42 results about "Scale-invariant feature transform" patented technology

A fingerprint biometric key generation method based on scale invariance and a deep neural network. The object is small sample data, and fingerprint image signal processing knowledge is combined with deep neural network feature stability; fingerprint features are further stabilized at feature values by means of layer-by-layer processing of a deep neural network; and finally, a normal fingerprint image high-intensity key sequence is generated by means of a fingerprint key fuzzy extractor. A high-intensity stable fingerprint biometric key can be directly extracted from a human fingerprint small sample by means of processing such as scale invariant feature conversion and deep neural network re-encoding, and fingerprint image signal processing knowledge is combined with deep neural network feature stability to better support generation of a small sample fingerprint biometric key, thereby significantly reducing the false recognition rate during fingerprint biometric key generation.

The invention provides a two-stage SAR (Synthetic Aperture Radar) image registration method based on SAR-SIFT (Scale Invariant Feature Transform) and template matching, which comprises a pre-registration stage and a fine registration stage, and is characterized in that in the pre-registration stage, a down-sampling image is calculated through an SAR-SIFT algorithm to obtain an initial registration image; updating the initial registration image to obtain a final matching pair; after the final matching pair is mapped back to the original image size, a pre-registration image is obtained according to the final affine transformation matrix; in the fine registration stage, stable feature points are extracted from the pre-registration image blocks, and mismatching pairs are eliminated to obtain a fine registration image; and judging whether a region with relatively large topographic relief exists through the clustering mismatching pair, sampling a candidate topographic relief region, and then carrying out local correction to obtain a fine registration image again. The SAR-SIFT algorithm and iterative optimization are utilized to realize initial registration, a similarity measurement method is utilized to realize accurate registration of a flat region, adaptive stretching transformation is utilized to transform a topographic relief region to realize accurate registration of the topographic relief region, and finally accurate registration of the SAR image is realized.

The invention belongs to the field of 3D display, and provides a dynamic eyeball tracking method for naked eye 3D and a tablet personal computer, and the method comprises the steps: obtaining a video frame shot by the tablet personal computer, and carrying out the detection of eye and nose regions of the video frame based on an error learning framework; performing content classification on the eye and nose areas, extracting SIFT features in the eye and nose areas based on a scale invariant feature transform algorithm, and selecting a corresponding supervised descent model according to a content classification result to obtain first coordinates of pupil centers of human eyes; estimating a three-dimensional coordinate of the pupil center according to the 3D face model and the first coordinate, and dynamically adjusting a pixel arrangement mode of a screen of the tablet computer and the position of a cylindrical lens grating based on the three-dimensional coordinate, so that a left eye image and a right eye image projected by the tablet computer are matched with the three-dimensional coordinate of the pupil center; according to the invention, the problem that the naked-eye 3D effect cannot be achieved when the existing naked-eye 3D screen generates ghosting when the position of an observer is not correct enough or moves or the deviation exceeds a certain degree can be solved.

The application provides a kind of building facade disease image recognition and positioning method, by unmanned aerial vehicle with camera lens is directly opposite the angle of building facade and is shot to obtain multiple building facade local images;Data set is obtained from positive image sample and negative image sample;Building facade disease image recognition network model is constructed, and building facade disease image recognition network model includes skeleton network, neck network and detection head network;The optimal building facade disease image recognition network model is obtained;Disease prediction image is obtained, and scale invariant feature transform algorithm SIFT and random sample consensus algorithm RANSAC are used for splicing to obtain the final building facade overall recognition image.The method can realize the rapid and accurate identification of building facade disease, can improve the detection efficiency, has relatively accurate identification effect, has strong practicality and engineering significance.

