40results about How to "Accurate face recognition" patented technology

Embodiments of the invention disclose a similarity calculation method and a similarity calculation system in face recognition. The method comprises the following steps: acquiring a target image of a target face and N to-be-compared images of a first to-be-compared face; based on the target image, the N to-be-compared images and a pre-stored face feature extraction model, converting the target image into a target feature vector, and converting the N to-be-compared images into N to-be-compared feature vectors; calculating the distance measure of the target feature vector and a to-be-compared feature vector set, wherein the to-be-compared feature vector set is composed of the N to-be-compared feature vectors; and determining the similarity between the target face and the first to-be-compared face based on the pre-stored mapping relationship between distance measure and similarity and the calculated distance measure. The embodiments of the invention further disclose a corresponding similarity calculation system in face recognition. By using the technical scheme provided by the embodiments of the invention, the accuracy of face recognition can be improved.

The present invention provides a face recognition method and system, an electronic device and a computer program product, and is applied to the technical field of face recognition. The method comprises the steps of: obtaining face images with a plurality of shooting angles of the same user; performing blur filtering of the face images; performing attitude filtering of the face images after blur filtering; and performing face recognition of the face images after attitude filtering. According to the invention, after the face images with shooting angles of the same user are obtained, the face images are subjected to blur filtering and attitude filtering, the face images after filtering are subjected to face recognition, and therefore, the quality of face recognition images can be effectivelycontrolled, and rapid and accurate face recognition in a natural dynamic scene can be achieved.

The invention provides a face identification method applied to an adaptive drive seat; the method comprises the following steps: loading a face identity characteristic head portrait database; building a face identification model, and training a loaded face identification database; obtaining a video image from a camera; using a cascade classifier to detect whether the video image contains face characteristic information or not; if yes, extracting the face portion, and forming a face image; carrying out dimension normalization for the extracted characteristic head portrait, and carrying out histogram equalization treatment; using two dimension discrete rapid Fourier transform to convert the face image from a space domain to a frequency domain, and extracting characteristics; comparing the extracted characteristics in the face identification database, if the similarity is higher than a preset threshold, a predicted ID label is outputted so as to confirm the passenger ID, thus starting the adaptive drive seat; asking to input the face identity if the similarity is smaller than the preset threshold. The face identification method is applied to an unmanned vehicle auxiliary driving system, and matched with the adaptive drive seat, thus fast and accurately carrying out face identification with high efficiency.

The invention relates to a method and a device of face recognition. The method comprises the following steps: acquiring a target image on which face recognition needs to be carried out; carrying out multi-layer convolution processing on the target image; combining a preset number of pooling results of convolution feature mapping in a multi-layer convolution processing to obtain a final external processed image; classifying the final external processed image; carrying out bbox regression processing on the final external processed image; and calculating and obtaining a face position of the target image according to a classification processing result and a bbox regression processing result. Multi-layer convolution processing is carried out on the initial target image, and image features after convolution processing of different levels are combined to obtain the final external processed image. The last final external processed image generated by the method contains image detail information after intermediate convolution processing, multi-scale feature fusion is formed, description for the original target image is more accurate, and thus face recognition and position determination are also more accurate.

The invention discloses a two-dimensional code and image fusion method used for counterfeiting prevention. The method includes the steps of obtaining an image and a two-dimensional code which are to be fused; dividing the image into a plurality of image blocks according to the original information point number of the two-dimensional code, the image blocks and the original information points of thetwo-dimensional code establishing a corresponding relation; performing gradual change processing on a central sub-region of the image according to attributes of the original information points of thetwo-dimensional code; and a non-central region adopts color of corresponding pixels of the image, thereby obtaining the two-dimensional code fused with the image. The two-dimensional code and image fusion method provided by the invention enables other image blocks except the central sub-region of each information point in the two-dimensional code to clearly display information of the image, the central sub-region is subjected to gradual change processing according to the attributes of the two-dimensional code original information points, image information is not completely shielded, the display range of the image itself is expanded, the area of the image that is covered is reduced, the image information is relatively easy to capture, face recognition is relatively accurate, and anti-counterfeiting processing of fields with relatively high security is facilitated.

