WO2020114135A1 - 特征识别的方法及装置 - Google Patents
特征识别的方法及装置 Download PDFInfo
- Publication number
- WO2020114135A1 WO2020114135A1 PCT/CN2019/112909 CN2019112909W WO2020114135A1 WO 2020114135 A1 WO2020114135 A1 WO 2020114135A1 CN 2019112909 W CN2019112909 W CN 2019112909W WO 2020114135 A1 WO2020114135 A1 WO 2020114135A1
- Authority
- WO
- WIPO (PCT)
- Prior art keywords
- target object
- image
- feature
- model
- machine learning
- Prior art date
Links
- 238000000034 method Methods 0.000 title claims abstract description 55
- 238000010801 machine learning Methods 0.000 claims abstract description 60
- 238000012545 processing Methods 0.000 claims abstract description 25
- 238000012549 training Methods 0.000 claims description 30
- 230000015654 memory Effects 0.000 claims description 18
- 238000004590 computer program Methods 0.000 claims description 11
- 230000001815 facial effect Effects 0.000 claims description 8
- 230000005021 gait Effects 0.000 claims description 4
- 238000001914 filtration Methods 0.000 abstract description 5
- 238000010586 diagram Methods 0.000 description 12
- 238000005516 engineering process Methods 0.000 description 8
- 230000006870 function Effects 0.000 description 8
- 238000013473 artificial intelligence Methods 0.000 description 6
- 230000005540 biological transmission Effects 0.000 description 6
- 238000013528 artificial neural network Methods 0.000 description 5
- 230000008569 process Effects 0.000 description 5
- 238000011161 development Methods 0.000 description 3
- 210000003128 head Anatomy 0.000 description 3
- 238000003909 pattern recognition Methods 0.000 description 3
- 238000006243 chemical reaction Methods 0.000 description 2
- 238000004891 communication Methods 0.000 description 2
- 238000013434 data augmentation Methods 0.000 description 2
- 238000013508 migration Methods 0.000 description 2
- 230000005012 migration Effects 0.000 description 2
- 238000012986 modification Methods 0.000 description 2
- 230000004048 modification Effects 0.000 description 2
- 238000005457 optimization Methods 0.000 description 2
- 230000036544 posture Effects 0.000 description 2
- 201000004569 Blindness Diseases 0.000 description 1
- 230000003321 amplification Effects 0.000 description 1
- 238000004458 analytical method Methods 0.000 description 1
- 230000003190 augmentative effect Effects 0.000 description 1
- 230000009286 beneficial effect Effects 0.000 description 1
- 230000008901 benefit Effects 0.000 description 1
- 238000012937 correction Methods 0.000 description 1
- 230000007547 defect Effects 0.000 description 1
- 238000001514 detection method Methods 0.000 description 1
- 210000000887 face Anatomy 0.000 description 1
- 230000008713 feedback mechanism Effects 0.000 description 1
- 230000006872 improvement Effects 0.000 description 1
- 238000007689 inspection Methods 0.000 description 1
- 238000010295 mobile communication Methods 0.000 description 1
- 238000003199 nucleic acid amplification method Methods 0.000 description 1
- 230000003287 optical effect Effects 0.000 description 1
- 230000003252 repetitive effect Effects 0.000 description 1
- 238000012360 testing method Methods 0.000 description 1
- 238000012795 verification Methods 0.000 description 1
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/21—Design or setup of recognition systems or techniques; Extraction of features in feature space; Blind source separation
- G06F18/214—Generating training patterns; Bootstrap methods, e.g. bagging or boosting
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V30/00—Character recognition; Recognising digital ink; Document-oriented image-based pattern recognition
- G06V30/10—Character recognition
- G06V30/19—Recognition using electronic means
- G06V30/192—Recognition using electronic means using simultaneous comparisons or correlations of the image signals with a plurality of references
- G06V30/194—References adjustable by an adaptive method, e.g. learning
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V40/00—Recognition of biometric, human-related or animal-related patterns in image or video data
- G06V40/10—Human or animal bodies, e.g. vehicle occupants or pedestrians; Body parts, e.g. hands
- G06V40/18—Eye characteristics, e.g. of the iris
- G06V40/197—Matching; Classification
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V20/00—Scenes; Scene-specific elements
- G06V20/60—Type of objects
- G06V20/62—Text, e.g. of license plates, overlay texts or captions on TV images
- G06V20/625—License plates
Definitions
- This application relates to, but not limited to, the computer field, and in particular, to a method and device for feature recognition.