The invention discloses a video three-dimensional reconstruction key frame extraction method and device based on mutual verification weighted matching, and the method comprises the steps: obtaining an input video stream, and carrying out the pre-classification of an image scene in the input video stream; calculating image feature points and descriptors by using an SIFT (Scale Invariant Feature Transform) detection algorithm, capturing dynamic changes of matching points of adjacent frames based on FLANN matching and pyramid LK optical flow calculation, and screening according to an intersection of FLANN matching and pyramid LK optical flow calculation results to obtain strong matching feature points; and based on the strong matching feature points, calculating vanishing points, endowing surrounding points within a preset range of the vanishing points with weights, constructing a robust index of the similarity of adjacent frames, obtaining the similarity of the adjacent frames, and realizing screening of the key frames of the video stream. According to the invention, the detection precision and stability are improved, and the robustness of the algorithm in diversified scenes is ensured.

The invention discloses an SIFT (Scale Invariant Feature Transform) algorithm hardware circuit implementation, which optimizes the hardware circuit implementation of an original SIFT algorithm, and solves the problems of slow operation of the SIFT algorithm and large resource consumption when the SIFT algorithm is deployed on an FPGA (Field Programmable Gate Array). According to the specific implementation scheme, the method comprises the steps that a one-layer six-group Gaussian pyramid is built, and an independent precalculated 21 * 21 Gaussian filtering kernel is used in each group; extreme points of three adjacent groups of differential pyramids are detected by adopting a threshold method, and edge response points are eliminated; a CORDIC algorithm is adopted to calculate the gradient direction and amplitude of a Gaussian image where the feature points are located, and a 16-row 16-column calculation result is output for gradient histogram statistics and descriptor generation; and normalization of a 128-dimensional descriptor is realized by using a single divider. The method has the advantages of low resource consumption, high real-time performance and the like, and is suitable for scenes, such as an FPGA (Field Programmable Gate Array), an ASIC (Application Specific Integrated Circuit) and the like, needing hardware acceleration of the SIFT algorithm.

The invention discloses a fast block division method for multi-view video coding, which comprises the following steps of: S1, reducing the resolution ratio of a coding frame in a sequence to be coded to be 1 / 2 of the original resolution ratio, and performing coarse intra-frame prediction and coarse inter-frame prediction by taking 16 * 16 pixels as units; s2, calculating an average offset vector Voffset (x, y) between the reference viewpoint and the referenced viewpoint by using an SIFT (Scale Invariant Feature Transform) feature extraction algorithm; s3, after the coding of the referenced viewpoint is completed, storing a depth map of the referenced viewpoint according to the division structure of each coding tree unit CTU of the referenced viewpoint; s4, for each CTU of the reference viewpoint, according to the position (x, y) of the CTU in the picture and the obtained average offset vector Voffset (x, y) before actual coding, constructing a depth region DR by taking the current to-be-coded block as a benchmark; s5, placing the constructed DR in the corresponding position of the referenced viewpoint, and summarizing the depth distribution of the depth map in the DR; s6, carrying out coding division on the reference viewpoint coding block; and S7, repeating the steps S4 to S6 on the coding blocks of all the reference viewpoints until all the coding blocks of the current viewpoint are coded.

The invention relates to the technical field of bioinformatics, in particular to a homologous protein searching method based on scale invariant feature transformation and pre-filtering, and the method comprises the steps: obtaining # imgabs0 #-carbon atom coordinates in a to-be-compared protein structure, calculating a corresponding distance matrix, extracting a feature description vector matrix of the distance matrix by using an SIFT (scale invariant feature transform) algorithm, and obtaining a feature description vector matrix of the to-be-compared protein structure; the matrix is mapped to a high-dimensional spherical surface, candidate protein structures meeting pre-filtering conditions are rapidly screened out from a large-scale protein database through a spherical surface partition pre-filtering algorithm, and the to-be-compared protein structure and the feature description vector matrix of the candidate protein structures are accurately compared through GPU acceleration. And generating a protein structure comparison result, and further determining the homologous protein corresponding to the to-be-compared protein. Therefore, the comparison speed is increased while the comparison precision is kept, the defect that efficiency and precision are difficult to consider in the prior art is overcome, and the method is suitable for rapid search and efficient comparison of a large-scale protein database.