The invention discloses a human face recognition system based on dynamic processing of an ARM (advanced RISC machines) processing platform and equipment thereof. The human face recognition system comprises a hardware platform, an image acquisition module, an image preprocessing module, a human face detection module, a feature extracting module, a human face recognition module, a database module and a video display module. The human face recognition system is characterized in that firstly, the corresponding hardware platform is selected, and the dynamic human face recognition system is designed; the human face recognition system is used for acquiring the image information through the image acquisition module, and the collected image is preprocessed by the image preprocessing module; the human face is detected by the human face detection module, and a human face image is captured; finally, the human face is recognized by the human face recognition module according to the features extracted by the feature extracting module, and the acquired and recognized information is stored into the database module; the human face database can be conveniently managed in the system by a user, and the database management function is researched and developed by an MySQL database in the system.

A face identification method under a shielding condition based on deep learning is disclosed. In reality, face image information collected in real time is interfered by a plurality factors of shielding of body hairs, a gauze mask, a hat, sunglasses and other frequent-wearing objects and other factors so that complexity of a face identification technology is greatly increased and realization difficulty is large. The method comprises the following steps of (1) carrying out face detection and characteristic point detection on a given partially-shielded face image; (2) according to detected positions of characteristic points on a face, carrying out partial five-sense-organ image capture; and (3) based on a ubuntu16.04 operation system, under GPU1080, constructing a CAFFE deep learning framework and acquiring a face identification model. The invention discloses the face identification method under the shielding condition based on the deep learning.

The invention discloses a human face recognition method based on the bi-directionally and two-dimensionally iterative and non-relevant discriminant analysis. Firstly, the image sample of a human face is acquired. After that, within-class scatter matrixes, inter-class scatter matrixes and overall scatter matrixes in the horizontal and vertical directions are respectively calculated. By utilizing the non-relevant discriminant transformation method, a first optimal differential vector and an optimal discriminant vector set in each direction are respectively calculated. Afterwards, the projection and dimensionality reduction of the image is conduced by the two optimal discriminant vector sets at the same time. Finally, the classified calculation is conducted by a nearest neighbor classifier to figure out the recognition rate of the image. According to the technical scheme of the invention, the non-relevant discriminant transformation on two-dimensional images is conducted based on the discrimination information in two directions. In this way, the dimensionality reduction and the discrimination information extraction of two-dimensional images both in the horizontal direction and in the vertical direction can be realized at the same time. Therefore, the correct facial recognition is realized. Both the processing time and the storage space are saved, and the recognition rate is ensured to be high.

The invention is suitable for the technical field of intelligent robots, and provides a deblurred face recognition method and system and an inspection robot, and the method comprises the steps: obtaining a video stream from a camera of the inspection robot, carrying out the video decoding of the video stream, and obtaining a decoded video image; performing face detection on the video image by adopting a face detection algorithm to obtain an area image containing a face ROI area in the video image; carrying out ambiguity judgment on the regional image; if the regional image is not the blurred image, performing face recognition on the video image to obtain a face recognition result; if the mode region image is a blurred image, extracting a plurality of continuous frames of images before andafter the video image; inputting the multiple frames of images into a deblurred network to obtain a deblurred image, the deblurred network being a deep learning network obtained by pre-training a plurality of blurred image samples and corresponding clear image samples as training samples; and performing face recognition on the deblurred image to obtain a face recognition result.

The invention provides a face recognition method based on improved mobile NetV3. The face recognition method comprises the following steps of S1, performing face detection and feature point detectionon an image; S2, performing face alignment by utilizing the detected face feature points; S3, making the aligned images and labels into a training set; and S4, building a network structure, changing the final pooling layer of the network into a convolution layer, and reducing the expansion multiple of the channel. According to the face recognition method based on the improved mobile NetV3, the recognition rate can be improved through the improved mobile NetV3 method in entrance and exit and video monitoring application, and accurate and rapid face recognition is achieved.

The present invention discloses a face identification method. The method comprises the following steps: receiving the face image of a user to be identified, and extracting a characteristic vector from the face image; matching the characteristic vector and at least two pre-stored compression vectors, and calculating the similarity of the characteristic vector and each compression vector, wherein the compression vector is obtained by compressing at least two corresponding characteristic vectors extracted from at least two face images of the same user; obtaining the maximum value in the at least two calculated similarities; and outputting the user information corresponding to the compression vector corresponding to the maximum value. The invention further discloses a face identification device. The face identification method and device are able to improve the speed and the accuracy of face identification.

