92 results about "Local feature descriptor" patented technology

The invention proposes a vehicle type recognition and tracking method and system based on a monitoring video, and the method comprises the steps: A, background modeling and foreground detection: enabling a relatively static part in a video sequence to serve as the background and to be separated from the foreground comprising a moving object; B, vehicle image feature extraction: extracting local feature descriptors in images, and employing an SIFT feature descriptor because the method can describe the content and features of the images better, thereby enabling the image noise and affine changes to be stable to some extent; C, vehicle image feature coding: enabling the local feature descriptors with different numbers to be coded as the fixed-length vectors, so as to adapt to the input of a classifier; D, vehicle image recognition: selecting and designing an appropriate classifier for the classification of the characteristic vectors of the images, thereby finally achieving the recognition purpose; E, vehicle image tracking: tracking a recognition image region with one vehicle type being recognized, thereby avoiding the repeated foreground detection and recognition, and improving the operation speed of the system.

The invention discloses a real-time pose estimation method and a positioning grabbing system for a three-dimensional target object. The real-time pose estimation method comprises the following steps:acquiring three-dimensional graphic information of the target object; calculating to obtain a local feature descriptor of the target object according to the graphic information; performing three-dimensional pose estimation on the target object through the local feature descriptor by utilizing a pre-established three-dimensional model database to obtain pose information of the target object; and outputting the obtained pose information. According to the application, the real-time pose estimation method requested to be protected is applied to the positioning and grabbing system of the three-dimensional target object; the controller can control the motion mechanism to accurately grab the target object according to the pose information output by the processor, the grabbing accuracy can be effectively improved while the grabbing efficiency is guaranteed, and the practical performance of the positioning grabbing system in the application process is enhanced.

The invention provides a method for obtaining compact global feature descriptors of an image and an image retrieval method. The method for obtaining the compact global feature descriptors of the image comprises the following steps: obtaining at least a local feature descriptor of the image; selecting one or more local feature descriptors from all the local feature descriptors; carrying on dimensionality reduction to the selected local feature descriptors; obtaining the local feature descriptors after dimensionality reduction; converting the local feature descriptors after dimensionality reduction into global feature descriptors for expressing visual feature of the image according to a first rule; carrying on data compression to the global feature descriptors; and obtaining the compact global feature descriptors of the image, wherein byte size of the global feature descriptors can be varied according to the variation of parameter values in the first rule. Through the adoption of the method, the obtained global feature descriptors are more compact and have scalability, and the defect that a mobile terminal with lower internal memory has insufficient space in the prior art can be solved.

The invention provides a point cloud automatic registration method based on a local feature descriptor. an initial matching point pair can be obtained based on a local fast point feature histogram descriptor and a sampling consistency algorithm; A coarse registration matrix is obtained through an error measurement loss function, and then a fine registration rigid body transformation matrix is obtained by combining an iteration nearest point algorithm, so that the problem that three-dimensional point clouds cannot be automatically registered under different visual angles can be effectively solved; Compared with an existing method, the method has the advantages that mark points do not need to be pasted, the method is not affected by a positioning device, an additional auxiliary device is notneeded, the requirement for the environment is not high, efficiency is high, and the method has high robustness in actual measurement.

The invention discloses a method for identifying a traffic sign and belongs to the field of image processing. The method comprises the following steps of: extracting key points from an acquired image to be matched, establishing local feature descriptors, color feature descriptors and position feature descriptors respectively, extracting key points from the image to be matched and a template image in a template image library respectively to form feature point pairs to be matched, and finding a template image with most feature point pairs to be matched by judging whether position feature descriptors, color feature descriptors and local feature descriptors of the feature point pairs to be matched meet certain conditions, wherein the template image with most feature point pairs to be matched is taken as a finally identified traffic sign image for the image to be matched. By the method, the advantage that scale invariant feature transform (SIFT) features are invariant for the scale change and rotation of images is retained, and color and spatial position features of extracted feature quantities can be conveniently distinguished; and the method is extremely effective to the identification of traffic signs with rich colors and different spatial position distribution changes.

