65 results about "Binary trait" patented technology

The present invention relates to a method of recognizing the human iris using the Daubechies wavelet transform. The dimensions of characteristic vectors are initially reduced by extracting iris features from the inputted iris image signals through the Daubechies wavelet transform. Then, the binary characteristic vectors are generated by applying quantization functions to the extracted characteristic values so that the utility of human iris recognition can be improved as the storage capacity and processing time thereof can be reduced by generating low capacity characteristic vectors. By measuring the similarity between the generated characteristic vectors and the previously registered characteristic vectors, characteristic vectors indicative of the iris patterns can be realized.

The invention relates to the technical field of image depth calculation, in particular to a depth estimation method and device based on near-infrared laser speckles. According to the depth estimation method based on the near-infrared laser speckles, through preprocessing conducted on a target speckle image, influences of ambient illumination are eliminated from the target speckle image, and therefore accuracy of depth estimation is improved. Through the fact that speckles are described by the utilization of binary system characteristics, huge difference of brightness and the contrast ratio between the target speckle image and a reference speckle image is effectively offset. Through quick information spread, the depth estimation can be conducted in high efficiency, and meanwhile, due to the fact that parallax error information of an area provided with fewer speckles in the target speckle image is added, the result of the depth estimation can be more comprehensive and more accurate.

The method comprises a first step (S1 in FIG. 1) of obtaining a transmission image, a second step (S2) of applying a quadratic differential filter to the transmission image, thereby to emphasize a part of large luminance change as a quadratic differential filter image, a third step (S3) of binarizing the quadratic differential filter image with a predetermined threshold value, and then storing the resulting binarized image, a fourth step (S4) of binarizing the transmission image with another predetermined threshold value, and then storing the resulting binarized image, a fifth step (S5) of performing the measurement of binary feature quantities for the binarized image stored at the third step (S3) and the binarized image stored at the fourth step (S4), and a sixth step (S6) of deciding the quality of the object to-be-inspected from the binary feature quantities.

The disclosure provides an approach for edge-aware spatio-temporal filtering. In one embodiment, a filtering application receives as input a guiding video sequence and video sequence(s) from additional channel(s). The filtering application estimates a sparse optical flow from the guiding video sequence using a novel binary feature descriptor integrated into the Coarse-to-fine PatchMatch method to compute a quasi-dense nearest neighbor field. The filtering application then performs spatial edge-aware filtering of the sparse optical flow (to obtain a dense flow) and the additional channel(s), using an efficient evaluation of the permeability filter with only two scan-line passes per iteration. Further, the filtering application performs temporal filtering of the optical flow using an infinite impulse response filter that only requires one filter state updated based on new guiding video sequence video frames. The resulting optical flow may then be used in temporal edge-aware filtering of the additional channel(s) using the nonlinear infinite impulse response filter.

The invention relates to a shielded workpiece identifying method and device based on binary system feature matching. The identifying problem of a shielded workpiece is solved, and meanwhile the computation complexity can be lowered, the storage space can be reduced, and the matching precision can be improved. In addition, good robustness is achieved on ambient light, view angle changing and partial shielding, the shielded workpiece can be identified in different interference environments, and a good identifying result is obtained. Thus, the shielded workpiece identifying method and device can be widely applied to the field of image identifying.

Embodiments of the invention disclose a system and a method for transforming a biometric of a user to a binary feature vector suitable for user authentication, comprising steps of: partitioning the biometric into a set of regions, wherein each region is a contiguous region confining a part of the biometric; determining, for each region, biometric parameters; applying, to each region, binary logical conditions to produce at least one bit of the binary feature vector, wherein the applying is performed by a processer; and outputting the binary feature vector.

The invention discloses an improved approximate projection transformation image splicing method, relating to the technical field of image processing. The image splicing method includes S1 image preprocessing, S2 image feature point extraction, and S3 image feature point matching based on an optical flow pyramid. The image processing in S1 includes noise suppression, enhancement of texture and contrast, and normalization of histograms, and further includes calibration of the camera lens, establishment of the corresponding mathematical model based on the cause of image distortion, and correction of contaminated or distorted images by using the calibrated parameters; and the image feature point extraction in S2 adopts a BRISK and/or ORB feature descriptor. The method provided is improved based on the approximate projection transformation (APAP) image splicing method, the pre-processing step is added before the feature extraction step, and the BRISK and/or ORB binary feature descriptor is creatively used in the feature extraction step, and the optical flow pyramid is creatively used in the matching step, so that the image splicing is timesaving and accurate in registering.