- the method to solve these problems is data augmentation technology, that is, based on relevant technical data, according to a certain rule or method, artificially increase data diversity.
- new image data can be generated by rotating, adjusting contrast, binarizing, etc.
- these methods often only pursue quantity, with a certain blindness, and cannot truly reflect the invariance of the data probability distribution.
- the models trained with these augmented data will have defects such as mean deviation or unstable variance when predicting results, which will affect the final prediction accuracy.
- FIG 1 is a training schematic diagram of an artificial intelligence model according to related technologies.
- the development model of related technologies only has two links of training and application, which are mostly one-time development or regular upgrades.
- the model cannot be used normally.
- the early products of automatic license plate recognition were incompatible for a long period of time. Unable to quickly obtain new samples and new models, so that such artificial intelligence products cannot quickly solve new problems.
- the embodiments of the present application provide a method and a device for feature recognition, so as to at least solve the problem of low accuracy when a feature recognition model recognizes various forms of features of a target object in the related art.
- a method for feature recognition includes: acquiring a first feature of a first target object; using a machine learning model to acquire a first feature that has an association relationship with the first feature of the first target object A second feature of a target object, wherein the machine learning model is a model obtained by training the original model using the first sample information as input information of the original model, and the first sample information includes the first rule and Multiple sets of first characteristics of the second target object, wherein the first rule is a rule for acquiring the second characteristics of the second target object according to the first characteristics of the second target object; the first The second feature of the target object is compared with the third feature, and when the two match, the third feature is mapped to the first target object.
- a method for feature recognition comprising: acquiring a fourth feature of a first type; using a machine learning model to acquire a second feature of a second type having an association relationship with the fourth feature Five features, wherein the machine learning model is a model obtained by training the original model using second sample information as input information of the original model, and the second sample information includes second rules and multiple sets of first-type features , Where the second rule is a rule for obtaining a second type feature based on the multiple sets of first type features; comparing the fifth feature with the first target object second type feature, in the two When they match, the fourth feature is mapped to the first target object.
- an apparatus for feature recognition including: a first acquisition module for acquiring first features of a first target object; and a second acquisition module for acquiring using a machine learning model A second feature of the first target object having an association relationship with the first feature of the first target object, wherein the machine learning model is obtained by training the original model using the first sample information as input information of the original model Model, the first sample information includes a first rule and first characteristics of multiple sets of second target objects, wherein the first rule is used to obtain the first rule based on the first characteristics of the second target object The rule of the second feature of the two target objects; the matching module is used to compare the second feature of the first target object with the third feature, and when the two match, the third feature is mapped to the The first target audience.
- a storage medium in which a computer program is stored, wherein the computer program is set to execute the steps in any one of the above method embodiments during runtime.
- an electronic device including a memory and a processor, the memory stores a computer program, the processor is configured to run the computer program to perform any of the above The steps in the method embodiment.
- the second feature may be a machine learning model
- the first feature is obtained after processing, such as image rotation, resolution enhancement, voice filtering, etc., and then compared with the currently acquired third feature according to the converted second feature, if the two are similar or the same, it means The third feature also describes the first target object.
- the feature recognition model is based on the virtual feature ratio For the actual third feature, it solves the problem of low accuracy when the feature recognition model in the related art recognizes various forms of features of the target object.
- Fig. 1 is a training schematic diagram of an artificial intelligence model according to related technologies
- FIG. 2 is a block diagram of a hardware structure of a computer terminal of a method for feature recognition according to an embodiment of the present application
- FIG. 3 is a flowchart of a method of feature recognition according to an embodiment of the present application.
- FIG. 5 is a schematic diagram of adjusting the license plate according to the first embodiment of the specific embodiment
- FIG. 6 is a schematic diagram of AI training of scheme three according to a specific embodiment
- FIG. 7 is a schematic diagram of the AI recognition image of scheme 3 according to a specific embodiment.
- the embodiments described in this application can be applied to urban security and other fields, such as obtaining images of multiple postures and multiple angles of the target person according to the target person’s ID photo, and then matching with the images taken by the camera on the street to identify Target person.