The invention discloses a shape matching and target recognition method based on a PCA-SC algorithm. The method comprises the steps of carrying out preprocessing on a target image, filtering part of noises in the target image, extracting the edge of the target image, extracting information of boundary contour points, working out the rectangular coordinate parameters of the contour points, converting the contour points from rectangular coordinates into polar coordinates, obtaining a corresponding logarithmic polar histogram of each point to forming a local feature descriptor, forming a covariance matrix, extracting a corresponding feature vector of a larger characteristic value of the matrix, adopting a linear transformation method to drop the matrix from high dimension to low dimension, forming a new characteristic matrix, wherein the new characteristic matrix is used for the shape matching and the target recognition, calculating matching degree, and obtaining a matching degree value between the target image and each template image. According to the shape matching and target recognition method based on the PCA-SC algorithm, characteristic extracting and effective representation for the image can be achieved, scale invariance, rotation invariance and translation invariance are achieved, accuracy rate and efficiency are improved, and interference of the noise is effectively restrained.

The invention relates to a method for constructing a compact image local feature descriptor. Compared with the prior art, the defects that an image local descriptor with high descriptive power is high in dimensionality, so that feature matching computing cost is large, and a common dimensionality reduction method influences the distinction degree and visuality of the image local feature descriptor are overcome. The method comprises the following steps that a feature region is determined; the feature region is partitioned and numbered; codes of leading central symmetry local binary patterns of points are worked out in the feature region; in units of partitioned sub-regions, a feature vector of the leading central symmetry local binary pattern of each partitioned sub-region is worked out; according to the sequence of the numbers of the partitioned sub-regions, the feature vectors of the leading central symmetry local binary patterns of all the partitioned sub-regions are arranged. The descriptor constructed with the method has the advantages of being low in dimensionality and high in descriptive power and distinction degree, high robustness of rotation transformation and illumination transformation of images is achieved, calculation is easy, and the matching speed is high.

The invention relates to a local feature description method based on three-dimensional point cloud. The method comprises the following steps: extracting a plurality of feature points from a scene point cloud, taking each feature point as a center, and establishing a three-dimensional local coordinate system according to points in a spherical neighborhood thereof; transforming the points in the spherical neighborhood of the feature points into the corresponding local coordinate system, and partitioning the spatial region of the spherical neighborhood along the radial direction. For each partitioned spatial region, the cosine values alpha and beta of the included angle between each point and the x-axis and z-axis of the coordinate system are calculated and mapped to two independent one-dimensional histograms, respectively. The one-dimensional histograms of all regions are connected in series and then divided by the total number of points in the spherical neighborhood of the feature points to normalize and obtain the final three-dimensional local feature description. The three-dimensional local feature descriptor provided by the invention has the characteristics of good discrimination, strong robustness and high computational efficiency, and improves the correct matching rate of similar parts between scene point clouds.

Feature descriptor matching described herein may include receiving a first input image and a second input image. A feature detector may detect features from the first and second input images. A descriptor extractor may learn local feature descriptors from the features of the first and second input images based on a feature descriptor matching model trained using a ground truth data set. The descriptor extractor may determine a listwise mean average precision (mAP) rank of a pool of candidate image patches from the second input image with respect to a queried image patch from the first input image based on the feature descriptor matching model, the first set of local feature descriptors, and the second set of local feature descriptors. The descriptor matcher may generate a geometric transformation between the first input image and the second input image based on the listwise mAP and a convolutional neural network.

The invention provides a massive image classification method based on a deep vector of locally aggregated descriptors (VLAD). The method is characterized in that through multi-layered image VLAD extraction, more abstract and richer image expression is finally obtained, so that the purpose of higher image classification accuracy is achieved. The method provided by the invention has the advantages that an adopted deep model can capture more abstract and richer feature information from an image during stepwise feature extraction, and average pooling operation is further conducted, so that captured image features achieve translation invariance, which is critical for the increasing image quantity and the image variety.