The application discloses a target identification method based on monitoring video and relates to the technical field of target identification. The method comprises the following steps: acquiring a plurality of regional monitoring videos, extracting a preliminary monitoring image key frame sequence by optimizing an interframe difference method, and screening an optimized monitoring image key frame sequence by a secondary clustering method; adopting a scale invariant feature transformation algorithm to perform feature matching on the optimized monitoring image key frame sequence, obtaining a monitoring image sequence with the same feature, and performing image splicing on the monitoring image sequence by a video splicing method based on a random sample consensus algorithm, thereby obtaining a spliced monitoring image sequence; performing fusion on the spliced monitoring image sequence by a video fusion algorithm based on an adaptive threshold, thereby obtaining a fused monitoring image sequence; and performing identification on the optimized key frame sequence and the fused monitoring image sequence respectively by a target identification algorithm, obtaining a first confidence degree and a second confidence degree, and obtaining a corrected confidence degree by using an adaptive weighted fusion method.

The embodiment of the invention discloses a material label detection method and device, electronic equipment and a storage medium, and relates to the technical field of industrial visual inspection.The method comprises the steps that a label sample of a to-be-detected material is obtained, a drawing template is obtained according to a product number, and region frame selection is conducted on the drawing template through a yolov8 model; carrying out feature point detection by adopting SIFT (Scale Invariant Feature Transform) to obtain feature points, matching the feature points to obtain matching points, and mapping the label sample to the space of the drawing template according to the matching points; comparing the label sample with a drawing template through contour comparison, pixel absolute value comparison and text comparison, combining differences and filtering allowable offset to obtain a difference region of the drawing template and the material label sample; and switching and displaying the difference areas by using a GIF dynamic display technology, and carrying out frame selection and highlighting on the difference areas. The problems that in the prior art, manual comparison is relied on, efficiency is low, mistakes and omissions are prone to occurring, and identification and rapid difference positioning are difficult are solved.

The invention belongs to the technical field of fault diagnosis, and discloses a double-path incremental fault diagnosis method fusing an attention mechanism and a prototype, and the method comprises the steps: designing a double-branch network, and carrying out the division of labor through a heterogeneous network: extracting a time domain feature through a Hu moment-based convolutional neural network, extracting a frequency domain feature through a scale invariant feature conversion-based convolutional neural network, and carrying out the classification of the time domain feature and the frequency domain feature; the time domain branch network captures the global features of the operation trend of the equipment, and the frequency domain branch network extracts the local features of abnormal signals of a specific frequency band to form a time-frequency complementary feature space, thereby achieving the comprehensiveness of feature coverage. Considering that local frequency domain features of a high frequency band are generally more sensitive to early faults and time domain global features can visually reflect the overall operation state, an attention mechanism is introduced to realize adaptive distribution of weights, and the accuracy of feature extraction is further enhanced. Random projection is adopted to increase feature dimensions, the linear separability of features is improved, the problem of category overlapping in a low-dimensional feature space is solved, and category boundaries are expanded.

The application provides a computer vision-based automatic identification and positioning method for chip components, which includes the following steps: establishing the conversion relationship between the pixel coordinate system and the manipulator coordinate system, constructing a matching template based on the scale invariant feature transform (SIFT), and calculating the angle recognition error through rotation calibration; the formal target identification and positioning includes the following steps: preprocessing (the preprocessing includes converting the color space, bilateral filtering, enhancing the contrast, performing the morphological opening and closing operation, and detecting the contour using the Canny operator), matching the target by using the matching algorithm provided by OpenCV, identifying the chip component, and then calculating the coordinates and angle of the identified chip component in the manipulator coordinate system to achieve the purpose of positioning the identified chip component. The application has stronger universality, wider application range, and can better protect the chip component to be tested. In addition, the application is also applicable to non-polar chip components, and only the corresponding template needs to be replaced.