- FIG. 2 is a block diagram of a hardware structure of a computer terminal of a method for feature recognition according to an embodiment of the present application.
- the computer terminal 20 may include one or more (in FIG. 2 Only one is shown) a processor 202 (the processor 202 may include but is not limited to a processing device such as a microprocessor MCU or programmable logic device FPGA) and a memory 204 for storing data, optionally, the computer terminal described above may also It includes a transmission device 206 for communication functions and an input and output device 208.
- FIG. 2 is merely an illustration, which does not limit the structure of the computer terminal described above.
- the computer terminal 20 may also include more or fewer components than those shown in FIG. 2 or have a configuration different from that shown in FIG. 2.
- the memory 204 may be used to store software programs and modules of application software, such as program instructions/modules corresponding to the method of feature recognition in the embodiments of the present application, and the processor 202 executes each program by running the software programs and modules stored in the memory 204 Various functional applications and data processing, that is, to achieve the above method.
- the memory 204 may include a high-speed random access memory, and may also include a non-volatile memory, such as one or more magnetic storage devices, flash memory, or other non-volatile solid-state memory.
- the memory 204 may further include memories remotely provided with respect to the processor 202, and these remote memories may be connected to the computer terminal 20 through a network. Examples of the above network include but are not limited to the Internet, intranet, local area network, mobile communication network, and combinations thereof.
- the transmission device 206 is used to receive or send data via a network.
- the specific example of the network described above may include a wireless network provided by a communication provider of the computer terminal 20.
- the transmission device 206 includes a network adapter (Network Interface Controller (NIC), which can be connected to other network devices through the base station to communicate with the Internet.
- the transmission device 206 may be a radio frequency (Radio Frequency, RF) module, which is used to communicate with the Internet in a wireless manner.
- RF Radio Frequency
- FIG. 3 is a flowchart of a method for feature recognition according to an embodiment of the present application. As shown in FIG. 3, the process includes the following steps:
- Step S302 Acquire the first feature of the first target object
- the first target object may be a target person, and the first feature may be an image of the target person, voice information, or the like.
- Step S304 Use a machine learning model to obtain a second feature of the first target object that has an association relationship with the first feature of the first target object, where the machine learning model uses the first sample information as an input of the original model Information on the model obtained by training the original model, the first sample information includes a first rule and a plurality of sets of second target object first features, wherein the first rule is used according to the second The rule of acquiring the second characteristic of the second target object by the first characteristic of the target object;
- the second feature of the first target object may be an image converted by a machine learning model, such as a photo of a person simulating a head down, a photo of a head up, a side face photo, etc.; if the feature is a target object’s
- a machine learning model such as a photo of a person simulating a head down, a photo of a head up, a side face photo, etc.
- the feature is a target object’s
- voice information you can use the machine learning model for voice amplification and voice filtering.
- the characteristics may also be characteristics of the target person, such as sports posture, etc.
- the original model can be pre-established by the staff based on experience.
- the original model can be divided into multiple layers. Each layer is used for different solutions. Taking the feature as image information as an example, the first layer can be used to identify the eyes in the front image. According to the first rule, the eyes of the side face are converted; the second layer can be used to identify the hair in the front image, and the hairstyle of the side face is converted according to the partial rules of the first rule. The recognition results are combined as the output of the original model.
- step S306 the second feature of the first target object is compared with the third feature, and when the two match, the third feature is mapped to the first target object.
- step S306 the image recognition method in the related art may be adopted. Since the basic material library of the image recognition model is expanded in the previous step S304, the image recognition model in step S306 may use photos of different angles of the target person, instead of being limited to use Take a head-on comparison with the street camera.
- the second feature may be a machine learning model
- the first feature is obtained after processing, such as image rotation, resolution enhancement, voice filtering, etc., and then compared with the currently acquired third feature according to the converted second feature, if the two are similar or the same, it means The third feature also describes the first target object.
- the feature recognition model is based on the virtual feature ratio For the actual third feature, it solves the problem of low accuracy when the feature recognition model in the related art recognizes various forms of features of the target object.
- the feature includes one of the following: an image of the target object, and voice information of the target object.
- the method includes: using a camera to acquire a first image of the first target object; using a machine learning model to acquire an association relationship with the first image of the first target object
- the second image of the first target object is compared with the third image, and when the two match, the third image is mapped to the first target object.