The invention belongs to the field of high resolution remote sensing image processing and particularly relates to a high resolution remote sensing image local feature extraction method based on 2D-Gabor. According to the method provided by the invention, a scale space pyramid expression of an image is firstly established; accelerated partition testing features of different feature scales are searched in the scale space, and a maximum value inhibition method is utilized to obtain a feature point and to determine the position and the scale of the feature point; then a local feature descriptor based on a binary system is established; and finally, a Hamming distance is used in a similarity measurement method to perform feature matching of images of the same scene under different perspective conditions, then an RANSAC algorithm is adopted to perform feature purification, and error matching point pairs are removed. The method provided by the invention can accurately simulate cognitive features of the visual cortex and the retina of human beings. In the feature detection process, an invariance property for change in brightness and scale is achieved, and optimal performances can be obtained at the same time in the time domain and the frequency domain.

The invention provides a method and a device for selecting local interest points. The method comprises the following steps of: acquiring all the local interest points of an image to be detected and the attributes of each of the local interest points; acquiring the importance of each of the local interest points according to the importance of the attributes; selecting at least one of the local interest points of the image to be detected by a preset selection rule according to the importance of the local interest points, wherein the importance of the attributes is obtained by training the attributes of each of the local interest points in each of a plurality of images; the preset selection rule is that the importance of the local interest points is greater than a set threshold value, and the quantity of the local interest points meets a set range. The local interest points selected by the method above are capable of reflecting the features of the image, and improving the accuracy of image indexing and image matching; meanwhile, local feature descriptors corresponding to the selected local interest points can be further polymerized to obtain global feature descriptors, thus improving the accuracy of the global feature descriptors in image indexing and image matching.

A method for compressing a local feature descriptor is provided in an embodiment of the present invention. The method comprises: selecting from a target image one local feature descriptor or a plurality thereof; according to a preset codebook, carrying out multistage vector quantization on the selected local feature descriptor, and quantizing same as a feature codestream, said codestream comprising serial numbers of codewords obtained by means of the multistage vector quantization. A device for compressing a local feature descriptor and a storage medium are also provided in embodiments of the present invention.

The invention discloses a non-rigid SAR image registration method based on region similarity and local spatial constraint. The problem that an existing rigid registration method is poor in registration effect when applied to a large SAR image is mainly solved. The method comprises the implementation steps that 1, two SAR images are input; 2, feature points of the input images are extracted, and local feature descriptor similarity is calculated; 3, feature point background region similarity is constructed; 4, feature point local spatial constraint conditions are constructed; 5, a matching cost function is constructed according to the step 2, the step 3 and the step 4; 6, iterative optimization is carried out on the matching cost function through a probabilistic relaxation algorithm to obtain the optimal matching point; 7, geometric deformation parameters are obtained according to the optimal matching point, and a registration result is obtained. Compared with the prior art, robustness on speckle noise and feature abnormality points is enhanced, the capacity of simulating non-rigid deformation is improved, the large actual measurement SAR image registration effect is improved, and the method can be used for image fusion and deformation detection.

The present disclosure generally relates to methods and computer program products for searching for a similar image among a plurality of stored images, and in particular to a method and computer program product used in a content based image retrieval system where roughly similar images are clustered and feature vectors for the clustered images are filtered based on a matching frequency for the feature vectors among the images in the cluster.

The invention discloses a three-dimensional identification and positioning method for sheet metal parts based on a PCL point cloud library. The three-dimensional identification and positioning methodcomprises the steps: obtaining a three-dimensional scene point cloud of an operation platform of a whole sheet metal part, carrying out the segmentation, and forming a point cloud cluster of each sheet metal part; calculating a local feature descriptor and a global feature descriptor of each point cloud cluster; performing feature-level fusion on the local feature descriptor and the global featuredescriptor of each point cloud cluster to obtain a fusion feature vector; inputting the fusion feature vector into a pre-trained SVM classifier for classification; solving an initial registration transformation matrix from point cloud clustering to point cloud model coarse registration; and determining an accurate transformation matrix conforming to the preset registration precision from the point cloud clustering to the point cloud model, and determining a pose information result of each sheet metal part according to the accurate transformation matrix. Through the technical scheme of the invention, the inaccuracy of two-dimensional image recognition is avoided, and the recognition efficiency and the positioning precision are greatly improved.