The invention discloses a method for detecting a pirated application program. The method comprises the following steps: acquiring a binary characteristic file of an installation package of a to-be-detected application program; analyzing the binary characteristic file of the installation package of the to-be-detected application program and the binary characteristic file of the installation package of a copyrighted application program so as to determine the similarity of the binary characteristic file of the installation package of the to-be-detected application program and the binary characteristic file of the installation package of the copyrighted application program; judging whether the installation package of the to-be-detected application program is a pirated application program installation package according to the similarity. By adopting the technical scheme of the invention, the method can be used for detecting whether the application program installation package is a pirated application program installation package even the pirated application program installation package is repackaged and changed in name.

The invention provides a graded image retrieval method based on deep features of a convolutional neural network. According to the basic principle, the method comprises the steps of firstly, training the convolutional neural network used for feature extracting, and determining network parameters; then using the trained convolutional neural network to extract image features, and obtaining multiple convolutional layer binary system features and one full-joint layer binary system feature; secondly, applying the convolutional layer binary system features to a preliminary screening retrieval stage,conducting multi-feature similarity fusion after further compressing the features, sifting out a candidate image set, and reducing the retrieval range; finally, using the full-joint layer binary system feature to accurately retrieve the candidate image set to obtain a final retrieval result. As is shown by an experiment result based on a public image retrieval dataset, compared with an existing image retrieval method, the representing mode of the images of the graded image retrieval method is more comprehensive, a feature compression method is simpler and more efficient, and the retrieval accuracy is high; meanwhile, by means of a graded retrieval mode, the system calculation amount is dispersed, parallel accelerating is achieved, and the graded image retrieval method has practical value.

The invention discloses a lightweight-class Android malicious software fast judging method. According to the method, software samples are collected from different sources, and two software bases including a malicious software base and a benign software base are respectively formed; configuration files of Android program installing files in the two sample bases are respectively obtained; the configuration files are decoded, permission vectors are extracted, and permission sets of each program are obtained; the sample base is subjected to feature extraction to obtain 46 permissions, each sample generates one corresponding 46-dimension binary feature vector; 10 feature vectors with the highest relevancy with the software property are worked out, and each sample feature is subjected to class calibration; classification models are trained for the sets of 10 feature sets; installing files of software to be tested are obtained, and the feature vectors corresponding to the program to be tested are generated; according to the classification models, the feature vectors are used as the input, the software to be tested is classified, and the detection result of the software to be tested is obtained. Compared with the prior art, the lightweight-class Android malicious software fast judging method has the characteristics of high speed and simplicity and can be used as a primary auditing tool for each Android application market.

A method comprising: obtaining a three-dimensional (3D) point cloud about an object; obtaining binary feature descriptors for feature points in a 2D image about the object; assigning a plurality of index values for each feature point as multiple bits of the corresponding binary feature descriptor; storing the binary feature descriptor in a table entry of a plurality of hash key tables of a database image; obtaining query binary feature descriptors for feature points in a query image; matching the query binary feature descriptors to the binary feature descriptors of the database image; reselecting one bit of the hash key of the matched database image; and re-indexing the feature points in the table entries of the hash key table of the database image.

The invention discloses a distributed face recognition system. The system comprises an underlying database module, a logical layer module, and a client, wherein the underlying database module comprises a face image database, an image metadata database, and a face feature database; the face image database is used for storing original face image data; the image metadata database is used for storing basic information of an original face image and relevant personal information; the face feature database is a high-speed in-memory NoSQL database and is used for storing binary feature values of the original face image; the logical layer module comprises a database proxy service module, a computing service node module, a proxy service node module, and a core service node module; and an interface for the face image database, the image metadata database and the face feature database is packaged in the database proxy service module.