- a first image of a first target object is acquired using a camera
- a second image of a first target object associated with the first image of the first target object is acquired using a machine learning model
- the second image may be a machine
- the learning model acquires the first image after some image processing, such as image rotation, resolution enhancement, etc., and then compares the converted second image with the currently acquired third image. If the two are similar, explain The third image also describes the first target object.
- using machine learning models and big data to obtain images of various poses of the first target object in time, providing a large amount of basic materials for image recognition.
- the image recognition model compares the third image with the second image of the first target object to solve In the related art, the problem of low accuracy when the image recognition model recognizes various forms of characters.
- the machine learning model is acquired in the following manner:
- Step 1 Use first images of multiple groups of second target objects included in the first sample information as input information of the original model
- Step 2 processing the input information according to the first rule in the original model to obtain the second image of the second target object;
- Step 3 Acquire the similarity between the actual second image and the second target object second image according to the actual second image actually obtained in advance of the second target object first image;
- Step 4 When the similarity is higher than the threshold, output the current model as a machine learning model; when the similarity is lower than the threshold, adjust the parameters in the current model, and repeat steps 1 to 4 repeatedly.
- using the camera to obtain the first image of the first target object includes: using the camera to capture the first target object at a first angle to obtain the first image.
- the first image may be an ID photo or the like in the related art, or a photo of a target person captured in a street camera, and based on the photo, a photo of a suspect in another scene may be obtained.
- a second image of the first target object is acquired: angle tilt, image rotation, contrast adjustment, image resolution Rate adjustment, off-axis processing, binarization processing, hue processing, perspective processing.
- the processing of the first image of the first target object by the machine learning model may include the processing described in the above embodiment.
- using the machine learning model to obtain the second image of the first target object having an association relationship with the first image of the first target object includes: using a machine learning model to analyze the following information of the first image of the first target object At least one of: facial features, gait, clothing, mobile phone MAC, mobile application account; acquiring the second image of the first target object according to the information.
- different weights can be set for facial features, gait, clothing and other information.
- the weights are used when the machine learning model converts the first image.
- the features with high weights focus on the conversion during conversion, which consumes more Computing resources, in the subsequent image matching, priority matching features with high weight.
- facial features are the features with the highest weight.
- comparing the second image of the first target object with the third image, when the two match, matching the third image to the first target object includes: acquiring the first An image feature of a second image of a target object, and an image feature of the third image, wherein the third image is an image including the target object; when the similarity of the image features of the two is greater than a threshold, the The target object in the third image is the first target object.
- a method for feature recognition includes the following steps:
- Step 1 Obtain the fourth feature of the first type
- the first type may be an image taken at an oblique angle
- the second type may be an ID photo
- Step 2 Use a machine learning model to obtain a fifth feature of the second type that has an association relationship with the fourth feature, where the machine learning model uses the second sample information as input information of the original model to the original model
- the second sample information includes a second rule and multiple sets of first-type features, where the second rule is a rule for acquiring second-type features according to the multiple sets of first-type features ;
- Step 3 Compare the fifth feature with the second target feature of the first target object, and when the two match, match the fourth feature to the first target object.
- the feature acquired in real time is taken as the fourth feature, and the machine learning model automatically acquires the fifth feature corresponding to the fourth feature, for example, converts the side view to the front view, and then uses the fifth feature to match the target person’s first feature.
- the two types of features are compared, and if they match, the fourth feature is determined to be the feature of the target person.
- FIG. 4 is a schematic diagram of AI training according to the specific scheme of this application.
- S1 and S2 are added to the model.
- S1 analyzes and judges the AI execution results, including identification accuracy analysis, statistical result probability distribution, etc.
- Inference methods also include two categories, one is direct inference of the original data results, and the other is the introduction of artificial, especially the experience feedback of model users, to guide and intervene in the optimization strategy.
- AI iteration training methods such as loss setting, data selection, etc.
- the following three implementation cases explain the principle and application of inferred AI. In actual operation, one or more of them can be selected depending on the situation.
- FIG. 5 is a schematic diagram of adjusting the license plate according to the first embodiment, as shown in FIG. 5, including two sub-strategies, for example, the original license plate image is obtained after the off-axis and binary processing to obtain new data similar to the black license plate, or after Grayscale and perspective processing resulted in images similar to black and white photos. These new data are used to train the deep neural network parameters, and the ultimate goal is to find the optimal sub-strategy by inferring the guidance module.