The invention relates to the technical field of medical lesion image recognition, in particular to a medical lesion image feature expression method based on region division and a Fisher vector. The method uses an expanded lesion region as a region of interest (ROI), divides the ROI into N sub regions according to the gray values of all pixels in the ROI; in each sub region, extracts small image blocks as local feature descriptors; aggregates the local feature set of each sub region into a vector by using a Fisher vector (FV) algorithm, and then connects the obtained N vectors end to end to obtain the feature expression of a medical lesion image. The feature expression method of the present invention utilizes the region information and space position information around the lesion, and uses the FV algorithm which is more effective than a conventional bag-of-word model to make the constructed feature expression more discriminative, thereby contributing to improving the accuracy of clinical adjuvant diagnosis.

The invention provides a video analysis method based on a local characteristic descriptor. The video analysis method mainly comprises video query, characteristic extraction based on deep learning, compact local characteristic descriptor encoding, video matching and video retrieval, and comprises the steps of firstly extracting a characteristic descriptor of a keyframe in a video, using a color histogram to conduct frame-level distance comparison, combining a manual design characteristic of a compact descriptor used for video analysis and deep learning based on a convolutional neural network, then achieving pair matching through comparison in a coarse-to-precise strategy, finally extracting a candidate keyframe in a database, and conducting sorting through video-grade similarity through further examination on local descriptor matching. In the video analysis method based on the local characteristic descriptor, the redundancy time of the video is eliminated, high-efficiency and low-delay mobile vision search is achieved, the memory size, the bandwidth resource and the cost during running are drastically saved, the compressibility is reduced, and the performance loss is lowered.

The invention provides a video motion recognition method based on fusion of sorting pooling and spatial features. The method comprises: on the basis of a video local feature descriptor algorithm, a basic visual feature vector set is extracted for each video; multi-scale segmentation is carried out on two-dimensional space for each frame of image in each video to construct a two-dimensional spatialpyramid model; video basic feature vector sets in each sub space in the pyramid model are arranged according to a frame sequence time sequence; smooth operation is carried out on an ordered basic feature vector sequence in each sub space independently; the ordered basic feature vector sequence after the smooth operation in each sub space is processed by using a sorting pooling algorithm and learning is carried out to obtain model parameters belonging to the sub space; the model parameters obtained in all sub space in the pyramid model are connected in series to obtain a feature vector as a final video feature vector; and a classifier is used for classifying the video feature vector and thus the motion type of the video is identified.

The invention discloses a robustness feature description method for images with noise, and belongs to the field of digital image processing. Firstly, a sampling method based on a local binary pattern is adopted to sample local images to construct sample point vectors; secondly, one-dimensional Fourier transformation is carried out on the collected sample point vectors to obtain local feature descriptors based on frequency domains; ultimately, a low-frequency filter is used, and a low-frequency part of feature vectors is selected as a final local feature descriptor. Because of frequency domain conversion and low-frequency filtering, the feature descriptor has good robustness to noise, and rotational invariance and gray scale invariance of an LBP descriptor remain.

An image recognition apparatus using a scalable compact local feature descriptor is provided. The image recognition apparatus includes a feature descriptor generator, a database, and a descriptor matcher. The feature descriptor generator extracts scalable compact local feature descriptor information for recognizing an object from input image information. The database includes information on a plurality of feature descriptors. The descriptor matcher compares a feature descriptor output from the feature descriptor generator with a plurality of feature descriptors stored in the database to recognize an object included in an image.