The present invention discloses a method for automatically replacing a reference photograph according to similarity concentration ratios. The method includes the following steps that: (1) a database is constructed in a system, a photograph archiving module, a photograph comparison module and a photograph replacement module are set in the database; (2) when a relevant person logins the system for the first time, the system takes a picture of the relevant person, wherein the picture is adopted as a reference photograph; (3) when the relevant person starts a process of multiple times of login; when the relevant person logins the system each time, the system identifies the face of the relevant person, an identified image is compared with the reference photograph, if the similarity of the identified image and the reference photograph reaches a specified value, the relevant person logins and enter the system; and (4) when the number of successful logins of the relevant person in the step 3 reaches a specified value of the system, images stored in the photograph comparison module are automatically compared, and a photograph with the highest similarity concentration ratio in all the images is identified and is adopted as a new reference photograph. With the method of the invention adopted, face recognition can be carried out more quickly, more accurately and conveniently.

The invention relates to a face recognition method, a face recognition device, a storage medium and electronic equipment, and belongs to the technical field of face recognition. The method comprises the following steps: carrying out the key point detection when a first face image is received; obtaining an recognition score of each detectable key point and a number of a missing key point; obtainingthe influence score of the shielded area of part of the face features on the recognized face; when the influence score is higher than a predetermined score threshold, obtaining a plurality of targetkey points having a predetermined face feature association relationship with the missing key points in the plurality of detectable key points; obtaining a target face feature template of which the position combination degree with the plurality of target key points is greater than a preset combination degree threshold; and stitching the target face feature template with the first face image to obtain a second face image, and performing face recognition after detecting all key points according to the second face image. According to the invention, when the human face features are missing, the features are accurately supplemented, so that the human face recognition can be efficiently and accurately carried out.

The invention relates to a face image recognition method and device, electronic equipment and a storage medium, and the method comprises the steps: obtaining a plurality of face images; according to aplurality of feature vectors obtained by performing feature extraction on the plurality of face images correspondingly, obtaining a plurality of to-be-identified target objects; obtaining a gradientparameter according to the plurality of feature vectors and a classification reference vector; and classifying the plurality of to-be-identified target objects according to the gradient parameters toobtain a classification result. According to the embodiment of the invention, the face recognition effect can be improved.

The invention provides an intelligent face recognition method and system suitable for a facial paralysis patient, and the method comprises the steps: obtaining a transverse corner image of the facial paralysis patient, and enabling the transverse corner image to be the same as the front face image of the patient in size, wherein each pixel value in the transverse rotation angle image represents the transverse rotation angle of the face of the patient when the pixel is located at the boundary of the side face; obtaining an optimal face deflection angle according to the condition information of the patient and the transverse rotation angle image, enabling the patient to transversely rotate at the optimal face deflection angle to obtain a side face image of the patient, and shielding the lesion position of the patient in the side face image of the patient; and correcting the side face image of the patient to obtain a healthy face image of the patient, and performing face recognition on the patient according to the healthy face image of the patient. According to the face recognition method, a face recognition technical scheme for the facial paralysis patient is provided, the influence of the lesion position of the facial paralysis patient on face recognition is eliminated, and face recognition can be accurately carried out on the facial paralysis patient.

The invention discloses a face recognition method for an advertisement screen based on autonomous learning, and the method comprises the steps: carrying out the secondary combination: carrying out thelearning of sample feature values accumulated in the period time, carrying out the secondary screening of the feature values in a same face angle interval, and carrying out the combination of the feature values of the same face ID; and judging the similarity of the human face features by adopting a human face recognition algorithm, and if the similarity is greater than a set value, merging the recognized human face features below the same human face ID. Accurate data statistics is realized through an autonomous learning method, so that more accurate audience analysis is provided for advertisement putting on a display screen. According to the method, the recognition rate is improved through autonomous learning, the recognition recall rate can be continuously improved through learning of the sample characteristic values, more accurate face recognition can improve the understanding of an advertiser on audiences, personalized advertisement putting is more accurate, and meanwhile more reliable data support is provided for calculating the putting effect of advertisements.

The invention relates to a face recognition method based on block collaborative representation. In practical application, the information loss (such as pixel information loss, corrosion block and shielding) of the face image can influence the pixel value of the image so as to influence the face recognition effect. The method comprises the following steps: firstly, dividing images in a face image library into a training set and a test set, searching an optimal blocking mode to block the images, and constructing a sub-block dictionary for the images after the training samples are blocked; Secondly, separating an error generated by information loss of the human face image from the original image by adopting cooperative representation (CR) of sub-blocks, so that the influence of a human face recognition result of the missing information image can be effectively reduced; Then classifying the sparse sparsity obtained by each sub-block; And finally, obtaining a final identification result based on a maximum voting criterion, thereby effectively reducing the influence of invalid classification on the identification result of the whole image caused by a large minimum coefficient error obtained by one or more sub-block characteristics, and improving the identification rate.

