in a short time can be reduced, the calculation amount of target detection can be greatly reduced, and the target detection efficiency can be improved.
Owner:ANHUI IFLYTEK INTELLIGENT SYST
Marine radar target rapid detection method based on support vector machine
ActiveCN110221266AImprove target detection efficiencyReduce false alarm rateICT adaptationRadio wave reradiation/reflectionFeature parameterFalse alarm
The invention discloses a marine radar target rapid detection method based on a support vector machine. The method includes the steps that first, a field observation test is carried out offline, a certain amount of two types of radial echo samples on a single scanning line are selected, one type of radial echo samples contain target radar echoes, the other type of radial echo samples do not contain target pure sea clutter radar echoes, two appropriate radial echo characteristic parameters are extracted according to the characteristics of the samples, and a classifier model is obtained by training through an SVM; second, data to be measured on the radar single scanning line are acquired, the values of the data to be measured in the two radial echo characteristics in the first step are calculated and input to the classifier model, and a corresponding decision function value K is obtained; and final, the obtained decision function value K is compared with a threshold value of a classifier, and whether a target exists on the single scanning line or not is judged. According to the method, the target detection efficiency can be effectively improved, and false alarms can be reduced.
Owner:HARBIN ENG UNIV
Multi-scale target detection method and device
ActiveCN110263732AExpand the receptive fieldScaling DensificationCharacter and pattern recognitionComputer visionConvolution
The invention provides a multi-scale target detection method and device. The method comprises the steps of obtaining an image to be subjected to target detection; carrying out first convolution processing on the image to obtain the feature maps of multiple scales; for each scale in the plurality of scales, carrying out hole convolution processing on the feature map of the scale to obtain a feature map after hole convolution; processing the feature map of the scale according to the feature map subjected to the cavity convolution to obtain a processed feature map of the scale; carrying out fusion processing on a processed feature map of each scale in a plurality of scales, determining a target detection result of an image according to the fused feature map of each scale, wherein a receptive field is increased through the adoption of the cavity convolution, the scale division is densified, and the target detection efficiency is improved.
Owner:BOE TECH GRP CO LTD
Target detection method and target detection device
InactiveCN108875763AImprove target detection efficiencyAvoid errorsCharacter and pattern recognitionNeural architecturesPattern recognitionConsecutive frame
Owner:BEIJING KUANGSHI TECH +1
A method and apparatus for image processing
InactiveCN109308516AImprove target detection efficiencyReduce calculationImage analysisNeural architecturesImaging processingImage detection
Owner:HUAWEI TECH CO LTD
Remote sensing video object detection method based on overlapping slices
ActiveCN109409285AImprove detection rateComprehensive detectionScene recognitionTest sampleComputer science
The invention discloses a remote sensing video target detection method based on overlapping slices, which mainly solves the problems of missed detection rate of targets, high false detection rate of targets and tedious, complex, time-consuming and low efficiency of target detection process in the prior art. The specific steps of the invention are as follows: (1) selecting source data; (2) generating a training sample set and a test sample set; (3) preprocessing the training sample; (4) pretreating the test sample with overlapping slices; (5) training target detection model; (6) detecting targets in overlapping slices; (7) detecting the target in the test sample; (8) converting the test sample picture to video. The invention can reduce the missed detection rate of the target and the false detection rate of the target, improve the detection accuracy of the remote sensing video target, shorten the time consuming of the target detection process, and realize the efficient detection of the airplane and the ship targets in the remote sensing video.
Owner:XIDIAN UNIV
Video monitoring target detection method and device
ActiveCN103336951AReduce the number of zoomsImprove target detection efficiencyCharacter and pattern recognitionClosed circuit television systemsVideo monitoringSlide window
The invention discloses a video monitoring target detection method and device. The method comprises the following steps: detecting the first N frames of input images by a sliding window detection method, determining the target scale range of each pixel point in the images and determining image scaling parameters corresponding to target scales, wherein N is greater than 1 and is an integer; according to the target scale range of each pixel point, dividing the images into a plurality of scanning zones; determining the zone target scale range of each scanning zone according to the target scale ranges of the pixel points in the scanning zone and determining an image scaling parameter corresponding to the scanning zone; as for the scanning zones of subsequently input images, detecting targets in the images by the sliding window detection method according to corresponding image scaling parameters. Through application of the technical scheme provided by the invention, the target detection efficiency can be improved.