- the constant characteristics of the target data can be found, which is helpful for model migration.
- the data we have is ID photo data
- the single target data cannot achieve model training and accurate recognition.
- the image of the target person's neck after rotation can be automatically generated. It can also accurately match facial features when facing pattern recognition problems in complex scenes.
- the traditional verification method is generally the label traversal method, that is, the predicted classification is compared with the real classification one by one.
- the available features include facial features, gait, clothing, mobile phone MAC, and mobile application account.
- This AI can automatically associate the digitized information with the image under inspection and match it through the physical state, geographic location and other information. At the same time, the matching information is fed back to the model, which will remove the image that affects the correct judgment, and modify the parameters accordingly, so as to improve the recognition accuracy.
- FIG. 6 is a schematic diagram of AI training according to scheme 3 of a specific embodiment. As shown in FIG. 6, it is a closed-loop process.
- AI guidance of the staff
- the data output by the AI model is obtained.
- FIG. 7 is a schematic diagram of an AI recognition image according to scheme 3 of a specific embodiment.
- the AI model recognizes multiple forms of images of target persons with the same ID and the same MAC, including people in the algorithm training library Multi-angle training data such as face, head, mobile phone base station positioning, etc., based on the training data, the previously captured images are verified one by one.
- Each image includes the corresponding ID, time, MAC, portrait and location at that time.
- the method according to the above embodiments can be implemented by means of software plus a necessary general hardware platform, and of course, it can also be implemented by hardware, but in many cases the former is Better implementation.
- the technical solution of the present application can essentially be reflected in the form of a software product that contributes to the existing technology.
- the computer software product is stored in a storage medium (such as ROM/RAM, magnetic disk,
- the CD-ROM includes several instructions to enable a terminal device (which may be a mobile phone, computer, server, or network device, etc.) to perform the methods described in the embodiments of the present application.
- a device for identifying features is also provided.
- the device is used to implement the above-mentioned embodiments and preferred implementation modes, and those that have already been described will not be repeated.
- the term "module” may implement a combination of software and/or hardware for a predetermined function.
- the devices described in the following embodiments are preferably implemented in software, implementation of hardware or a combination of software and hardware is also possible and conceived.
- a device for identifying features including:
- a first acquiring module configured to acquire the first characteristic of the first target object
- a second acquisition module for acquiring a second feature of the first target object having an association relationship with the first feature of the first target object using a machine learning model, wherein the machine learning model uses the first sample information as the original
- the input information of the model is a model obtained by training the original model, and the first sample information includes a first rule and first features of multiple groups of second target objects, wherein the first rule is used to A rule that the first characteristic of the second target object acquires the second characteristic of the second target object;
- the matching module is configured to compare the second feature of the first target object with the third feature, and when the two match, match the third feature to the first target object.
- the second feature may be a machine learning model
- the first feature is obtained after processing, such as image rotation, resolution enhancement, voice filtering, etc., and then compared with the currently acquired third feature according to the converted second feature, if the two are similar or the same, it means The third feature also describes the first target object.
- the feature recognition model is based on the virtual feature ratio For the actual third feature, it solves the problem of low accuracy when the feature recognition model in the related art recognizes various forms of features of the target object.
- the above modules can be implemented by software or hardware. For the latter, they can be implemented by the following methods, but not limited to this: the above modules are all located in the same processor; or, the above modules can be combined in any combination The forms are located in different processors.
- the embodiments of the present application also provide a storage medium.
- the above storage medium may be set to store program code for performing the following steps:
- S2 Use a machine learning model to obtain a second feature of the first target object that has an association relationship with the first feature of the first target object, where the machine learning model uses the first sample information as the input information of the original model.
- a model obtained by training the original model, the first sample information includes a first rule and first features of multiple sets of second target objects, wherein the first rule is used to determine the second target object
- the first feature acquires the rule of the second feature of the second target object;
- S3 Compare the second feature of the first target object with the third feature, and when the two match, match the third feature to the first target object.
- the above storage medium may include, but is not limited to: U disk, read-only memory (ROM, Read-Only Memory), random access memory (RAM, Random Access Memory), mobile hard disk, magnetic Various media such as discs or optical discs that can store program codes.