The present invention discloses an end-to-end 3D reconstruction method and device based on drone images, the method comprising: obtaining drone aerial images; screening key images of the aerial images based on a parallax analysis strategy to obtain a key image set; extracting features of the key images based on a scale-invariant feature conversion operator and a parallel processing strategy to obtain feature points of each key image, and performing pairwise matching calculations on the feature points between the key images based on a fast approximate nearest neighbor algorithm to obtain matching feature point pairs; estimating the camera pose based on the matching feature point pairs and a normalized RANSAC algorithm, and calculating the 3D coordinates of the feature points according to the camera pose to obtain a sparse point cloud; performing depth map calculation on the sparse point cloud based on a depth map strategy to obtain a dense point cloud; and constructing a triangular mesh on the dense point cloud based on a voxel reconstruction algorithm to generate a scene 3D model. The present invention improves the efficiency and accuracy of 3D reconstruction.

The invention discloses an unmanned aerial vehicle image retrieval method and system based on residual vector weighted aggregation, and the method comprises the steps: carrying out the feature extraction and description of an unmanned aerial vehicle image based on a scale invariant feature conversion algorithm, and generating a local feature of a preset scale; based on a feature aggregation algorithm, performing weighting processing on the residual vectors by using a residual vector weighting strategy to obtain global features after weighted aggregation of the residual vectors; performing global L2 removal normalization processing on the global feature to obtain a global feature descriptor vector of the unmanned aerial vehicle image; and performing image matching pair retrieval based on the global feature descriptor vector, and outputting a retrieval result corresponding to the unmanned aerial vehicle image. According to the unmanned aerial vehicle image matching pair retrieval method based on feature selection and residual vector weighted aggregation, the accuracy of matching pair retrieval of unmanned aerial vehicle images is improved.

The invention discloses a solidified soil damage mode recognition method and system based on digital speckles, and the method comprises the steps: obtaining a left group of speckle image series and a right group of speckle image series of measured solidified soil through a calibrated camera, and obtaining two groups of preprocessed image series; decomposing the reference image of the first group of preprocessed image series into first reference sub-region images through a sub-region matching algorithm; performing stereo matching on the second group of preprocessed image series based on a feature algorithm of scale invariant feature transformation to obtain a second reference subarea image; and performing time sequence matching on the first reference subarea image and the second reference subarea image to obtain an initial deformation displacement value of the solidified soil so as to confirm an identification damage mode of the measured solidified soil according to the displacement deformation value of the solidified soil. The method solves the problem that a traditional determination method of a traditional solidified soil damage mode mainly depends on analysis of a damage form and a stress-strain curve of cement solidified soil, but the analysis is only on a qualitative level, and the damage mode of the solidified soil is difficult to accurately identify.

The invention discloses a point cloud construction-based hook center line extraction method. The method comprises the following steps of firstly obtaining a hook image and performing preprocessing; feature point extraction is carried out on the preprocessed image by adopting an SIFT (Scale Invariant Feature Transform) algorithm; preliminary matching is carried out through similarity measurement of feature point descriptors, and wrong matching is removed by using an RANSAC (Random Sample Consensus) algorithm; calculating the three-dimensional coordinates of the feature points by using a triangulation method, generating initial three-dimensional point clouds, and registering and merging the initial point clouds obtained from different angles by using an ICP algorithm to generate high-precision hook three-dimensional point clouds; performing layering processing on the high-precision hook three-dimensional point cloud, performing circle fitting on a point set on each layer of cross section, determining the circle center of the cross section, and extracting the center line of the hook through a B spline curve fitting method; according to the method, by introducing a multi-angle imaging and controllable illumination system, details and contrast in the image are enhanced, and thus the feature point extraction precision is improved.