Owner:ZMODO TECH SHENZHEN CORP
Grassland spodoptera litura adult image detection method used under lamp trapping equipment
ActiveCN110428413AImprove reliabilityGuaranteed detection accuracyImage enhancementImage analysisLawnRapid detection
The invention relates to a grassland spodoptera litura imago image detection method used under lamp trapping equipment. Compared with the prior art, the defect that grassland spodoptera litura imago rapid detection is difficult to conduct under the lamp trapping equipment is overcome. The method comprises the following steps: collecting and preprocessing training samples; constructing and traininga grassland spodoptera litura adult detection model; collecting and preprocessing adult images of the grassland spodoptera litura under the lamp inducing equipment; marking the grassland spodoptera litura adults under the lamp trapping equipment. According to the method, the defect that a large number of redundant candidate boxes generally need to be generated in a target detection method is overcome, the reliability of data is enhanced, candidate areas are greatly reduced while the detection precision of the grassland spodoptera litura adult image is guaranteed, and the detection speed is increased.
Owner:HEFEI INSTITUTES OF PHYSICAL SCIENCE - CHINESE ACAD OF SCI
Remote sensing image ship detection method based on block extraction
ActiveCN111027511AKeep targetOptimize water and land segmentation resultsImage enhancementImage analysisData setImage scale
The invention discloses an optical remote sensing image ship detection method based on block extraction of interest, and mainly solves the problems of low detection precision and more false alarms inthe prior art. The method comprises the following steps: constructing an optical remote sensing image ship detection data set; carrying out downsampling and defogging enhancement on the wide remote sensing image, and carrying out land and water segmentation by using context information and image global features; training an SCRDet-based target detection model by using the constructed data set; according to the land and water segmentation result, scanning the original wide remote sensing image by using a partially overlapped sliding window to extract an interested block as a to-be-detected area, and inputting the to-be-detected area image into the detection model to obtain an area detection result; mapping a region result to an original wide image scale, and performing improved non-maximumsuppression to optimize a preliminary detection result; and optimizing the detection result again according to the structural characteristics of the ship. The method is high in detection precision and low in false alarm rate, and can be used for acquiring ship targets of interest and positions thereof in large-format remote sensing images.
Owner:XIDIAN UNIV
Target detection method of image data and related device
InactiveCN111339891AGuaranteed accuracyImprove target detection efficiencyCharacter and pattern recognitionNeural architecturesMachine learningImaging data
The invention discloses a target detection method for image data, and the method comprises the steps: processing a to-be-detected image through employing an anchor-frame-free target detection network,and obtaining an initial candidate frame; detecting the to-be-detected image according to the initial candidate box by adopting a convolutional network and a full-connection network respectively to obtain a convolution point type result, a convolution regression result, a full-connection classification result and a full-connection regression result respectively; and screening the convolution integral result, the convolution regression result, the full-connection classification result and the full-connection regression result through a score function to obtain a classification result and a regression result. By combining an anchor-frame-free target detection algorithm and a two-stage target detection algorithm, the precision and accuracy of target detection are improved on the premise of ensuring the target detection efficiency. The invention further discloses an image data target detection device, a server and a computer readable storage medium, which have the above beneficial effects.
Owner:INSPUR SUZHOU INTELLIGENT TECH CO LTD
Reflective real-time infrared polarization dual-separation imaging optical system
PendingCN109884803ASolve the error that is easy to introduce due to environmental factorsIncreased complexityOptical detectionOptical elementsInformation processingWollaston prism
The invention discloses a reflective real-time infrared polarization dual-separation imaging optical system. The reflective real-time infrared polarization dual-separation imaging optical system comprises a Cassegrain reflecting assembly, a Wollaston prism, an imaging infrared lens set and a detector assembly; the Cassegrain reflecting assembly is used for compressing parallel light and then outputting the parallel light, and the parallel light is emitted into the Wollaston prism; the Wollaston prism is used for decomposing the parallel light into two beams of polarized light; the imaging infrared lens set is used for converging the two beams of polarized light, and separating and imaging the two beams of polarized light to the detector assembly; the detector assembly comprises a detectorimage plane and a detector cold shield, the detector image plane is used for detecting two polarizing images formed by the two beams of polarized light, and the detector cold shield is used for reducing radiation. According to the reflective real-time infrared polarization dual-separation imaging optical system, light rays are collected through a Cassegrain reflecting system, and more infrared energy can be obtained; and the two polarizing images of the same scene can be obtained in real time through the Wollaston prism, the two polarizing images are subjected to image information processing,and the target detecting efficiency can be improved.