The invention discloses an authentication method based on full face binary matching. The method comprises the following steps of when an authentication triggering instruction of a user is received, obtaining the face image of the user; preprocessing the face image to obtain an optimized full face grayscale image; determining the skin pore features of a plurality of pore feature points in the full face grayscale image; independently carrying out Hash processing on the skin pore feature of each pore feature point to obtain the skin pore binary feature of each pore feature point; independently carrying out matching on the skin pore binary feature of each pore feature point with the pre-obtained reference binary feature of the corresponding pore feature point in a feature library; and according to a matching result, determining whether the user is a legal user or not. When the method provided by the embodiment of the invention is applied, identification efficiency and accuracy can be improved, and system safety is enhanced. The invention also discloses an authentication device based on the full face binary matching, and the authentication device has a corresponding technical effect.

The invention provides a face recognition working method through fusion of binary features and a joint laminated structure and an intelligent chip. The method comprises the following steps: S1, acquiring face feature data in image features, and constructing a face feature data hash layer in a convolutional neural network, S2, enabling a face feature data hash function to meet the constraints of independence and minimum quantization error, and solving the face feature data hash function; S3, performing deep face feature data transformation on the solved face feature data hash function by usingcascade structure data operation; s4, through a depth binary human face feature data hash function, extracted human face feature data and human face feature data key point detection, human face feature data posture estimation and human face feature data classification are carried out; and S5, after face feature data mining, face image comparison and recognition are carried out by utilizing similarity, face feature data are identified and extracted, and required face feature data are obtained.

Systems and methods for identifying similarities in program binaries, including extracting program binary features from one or more input program binaries to generate corresponding hybrid features. The hybrid features include a reference feature, a resource feature, an abstract control flow feature, and a structural feature. Combinations of a plurality of pairs of binaries are generated from the extracted hybrid features, and a similarity score is determined for each of the pairs of binaries. A hybrid difference score is generated based on the similarity score for each of the binaries combined with input hybrid feature parameters. A likelihood of malware in the input program is identified based on the hybrid difference score.

The invention provides a scene information searching method based on binary feature codes, and belongs to the technical field of mobile augmented reality technologies. The scene information searching method specifically includes the steps of firstly, collecting an image to be recognized of a current scene, and obtaining GPS information and gravity direction information; secondly, obtaining a descriptor feature vector of the image to be recognized; thirdly, packaging the GPS information, the gravity direction information and the descriptor feature vector into a descriptor file, and sending the descriptor file to a server; fourthly, calculating the included angle between the main direction of the descriptor feature vector and the gravity direction of the descriptor feature vector through the server; fifthly, conducting mapping on the descriptor feature vector through the Hash function; sixthly, finding a corresponding GPS information chain table closest to the GPS information; seventhly, obtaining a Hash table set; eighthly, conducting filtering on the Hash table set based on the included angle; ninthly, finding the best matching sample image in the Hash table set which is filtered, and feeding the scene information corresponding to the sample image back to a mobile terminal.

The invention discloses a building extraction method based on a binary feature classifier. The building extraction method comprises the following steps: firstly, an original image is subjected to downsampling and transformed into LAB color space, then SLIC super segmentation is implemented in the LAB color space, intersections are obtained and subjected to dichotomy by the binary feature classifier, building intersections and non-building intersections are obtained, and a building area is obtained; then an extracted building area image is transformed into the LAB color space, SLIC super segmentation is implemented in the LAB color space, new intersections are obtained and subjected to trichotomy by the binary feature classifier, building interior points, building edge points and non-building points are obtained, and a building contour is obtained. The building extraction method has the remarkably effects as follows: rapid and accurate extraction of the building contour is realized; compared with the traditional method, the method is higher in precision and speed, and interaction is avoided in an algorithmic process.

The invention discloses a binary image feature similarity discrimination method based on a random forest algorithm, comprising an off-line index phase: extracting the scale invariant feature transform conversion features of an image, employing each dimension of all features as a vector to be used for clustering through a K mean value method, and quantifying scale invariant feature transform conversion features into 512 dimension binary system features; and writing quantified features, feature indexing, image names corresponding to features and adjacent features into a database to be used as an image feature database. In the off-line index phase, the scale invariant feature transform conversion features of the image are extracted, and are quantified to be 512 dimension binary system features to match features in the image feature database, and find adjacent features; and then a random forest algorithm is employed to discriminate the adjacent features, and a voting mechanism retrieves similar images.