The invention is suitable for the technical field of computer vision, and provides a GraphMama-based dual-view image matching method, which comprises the following steps of: constructing a model which comprises an InitialProjection unit, a Preminarial Local ContextEncoding unit, a Consensus-Aware Learning Block unit, a GraphMama unit and an Inlier Predictor unit, and constructing a model which comprises an InitialProjection unit, a Preminarial Location unit, a Consensus-Aware Learning Block unit, a GraphMama unit and an Inlier Predictor unit; the method comprises the following steps of: performing high-dimensional feature mapping, local neighborhood enhancement, global consistency optimization and long-range dependence modeling by taking initial matching point pairs of a double-view image extracted by SIFT (Scale Invariant Feature Transform) as input, and performing Consensus-Aware Learning Block and GraphMama iteration for four times; and training the model by using supervised data sets such as YFCC100M, SUN3D and the like, and finally outputting correct matching point pairs and an essential matrix. According to the GraphMama-based double-view-angle image matching method, through the limitation that traditional MLP independently processes matching pairs and RANSAC robustness is insufficient, long-range dependence of long-sequence matching pairs is accurately captured, on YFCC100M and SUN3D data sets, mAP indexes of relative attitude estimation and feature matching are remarkably superior to those in the prior art, complex scenes such as view angle changes and shielding can be dealt with, and the method has the advantages of being high in adaptability, high in robustness and the like. The method can adapt to computer vision tasks such as image stitching and vision positioning.

The invention relates to the technical field of information hiding and digital watermarking, and discloses a tamper localization reversible watermarking method based on SIFT and quadrant partitioning, and the method comprises the steps: carrying out the quadrant partitioning of an original image, calculating the feature information of each quadrant, dividing the whole image into a checkerboard shape, and calculating the feature information of a black region of each block. And taking the two as the authentication watermark of the black area. Then, the authentication watermark of the black area is embedded into the black area through an inclined histogram, the Chinese remainder theorem is combined with ICMIC mapping to generate a new mapping mode, the watermark of the black area is embedded into the black area of the corresponding block through the mapping, the operation of the white area is the same as that of the black area, and the operation of the white area is the same as that of the black area. And secondary embedding is completed in the black and white areas. And respectively carrying out secondary authentication on the black and white areas of one block at a receiver. Compared with the prior art, the tampering area can be effectively and accurately detected and positioned, and certain invisibility is achieved.

The invention provides a digital ray image splicing and identification method and system based on high contrast characteristics. The method comprises the following steps: carrying out Gaussian filtering denoising and adaptive histogram equalization preprocessing on an aerospace digital ray image; extracting multi-scale high-contrast features by using a local binary pattern algorithm and carrying out series fusion; matching feature points through a scale invariant feature transformation algorithm, and estimating an affine transformation model in combination with a random sampling consistency algorithm to realize panoramic stitching; and constructing a convolutional neural network, and training the model by using a cross entropy loss function to complete panoramic image intelligent identification. The system correspondingly comprises a preprocessing module, a feature extraction module, a panoramic stitching module, an intelligent identification module and a data storage module. According to the method, the problems of low splicing precision and low recognition accuracy when the spaceflight digital ray image is processed by a traditional method are effectively solved, the image recognizability, the splicing precision and the target recognition accuracy are improved, and the requirements of detection of the internal quality of spaceflight parts and defect recognition are met.

The application discloses a kind of aerial photograph image difference comparison method, device, equipment and medium, by obtaining the first original aerial photograph image and second original aerial photograph image to be compared, using scale invariant feature transform algorithm to the first original aerial photograph image after pre-processing and the second original aerial photograph image after pre-processing are handled, obtain first feature image and second feature image;The similarity index between the first feature image and the second feature image is calculated;When the similarity index exceeds preset similarity threshold, the difference gray scale chart between the first original aerial photograph image and the second original aerial photograph image is calculated, and the difference gray scale chart is segmented according to threshold value to obtain gray contour line, and is labeled according to gray contour line, obtains aerial photograph image difference comparison result, the difference point between the aerial photograph of different period can be quickly identified and labeled in the present application, reduce the workload of relevant personnel, improve production efficiency.