Owner:11TH RES INST OF CHINA ELECTRONICS TECH GROUP CORP
Material quantity detection method and device, electronic equipment and storage medium
PendingCN112348835AAvoid the phenomenon of target missed detectionFlexible detection and countingImage enhancementImage analysisEngineeringImage pair
The invention discloses a material quantity detection method and device, electronic equipment and a storage medium. The method comprises the steps of acquiring a to-be-detected image and the size of the to-be-detected image; when the size of the to-be-detected image is greater than a preset value, performing segmentation processing on the to-be-detected image to obtain at least two to-be-detectedsub-images; performing target detection on each to-be-detected sub-image, and determining position information of a target in each to-be-detected sub-image; and determining the number of the targets in the to-be-detected image according to the position information of the targets in each to-be-detected sub-image. According to the invention, the to-be-detected image greater than the preset value issegmented, target detection is carried out on each segmented to-be-detected sub-image, the position information of each target is determined, and the number of the targets is finally determined; the phenomenon that the target information is not significant and target missing detection occurs due to the fact that the to-be-detected image with an overlarge size is compressed is avoided, and the targets in the to-be-detected image can be flexibly, efficiently and accurately detected and counted at the mobile terminal.
Owner:GLODON CO LTD
High and low frequency interleaved edge feature enhancement method suitable for pedestrian target detection and method for constructing enhancement network
PendingCN112115871AImprove feature extraction efficiencyImprove generalization abilityBiometric pattern recognitionNeural architecturesPattern recognitionEngineering
Owner:DALIAN NATIONALITIES UNIVERSITY
Comprehensive feature target detection method and system in intelligent monitoring network
ActiveCN110008888AReduce resource usageImprove accuracyCharacter and pattern recognitionSmart surveillanceMonitor equipment
The invention provides a comprehensive feature target detection method and system in an intelligent monitoring network. The method comprises the following steps: defining and extracting a plurality ofpieces of feature information related to target detection in a monitoring network; screening monitoring devices in combination with the plurality of pieces of feature information, releasing tasks, and unloading the tasks of the monitoring devices to a server; and judging by utilizing the feature information of the monitoring equipment to obtain a detection result, and carrying out information fusion on the detection result. According to the method and the system, comprehensive characteristics such as time characteristics of a detection target and spatial characteristics of monitoring equipment are utilized, so that the target detection task can be better distributed and executed on the monitoring devices, and the spatial range of the detection target is finally determined by utilizing thecharacteristics and the detection result, so that the resource use of the monitoring network is effectively reduced while the higher accuracy is ensured. And the method has a good application value for a target detection task of deep learning, especially for target detection of video contents.
Owner:SHANGHAI JIAO TONG UNIV
Quick deep learning remote sensing image target detection method based on candidate region screening
ActiveCN108596952AImprove target detection efficiencyImprove object detection performanceImage enhancementImage analysisImage resolutionNetwork model
The invention relates to a quick deep learning remote sensing image target detection method based on candidate region screening and belongs to the field of remote sensing image processing. The methodcomprises the steps that (1) target training samples are utilized to train a deep convolution network model; (2) target geometric feature conditions are input, including target length, width and a length-to-width ratio range; (3) image parameters are analyzed, mainly including resolution information in two directions; (4) according to the resolution information, target geometric features are converted into pixel limiting conditions; (5) a remote sensing image target segmentation technology is utilized to realize target candidate region extraction; (6) according to the target geometric information size limiting conditions in the step 4, target candidate regions in the step 5 are screened; and (7) feature recognition is performed on the target candidate regions obtained after screening in the step 6 to determine a target. According to the method, target geometric feature information is utilized to screen the candidate regions, so that target detection efficiency is improved, and meanwhile false warning is suppressed to some extent.