A local feature descriptor for a point in an image is generated over multiple levels of an image scale space. The image is gradually smoothened to obtain a plurality of scale spaces. A point may be identified as the point of interest within a first scale space from the plurality of scale spaces. A plurality of image derivatives is obtained for each of the plurality of scale spaces. A plurality of orientation maps is obtained (from the plurality of image derivatives) for each scale space in the plurality of scale spaces. Each of the plurality of orientation maps is then smoothened (e.g., convolved) to obtain a corresponding plurality of smoothed orientation maps. Therefore, a local feature descriptor for the point may be generated by sparsely sampling a plurality of smoothed orientation maps corresponding to two or more scale spaces from the plurality of scale spaces.

A compact image sequence descriptor (101), used for describing an image sequence, comprises a segment global descriptor (113) for at least one segment within the sequence, which includes global descriptor information for respective images, relating to interest points within the video content of the images. The segment global descriptor (113) includes a base descriptor (121), which is a global descriptor associated with a representative frame (120) of the image sequence, and a number of relative descriptors (125). The relative descriptors contain information of a respective global descriptor relative to the base descriptor allowing to reconstruct an exact or approximated global descriptor associated with a respective image of the image sequence. The image sequence descriptor (101) may further include a segment local descriptor (114) for a segment, comprising a set of encoded local feature descriptors.

The invention provides a video static logo detecting method based on local feature description. The video static logo detecting method includes the following steps: (1) an image is divided into blocks for logo detecting protection; (2) local features of the blocks are calculated, local feature descriptors of the blocks at the same positions of a front frame and a back frame are compared, and it is more likely to be detected to be local feature similar blocks if the similarity is higher; (3) single-frame similarity results of the blocks are accumulated on a timer shaft, and it is considered that a macro block is a logo after a certain threshold is reached. By means of the video static logo detecting method, the logo in a video sequence can be effectively detected, then the logo can be protected in a motion estimation and motion compensation module, and interpolation frame logo breaking is avoided.

The invention provides a global feature descriptor polymerization method. The global feature descriptor polymerization method includes the steps of obtaining local feature descriptors of a to-be-processed image, according to according to the significance of the local feature descriptors, ranking all the local feature descriptors, and obtaining the ranked local feature descriptors; according to interceptive values, selecting a plurality of local feature descriptors used for global feature descriptor polymerization from the ranked local feature descriptors; conducting polymerization on the multiple local feature descriptors through the adoption of a Gaussian mixture model so as to obtain a global feature descriptor of the to-be-processed image, wherein the interceptive values are obtained by training retrieved results of images in a preset image data set. By means of the method, the time complexity in the global feature descriptor polymerization process in the prior art can be reduced, and the discriminability and the robustness of the global feature descriptor are improved.

Feature descriptor matching described herein may include receiving a first input image and a second input image. A feature detector may detect features from the first and second input images. A descriptor extractor may learn local feature descriptors from the features of the first and second input images based on a feature descriptor matching model trained using a ground truth data set. The descriptor extractor may determine a listwise mean average precision (mAP) rank of a pool of candidate image patches from the second input image with respect to a queried image patch from the first input image based on the feature descriptor matching model, the first set of local feature descriptors, and the second set of local feature descriptors. The descriptor matcher may generate a geometric transformation between the first input image and the second input image based on the listwise mAP and a convolutional neural network.

The invention provides a method for obtaining global feature descriptors. The method comprises the steps of obtaining the local feature descriptors of an image to be processed and using the local feature descriptors to from a descriptor set, carrying out conversion on all the local feature descriptors in the descriptor set according to a global feature descriptor generating rule to obtain an accumulation gradient vector set including accumulation gradient vectors, and structuring one global feature descriptor of the image to be processed according to the accumulation gradient vectors in the accumulation gradient vector set. The global feature descriptor obtained through the method includes the statistical information of the local feature descriptors of the image to be processed, has strong discriminating ability and improves the processing performance of image retrieval or image classification.