The invention provides a face recognition device capable of automatically adjusting a visual angle. The device comprises an induction module, a distance measurement module, a motor module, a camera module, and a main control module, and the induction module, the distance measurement module, the motor module, and the camera module are respectively connected with the main control module. The sensingmodule is used for sensing a human body; the camera module comprises a lens, and the distance measuring module is used for measuring the distance between a human body and the lens; the motor module is connected with the camera module and is used for controlling the movement or rotation of the lens; and the main control module is used for collecting a face image collected by the lens so as to realize face recognition. According to the face recognition device capable of automatically adjusting the visual angle, face image collection and face recognition are automatically completed by automatically controlling movement or rotation of the lens, so that lens installation is not limited by the angle and the position, use is very convenient, and the user experience is improved.

The invention relates to a face recognition device with strong adaptability for vehicle access. The device includes a housing, a light supplement assembly and two close mechanisms. The light supplement assembly comprises a rotating shaft, a switch component, a complementary light module, two support assemblies, two lampshades and a number of beads. Each support assembly includes a vertical bar anda cannula. The light supplement assembly includes a lamp tube, a power supply and a regulation unit. The switching assembly includes gears, a rack, a translation unit and an upright plate. Each closure mechanism comprises a fixing plate, a hinge assembly, a baffle, a first motor, a bearing, a first drive shaft, moving blocks and struts. The face recognition device with strong adaptability for vehicle access facilitates the camera to take clear images in a dim environment through a light compensation mechanism, and can easily carry out accurate face recognition. Moreover, when the light is strong, the baffle can block the opening through the closing mechanism, the camera is prevented from being damaged, the safe operation of the equipment is protected, and the practicability of the equipment is improved.

The invention discloses a social intimacy determination method and system. The method comprises the steps of obtaining an image collected by a camera; Detecting face information in the image by adopting a cascade neural network; Determining sparse representation of the detected face information by adopting a convolutional neural network; adopting The sparse representation of the face information as input, and adopting a classifier to classify faces to obtain faces representing different persons; Taking faces of different persons as targets for tracking, and obtaining behavior information of each person, the behavior information comprising distance information between persons and conversation behaviors between persons; And determining the intimacy between the walking persons according to the behavior information of each walking person. According to the social intimacy determination method and system provided by the invention, the intimacy between the pedestrians can be automatically obtained based on camera monitoring.

The invention discloses an access control system based on video photography. The access control system comprises a camera module used for acquiring video pictures. The access control system effectively fuses face recognition with artificial composition, can automatically open an access for a person after a face of the person in a video region is recognized and judges whether to open the access forthe person according to a monitoring picture transmitted in real time, thereby being more accurate compared with the face recognition; besides, the effect that a user acquires conditions of a deviceand updates own information can be realized, so that the access control system is anti-interference, effect of pattern recognition is better, face matching speed is increased, consumption of a networkbandwidth resource is reduced, network service quality can still be improved even if quantity of monitoring devices is increased or decreased, and real-time monitoring demand is met.

The invention discloses an unlocking method, device and equipment based on face recognition and a computer storage medium, and the method comprises the steps of obtaining video information sent by preset collection equipment, and the video information comprises standard video information and to-be-recognized video information; when the standard video information comprises a standard face image, processing the standard face image to obtain standard texture features of the standard face image, and updating and storing the standard texture features in a preset database; when the to-be-recognizedvideo information comprises a to-be-recognized face image, obtaining to-be-recognized texture features of the to-be-recognized face image; and comparing the to-be-identified texture feature with eachpre-stored standard texture feature in a preset database, and when a target standard texture feature matched with the to-be-identified texture feature exists in the preset database, executing an unlocking instruction. According to the invention, the accuracy of unlocking based on face recognition is improved.