The invention relates to the field of aircraft cluster cooperative localization, and discloses a multi-sensor vision-assisted cooperative localization method based on factor graph optimization, and each aircraft constructs an inertial navigation pre-integration information resolving model by using an inertial measurement unit add table and gyro measurement information. A visible light camera is utilized to realize synchronous observation of the same area from different visual angles, observation images are transmitted through wireless communication, a relative distance information measurement model between aircrafts is constructed, and based on a scale invariant feature transform (SIFT) algorithm, the same feature point in the observation images is identified and matched, and the relative distance information measurement model between the aircrafts is obtained. And constructing a sight line information measurement model of the relative feature points of the aircraft. In addition, the wireless communication device can communicate with star chain satellites, and a distance information measurement model between the aircraft and the satellites is constructed. By constructing a factor graph model, available navigation information is fused by utilizing a factor graph optimization algorithm, and state information, namely position, speed and attitude information, of the aircraft is estimated.

The embodiment of the invention relates to the field of material engineering, and provides a training method of a chopped fiber order degree evaluation model in meta-aramid insulation paper and a chopped fiber order degree evaluation method, and the method comprises the steps: obtaining a sample image of the meta-aramid insulation paper from a training sample of a data set; based on a scale invariant feature conversion algorithm, performing multi-scale feature extraction on the sample image; carrying out feature fusion on the sample features of each dimension; based on a principal component analysis algorithm, screening each sample feature in the sample comprehensive feature vector to obtain a sample dimensionality reduction feature vector; and training a chopped fiber order degree evaluation model constructed based on a support vector machine algorithm according to the sample label and the sample dimensionality reduction feature vector associated with the pre-labeled sample image. By adopting the method, the damage to the to-be-detected sample can be avoided, and the efficiency and accuracy of order evaluation are improved.

The invention discloses a fast motion and parallax estimation method for multi-view video coding, which comprises the following steps of: S1, taking all view images of the first ten frames of a video sequence, performing quadruple downsampling, and calculating an average offset vector Voffset (x, y) between a reference view and a referenced view by using an SIFT (Scale Invariant Feature Transform) feature extraction algorithm; s2, executing predictive coding of the first I frame of the main viewpoint and the first I frame of the auxiliary viewpoint to obtain a predictive frame after the predictive coding; s3, preparing non-I frame coding of the auxiliary viewpoint; and S4, sequentially executing the optimization on all CTUs of the current coding frame until all CTUs are coded.

The invention belongs to the technical field of image processing, and particularly relates to an image acquisition processing and contrastive analysis method of a private nursing instrument, which comprises the following steps of: synchronously acquiring six-axis attitude data, laser ranging object distance data and light supplementing lamp illumination parameters, performing SIFT (scale invariant feature transform) feature matching and attitude threshold accurate detection jitter, and performing image processing; then eliminating jitter angle deviation through motion model construction and pixel reprojection, realizing scale normalization in combination with an object distance proportion and bilinear interpolation, correcting illumination deviation by means of brightness compensation and a gray world algorithm, and finally respectively correcting pixel position deviation, scale difference, gray mean value difference and color temperature deviation of the two images, so as to obtain a final image. The same scene data basis is provided for subsequent comparison; on the premise, on the basis of pixel-level and feature-level multi-dimensional comparison, the misjudgment problem caused by direct comparison of non-uniform reference images in a traditional algorithm is reduced, and the comparison misjudgment rate is further reduced.

The invention relates to the technical field of soil moisture detection, in particular to a method and a system for detecting moisture distribution pattern and moisture content among soil particles. Firstly, a stress balance method is adopted for preparing and drying a test soil sample, and the structural integrity of the soil sample is kept; gradually adding water by using a dripping method, and synchronously collecting the weight change of the soil sample and microscopic image data; performing registration on the image sequence by using an SIFT (Scale Invariant Feature Transform) algorithm, including feature point detection, descriptor generation, feature point matching and RANSAC (Random Sample Consensus) optimization, and eliminating the influence of expansion deformation of the soil sample; and finally, establishing a moisture-optical correlation model, converting the image into an HSV color space, establishing a mapping relationship between the pigment tracer parameter and the moisture content by using a convolutional neural network, and generating a moisture distribution thermodynamic diagram. According to the method, interference of detection on moisture distribution is avoided, registration under the expansion condition is achieved, and the moisture distribution pattern is captured more accurately.