- An embodiment of the present application further provides an electronic device, including a memory and a processor, where the computer program is stored in the memory, and the processor is configured to run the computer program to perform the steps in any one of the foregoing method embodiments.
- the electronic device may further include a transmission device and an input-output device, where the transmission device is connected to the processor, and the input-output device is connected to the processor.
- the foregoing processor may be configured to perform the following steps through a computer program:
- S2 Use a machine learning model to obtain a second feature of the first target object that has an association relationship with the first feature of the first target object, where the machine learning model uses the first sample information as the input information of the original model.
- a model obtained by training the original model, the first sample information includes a first rule and first features of multiple sets of second target objects, wherein the first rule is used to determine the second target object
- the first feature acquires the rule of the second feature of the second target object;
- S3 Compare the second feature of the first target object with the third feature, and when the two match, match the third feature to the first target object.
- modules or steps of the present application can be implemented by a general-purpose computing device, they can be concentrated on a single computing device, or distributed in a network composed of multiple computing devices Above, optionally, they can be implemented with program code executable by the computing device, so that they can be stored in the storage device to be executed by the computing device, and in some cases, can be in a different order than here
- the steps shown or described are performed, or they are made into individual integrated circuit modules respectively, or multiple modules or steps among them are made into a single integrated circuit module to achieve. In this way, this application is not limited to any specific combination of hardware and software.
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- Multimedia (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Data Mining & Analysis (AREA)
- Databases & Information Systems (AREA)
- General Engineering & Computer Science (AREA)
- Evolutionary Computation (AREA)
- Evolutionary Biology (AREA)
- Bioinformatics & Computational Biology (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Artificial Intelligence (AREA)
- Life Sciences & Earth Sciences (AREA)
- Health & Medical Sciences (AREA)
- General Health & Medical Sciences (AREA)
- Ophthalmology & Optometry (AREA)
- Human Computer Interaction (AREA)
- Image Analysis (AREA)
Abstract
Description
Claims (12)