The invention discloses a depression angle face recognition method and system based on an axial attention weight distribution network. First, CNN is used to extract features from face image sequences of different angles and resolutions to generate a set of normalized feature vectors; Secondly, the cascaded axial attention weight assignment module is used to adaptively assign weights to the image feature sequence, and obtain the weight vector (horizontal) of each feature map itself and the weight vector (vertical) between features to form an axial weight matrix; The weighted aggregation is carried out by using the axial weight to obtain a more discriminative feature vector for identification. The axial matrix weight of the present invention can more delicately represent the weight distribution result between the face sequence images, so the fused features have stronger discrimination, which is beneficial to more accurate depression angle face recognition.

The invention provides a face feature extraction method and device, equipment and a storage medium. The image processing method comprises the following steps: acquiring a to-be-processed image; inputting the to-be-processed image into the face enhancement model, and performing face enhancement processing on the to-be-processed image through the face enhancement model to obtain an enhanced image, wherein the image quality of the enhanced image is higher than that of the to-be-processed image; and inputting the enhanced image into the face feature extraction model, and performing face feature extraction on the enhanced image through the face feature extraction model to obtain a target feature image. According to the method, the image quality can be improved to a great extent, the image information is increased, when the method is applied to the field of face recognition, due to the fact that the image quality is high, the accuracy of face recognition can be effectively improved, and the method can be better suitable for being applied to face recognition in environments and scenes.

The invention discloses a face recognition and analysis method used in monitoring video data, which uses two neural networks to compare and score the high-latitude face features of the face, and integrates the K-adjacent algorithm and the screening mechanism with the Euclidean distance as the threshold to achieve Face comparison and recognition, so as to obtain accurate face recognition results; in addition, the present invention also performs further screening according to the time of video data and the frequency of appearance of faces, finally obtains the recognition results, and then obtains more accurate face recognition result. The invention can perform relatively accurate face recognition on video data or picture data captured by real-time or historical monitoring cameras.

The invention discloses a method and system for determining social intimacy. The method includes: acquiring an image collected by a camera; using a cascaded neural network to detect face information in the image; using a convolutional neural network to determine a sparse representation of the detected face information; using the sparse representation of the face information as an input , use the classifier to classify the faces, and get the faces representing different actors; use the faces of different actors as targets to track, and obtain the behavior information of each actor, the behavior information includes the behavior between the actors Distance information and conversation behavior between actors; determine the intimacy between actors according to the behavior information of each actor. The method and system for determining social intimacy provided by the present invention can automatically acquire the intimacy between actors based on camera monitoring.

The invention discloses a human face recognition method based on the bi-directionally and two-dimensionally iterative and non-relevant discriminant analysis. Firstly, the image sample of a human face is acquired. After that, within-class scatter matrixes, inter-class scatter matrixes and overall scatter matrixes in the horizontal and vertical directions are respectively calculated. By utilizing the non-relevant discriminant transformation method, a first optimal differential vector and an optimal discriminant vector set in each direction are respectively calculated. Afterwards, the projection and dimensionality reduction of the image is conduced by the two optimal discriminant vector sets at the same time. Finally, the classified calculation is conducted by a nearest neighbor classifier to figure out the recognition rate of the image. According to the technical scheme of the invention, the non-relevant discriminant transformation on two-dimensional images is conducted based on the discrimination information in two directions. In this way, the dimensionality reduction and the discrimination information extraction of two-dimensional images both in the horizontal direction and in the vertical direction can be realized at the same time. Therefore, the correct facial recognition is realized. Both the processing time and the storage space are saved, and the recognition rate is ensured to be high.

The invention discloses a sitting posture recognition reminding method and device. The standard sitting posture of a recognized person is firstly input through a depth module and a Sensor lens, the feature value of the sitting posture and other information of a corresponding person are automatically stored after the inputting is completed, face recognition recording is carried out through the lens, and the face features shot by the lens are compared with inputted feature information one by one, so that cost is reduced, accurate sitting posture recognition is realized, accuracy of output recognition results is guaranteed, users are prompted to correct sitting posture conditions, and health problems of people with high myopia rate, cervical vertebra and the like in the current society are well solved.

The invention discloses a face recognition method, device, device and storage medium based on radar technology. The method detects the face through two wireless detection modules and generates two corresponding depth maps, and according to the two depth maps Get the shooting angle of the camera to the face; according to the shooting angle, get the corresponding feature distance ratio correction coefficient, and calibrate the face recognition process of the face image captured by the camera according to the feature distance ratio correction coefficient. The present invention can judge whether the face to be recognized is a real face, can effectively avoid the phenomenon of using an image face to pretend to be a real face for face recognition, and can improve the safety and accuracy of face recognition.