The invention discloses a protein phosphorylation site identification method, system and device and a storage medium. The method comprises the following steps: acquiring an amino acid sequence fragment of a protein phosphorylation site to be identified; performing logic operation on binary codes of amino acids in the amino acid sequence fragment to obtain logic binary feature vectors of the aminoacid sequence fragment; performing kernel principal component analysis on the logic binary feature vector according to a preset kernel function to obtain a kernel principal component logic binary feature vector; and inputting the kernel principal component logic binary feature vector into a random forest model for processing to obtain an identification result of the protein phosphorylation site. Based on theoretical calculation of the random forest model, the method can achieve the quick and accurate recognition of the information of a large number of protein phosphorylation sites, is low in cost, facilitates the development of phosphorylation mechanism and phosphorylation and disease relation research, and is widely applied to the field of protein phosphorylation site recognition.

The invention discloses a digital image tamper detection method based on adaptive feature points, which comprises the steps of firstly extracting SURF feature points through an adaptive threshold selection algorithm, and enabling the image feature points to be uniformly distributed by using a distributed processing algorithm at the same time; secondly, extracting binary feature descriptors of allfeature points by using a BRISK method; then performing fast feature matching by using a distinguishable embedded random fern algorithm; and finally, eliminating incorrect matched pairs through an RANSAC algorithm, marking matched areas by further using a fast NNPROD algorithm and a morphological method. Experimental results show that the method disclosed by the invention not only can effectivelyimprove the smooth tamper area detecting performance, but also has invariance for the post-processing operations such as JPEG compression, rotation and scaling, and has high detection precision and low time complexity.

The invention discloses a star pattern identification method of a hybrid redundancy feature. The star pattern identification method comprises the following steps of: based on the geometric position relation of the navigation star points, establishing a hybrid redundancy pattern feature of a navigation reference star point, and then, creating a binary feature database of the navigation reference star point; based on the geometric position relation of a star map and the star points, calculating the hybrid feature of the star map reference star point; according to the hybrid feature of the star map reference star point, selecting candidate star points corresponding to the star map reference star point from the binary feature database by using similar score measurement; by a maximum matching pair algorithm, selecting a matched star point for the star map reference star point from all the candidate star points. The invention also discloses a star pattern identification device of the hybrid redundancy feature. With the method and the device, high accurate identification rate can be realized under the condition of high star point position noise and star magnitude noise, moreover, acquisition of the matched star point is guaranteed when few accompanying star points exist in the radius of the navigation star point pattern.

The invention provides a method of extracting picture features by using a binary bottleneck neural network and belongs to the technical field of video processing. By building a binary bottleneck neural network, a picture is automatically extracted as a feature vector including multiple binary bits. When comparing two images for obtaining the similarity degree thereof, only the binary feature vectors of the two pictures need to be compared and then the Hamming distance between the two binary feature vectors is calculated; the smaller the Hamming distance, the higher the similarity degree of the two pictures. The method solves the technical problem in extracting binary feature vectors of images. The method calculates feature binary sequences of images and good performance can be obtained without manual design based on experience of researchers; image feature binary sequences calculated by the method can be used for calculating the similarity of images rapidly, so that the method is of great importance for similarity retrieval of pictures and video.

The present invention provides a feature extraction method for face recognition, which adopts the method of synchronous local binary feature learning and coding (SLBFLE) to extract a pixel difference vector of a face image, sets a mapping matrix W to convert the pixel difference vector into a low dimension local binary feature matrix B, converts a dictionary matrix D into a feature histogram and optimizes the mapping matrix W, the dictionary matrix D and a coefficient matrix A through an iterative method to obtain the feature histogram. The method can better prevent the influence of light variation and expression variation, improve the accuracy of face recognition, and have better anti-interference than that of the LBP method. The method is more suitable for the places with complex environments, and can be better applied to the classroom face check-in system feature extraction.