- 一种特征识别的方法,其特征在于,包括:获取第一目标对象的第一特征;使用机器学习模型获取与所述第一目标对象的第一特征具有关联关系的第一目标对象的第二特征,其中,所述机器学习模型是使用第一样本信息作为原始模型的输入信息,对所述原始模型进行训练得到的模型,所述第一样本信息包括第一规则和多组第二目标对象的第一特征,其中,所述第一规则是用于依据所述第二目标对象第一特征获取所述第二目标对象的第二特征的规则;将所述第一目标对象的第二特征与第三特征进行特征比对,在二者匹配时,将所述第三特征对应至所述第一目标对象。
- 根据权利要求1所述的方法,其特征在于,所述特征包括以下之一:目标对象的图像,目标对象的语音信息。
- 根据权利要求2所述的方法,其特征在于,在所述特征为目标对象的图像时,所述方法包括:使用摄像头获取第一目标对象的第一图像;使用机器学习模型获取与所述第一目标对象第一图像具有关联关系的第一目标对象第二图像,其中,所述机器学习模型是使用第一样本信息作为原始模型的输入信息,对所述原始模型进行训练得到的模型;所述第一样本信息包括第一规则和多组第二目标对象的第一图像,其中,所述第一规则是用于依据所述第二目标对象第一图像获取所述第二目标对象的第二图像的规则;将所述第一目标对象第二图像与第三图像进行特征比对,在二者匹配时,将所述第三图像对应至所述第一目标对象。
- 根据权利要求3所述的方法,其特征在于,使用机器学习模型获取与所述第一目标对象第一图像具有关联关系的第一目标对象第二图像之前,通过以下方式获取所述机器学习模型:步骤一,将所述第一样本信息中包括的多组第二目标对象的第一图像,作为所述原始模型的输入信息;步骤二,在所述原始模型中依据所述第一规则处理所述输入信息,获取所述第二目标对象第二图像;步骤三,依据预先获取的所述第二目标对象第一图像实际对应的实际第二图像,获取所述实际第二图像与所述第二目标对象第二图像的相似度;步骤四,在所述相似度高于阈值时,输出当前模型为机器学习模型;在所述相似度低于阈值时,调整当前模型中的参数,并重复执行所述步骤一至步骤四。
- 根据权利要求3所述的方法,其特征在于,使用摄像头获取第一目标对象的第一图像,包括:使用摄像头以第一角度拍摄所述第一目标对象,获取所述第一图像。
- 根据权利要求3所述的方法,其特征在于,使用机器学习模型获取与所述第一目标对象第一图像具有关联关系的第一目标对象第二图像,包括:使用所述机器学习模型对所述第一目标对象第一图像执行以下操作至少之一之后,获取所述第一目标对象第二图像:角度倾斜,图像旋转,对比度调整,图像分辨率调整,离轴处理,二值化处理,色调处理,透视处理。
- 根据权利要求3所述的方法,其特征在于,使用机器学习模型获取与所述第一目标对象第一图像具有关联关系的第一目标对象第二图像,包括:使用机器学习模型分析所述第一目标对象第一图像的以下信息至少之一:面部特征,步态,衣着,手机MAC,移动应用账号;依据所述信息获取所述第一目标对象第二图像。
- 根据权利要求3所述的方法,其特征在于,将所述第一目标对象第二图像与第三图像进行特征比对,在二者匹配时,将所述第三图像对应至所述第一目标对象,包括:获取所述第一目标对象第二图像的图像特征,以及获取所述第三图像的图像特征,其中,所述第三图像为包括目标对象的图像;在二者的图像特征的相似度大于阈值时,判断所述第三图像中的目标对象为所述第一目标对象。
- 一种特征识别的方法,其特征在于,包括:获取第一类型的第四特征;使用机器学习模型获取与所述第四特征具有关联关系的第二类型的第五特征,其中,所述机器学习模型是使用第二样本信息作为原始模型的输入信息对所述原始模型进行训练得到的模型,所述第二样本信息包括第二规则和多组第一类型特征,其中,所述第二规则是用于依据所述多组第一类型特征获取第二类型特征的规则;将所述第五特征与第一目标对象第二类型特征进行特征比对,在二者匹配时,将所述第四特征对应至所述第一目标对象。
- 一种特征识别的装置,其特征在于,包括:第一获取模块,用于获取第一目标对象的第一特征;第二获取模块,用于使用机器学习模型获取与所述第一目标对象第一特征具有关联关系的第一目标对象第二特征,其中,所述机器学习模型是使用第一样本信息作为原始模型的输入信息对所述原始模型进行训练得到的模型,所述第一样本信息包括第一规则和多组第二目标对象的第一特征,其中,所述第一规则是用于依据所述第二目标对象第一特征获取所述第二目标对象的第二特征的规则;匹配模块,用于将所述第一目标对象第二特征与第三特征进行特征比对,在二者匹配时,将所述第三特征对应至所述第一目标对象。
- 一种存储介质,其特征在于,所述存储介质中存储有计算机程序,其中,所述计算机程序被设置为运行时执行所述权利要求1至9任一项中所述的方法。
- 一种电子装置,包括存储器和处理器,其特征在于,所述存储器中存储有计算机程序,所述处理器被设置为运行所述计算机程序以执行所述权利要求1至9任一项中所述的方法。
Applications Claiming Priority (2)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201811492201.2A CN111291773A (zh) | 2018-12-06 | 2018-12-06 | 特征识别的方法及装置 |
CN201811492201.2 | 2018-12-06 |
Publications (1)
Publication Number | Publication Date |
---|---|
WO2020114135A1 true WO2020114135A1 (zh) | 2020-06-11 |
Family
ID=70974901
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
PCT/CN2019/112909 WO2020114135A1 (zh) | 2018-12-06 | 2019-10-24 | 特征识别的方法及装置 |
Country Status (2)
Country | Link |
---|---|
CN (1) | CN111291773A (zh) |
WO (1) | WO2020114135A1 (zh) |
Cited By (5)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN111950591A (zh) * | 2020-07-09 | 2020-11-17 | 中国科学院深圳先进技术研究院 | 模型训练方法、交互关系识别方法、装置及电子设备 |