Owner:NO 54 INST OF CHINA ELECTRONICS SCI & TECH GRP
Weak and small moving target detection method based on superpixel adjacent frame feature comparison
ActiveCN110910417AEnhanced Representational CapabilitiesClustering is fastImage analysisAlgorithmGraph theoretic
The invention relates to a weak and small moving target detection method based on superpixel adjacent frame feature comparison. The method comprises the following steps of: 1, performing super-pixel segmentation on a target frame image by using an SLIC (simple linear iterative clustering) algorithm; 2, generating a graph theory model according to the adjacency relation of the superpixels, designing superpixel features, calculating feature differences among the superpixels, taking the feature differences as edge values in the graph theory model, and realizing potential target extraction by utilizing a graph segmentation algorithm; and 3, performing adjacent frame color feature comparison on the potential target area, and if the color feature difference exceeds a set threshold, marking the potential target area as a moving target. According to the method, inter-frame search is avoided, and the efficiency of detecting the weak and small moving target is improved.
Owner:NORTHWESTERN POLYTECHNICAL UNIV
Optical remote sensing image target detection method based on geometric structure double-path convolutional network
ActiveCN110188682AHigh positioning accuracyTo overcome the disadvantage of weak responseGeometric image transformationScene recognitionData setCanis lupus familiaris
The invention discloses an optical remote sensing image target detection method based on a geometric structure double-path convolutional network. The optical remote sensing image target detection method comprises the following steps: constructing a training data set T in an image block-sketch block-label mode by using a labeled optical remote sensing image data set; constructing a test data set Uin an image block-sketch block mode by using an optical remote sensing image to be detected; constructing a target detection model based on the geometric structure double-path convolutional network,wherein the target detection model based on the geometric structure double-path convolutional network comprises a regional convolutional module and a DoG ridgelet basis function convolutional module;using the training data set T to train a target detection model based on the geometric structure double-path convolutional network to obtain a trained target detection model based on the geometric structure two-way convolutional network; and inputting the test data set U into the trained target detection model based on the geometric structure two-way convolutional network to obtain a detection result of the optical remote sensing image to be detected. According to the invention, the positioning precision of the target detection model is effectively improved.
Owner:XIDIAN UNIV
Embeddable non-contact elevator key interaction method fused with face recognition
ActiveCN113220114AHigh precisionAvoid interferenceInput/output for user-computer interactionImage enhancementKey pressingEdge maps
The invention discloses an embeddable non-contact elevator key interaction method fused with face recognition, which comprises the following steps of: firstly, carrying out edge detection on a shooting area in an original image through a Laplace filtering operator to obtain an edge image, and filtering the edge image by utilizing a linear filtering operator in the horizontal direction and a linear filtering operator in the vertical direction; then adopting a Hough straight line detection algorithm to conduct straight line detection on the images obtained after horizontal filtering and vertical filtering so as to position the area of the elevator key panel, and solving a homography transformation matrix; then using an improved YOLOv3 algorithm for detecting and positioning the fingers of the elevator user, obtaining floor keys pointed by the fingers according to the homography transformation matrix, and meanwhile obtaining face information of the resident for dual verification. According to the elevator key recognition system, the elevator keys selected by the elevator user can be accurately recognized, non-contact elevator taking can be achieved, and the safety of residents is guaranteed through double verification of floor information and resident face information.
Owner:SOUTH CHINA UNIV OF TECH
Who we serve
- R&D Engineer
- R&D Manager
- IP Professional
Why Eureka
- Industry Leading Data Capabilities
- Powerful AI technology
- Patent DNA Extraction
Social media
Try Eureka
Browse by: Latest US Patents, China's latest patents, Technical Efficacy Thesaurus, Application Domain, Technology Topic.
© 2024 PatSnap. All rights reserved.Legal|Privacy policy|Modern Slavery Act Transparency Statement|Sitemap