The invention relates to a multispectral image registration method, in particular to a multispectral image registration method based on improved SIFT (Scale Invariant Feature Transform) features. The method comprises the following steps: step 1, preprocessing multispectral remote sensing image data, wherein the multispectral remote sensing image data comprises a reference multispectral image and a to-be-registered multispectral image; step 2, extracting image gradient features based on the preprocessed multispectral remote sensing image data; extracting edge corner point features based on image gradient features; step 3, extracting an improved SIFT feature based on the image gradient feature and the edge corner feature; 4, performing homonymy point matching calculation on the reference multispectral image and the multispectral image to be registered based on the improved SIFT features to obtain a transformation matrix; and step 5, converting the multispectral image to be registered into the reference multispectral image based on the transformation matrix to obtain a final registration result. According to the method, the number of effective feature points in the image registration process can be increased, and rapid and accurate image registration is realized.

The embodiment of the application discloses a seal image space registration method and device, which can be used in the financial field or other technical fields, and the method comprises the following steps: extracting scale invariant feature transform features of a reserved seal image to obtain a first feature set; extracting scale invariant feature transform features of a to-be-inspected seal image to obtain a second feature set; matching each feature in the first feature set and the second feature set, in the feature matching process, automatically selecting appropriate hyperparameters by using adaptive hyperparameter search, screening feature pairs according to the hyperparameters to obtain a first matching feature pair, and then generating a first homography transformation matrix according to the first matching feature pair, wherein the hyperparameters comprise a feature pair matching threshold; and performing perspective transformation on the to-be-inspected seal image based on the first homography transformation matrix to obtain a space-registered to-be-inspected seal image. The application helps to improve the efficiency of seal authenticity inspection.

The invention discloses an asteroid surface feature extraction and matching method, which comprises the following steps of: firstly, extracting candidate feature points with scale invariance and rotation invariance by using an SIFT (Scale Invariant Feature Transform) algorithm, then filtering the SIFT feature points through angular point positions detected by a FAST algorithm, and reserving the SIFT feature points which are close to FAST angular points and are consistent in direction; therefore, the reliability and matching precision of the feature points are improved. According to the method, the problem of feature matching under the conditions of cross-scale, cross-view and violent illumination change can be solved, and the method has high engineering application value.

The application relates to a radar target recognition method, device and equipment based on a scale-invariant feature transformation. The method comprises the following steps: constructing a radar target recognition model; the radar target recognition model comprises a scale-invariant feature transformation feature extraction module, an encoder, a projection head and a classifier; after feature extraction on the augmented sample pair by the scale-invariant feature transformation feature extraction module and the encoder, the first mixed features are spliced, the first loss function is set after the first mixed features are mapped to the embedding space by the projection head; the encoder is trained by using the first loss function, and the trained encoder is obtained; the input radar image is subjected to feature extraction and splicing according to the trained encoder and the scale-invariant feature transformation feature extraction module, the second mixed features are input into the classifier, the second loss function is designed, the classifier is trained, and radar target recognition is performed according to the trained radar target recognition model. The method can improve the radar target recognition accuracy.

The invention belongs to an automatic assembly technology, and particularly relates to a high-speed workpiece interface positioning method based on scale invariant feature transformation, which comprises the following steps: S1, binocular image acquisition and distortion correction; s2, performing target detection on the left image based on a template matching algorithm; s3, feature extraction and matching are carried out on the left image and the right image based on scale invariant feature conversion; s4, calculating three-dimensional position information of the target based on a parallax method; and S5, performing hardware acceleration and optimization on the workpiece interface position detection method in the FPGA to obtain real-time target position information. According to the invention, the template matching algorithm and the SIFT algorithm are combined and realized on the FPGA, so that the real-time performance and flexibility of target position detection are effectively improved. The method has the advantages of high real-time performance, high flexibility and the like, and the efficiency and the flexibility of workpiece assembly can be remarkably improved.