CN112465565A (zh) * | 2020-12-11 | 2021-03-09 | 加和(北京)信息科技有限公司 | 一种基于机器学习的用户画像预测的方法及装置 |
CN112597542A (zh) * | 2020-12-04 | 2021-04-02 | 光大科技有限公司 | 目标资产数据的聚合方法及装置、存储介质、电子装置 |
CN113709367A (zh) * | 2021-08-24 | 2021-11-26 | 浙江大华技术股份有限公司 | 一种控制聚焦的方法、装置及电子设备 |
CN117330507A (zh) * | 2023-10-12 | 2024-01-02 | 苏州星帆华镭光电科技有限公司 | 手持激光仪的远程测试管控方法 |
Citations (5)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
US20070086627A1 (en) * | 2005-10-18 | 2007-04-19 | Samsung Electronics Co., Ltd. | Face identification apparatus, medium, and method |
WO2015122789A1 (en) * | 2014-02-11 | 2015-08-20 | 3Divi Company | Facial recognition and user authentication method |
CN106203400A (zh) * | 2016-07-29 | 2016-12-07 | 广州国信达计算机网络通讯有限公司 | 一种人脸识别方法及装置 |
CN108764031A (zh) * | 2018-04-17 | 2018-11-06 | 平安科技(深圳)有限公司 | 识别人脸的方法、装置、计算机设备和存储介质 |
CN108805977A (zh) * | 2018-06-06 | 2018-11-13 | 浙江大学 | 一种基于端到端卷积神经网络的人脸三维重建方法 |
-
2018
- 2018-12-06 CN CN201811492201.2A patent/CN111291773A/zh active Pending
-
2019
- 2019-10-24 WO PCT/CN2019/112909 patent/WO2020114135A1/zh active Application Filing
Patent Citations (5)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
US20070086627A1 (en) * | 2005-10-18 | 2007-04-19 | Samsung Electronics Co., Ltd. | Face identification apparatus, medium, and method |
WO2015122789A1 (en) * | 2014-02-11 | 2015-08-20 | 3Divi Company | Facial recognition and user authentication method |
CN106203400A (zh) * | 2016-07-29 | 2016-12-07 | 广州国信达计算机网络通讯有限公司 | 一种人脸识别方法及装置 |
CN108764031A (zh) * | 2018-04-17 | 2018-11-06 | 平安科技(深圳)有限公司 | 识别人脸的方法、装置、计算机设备和存储介质 |
CN108805977A (zh) * | 2018-06-06 | 2018-11-13 | 浙江大学 | 一种基于端到端卷积神经网络的人脸三维重建方法 |
Cited By (9)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN111950591A (zh) * | 2020-07-09 | 2020-11-17 | 中国科学院深圳先进技术研究院 | 模型训练方法、交互关系识别方法、装置及电子设备 |
CN111950591B (zh) * | 2020-07-09 | 2023-09-01 | 中国科学院深圳先进技术研究院 | 模型训练方法、交互关系识别方法、装置及电子设备 |
CN112597542A (zh) * | 2020-12-04 | 2021-04-02 | 光大科技有限公司 | 目标资产数据的聚合方法及装置、存储介质、电子装置 |
CN112597542B (zh) * | 2020-12-04 | 2023-10-24 | 光大科技有限公司 | 目标资产数据的聚合方法及装置、存储介质、电子装置 |
CN112465565A (zh) * | 2020-12-11 | 2021-03-09 | 加和(北京)信息科技有限公司 | 一种基于机器学习的用户画像预测的方法及装置 |
CN112465565B (zh) * | 2020-12-11 | 2023-09-26 | 加和(北京)信息科技有限公司 | 一种基于机器学习的用户画像预测的方法及装置 |
CN113709367A (zh) * | 2021-08-24 | 2021-11-26 | 浙江大华技术股份有限公司 | 一种控制聚焦的方法、装置及电子设备 |
CN117330507A (zh) * | 2023-10-12 | 2024-01-02 | 苏州星帆华镭光电科技有限公司 | 手持激光仪的远程测试管控方法 |
CN117330507B (zh) * | 2023-10-12 | 2024-04-05 | 苏州星帆华镭光电科技有限公司 | 手持激光仪的远程测试管控方法 |
Also Published As
Publication number | Publication date |
---|---|
CN111291773A (zh) | 2020-06-16 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
WO2020114135A1 (zh) | 特征识别的方法及装置 | |
WO2021077984A1 (zh) | 对象识别方法、装置、电子设备及可读存储介质 | |
CN111079639B (zh) | 垃圾图像分类模型构建的方法、装置、设备及存储介质 | |
CN112069929B (zh) | 一种无监督行人重识别方法、装置、电子设备及存储介质 | |
KR101185525B1 (ko) | 서포트 벡터 머신 및 얼굴 인식에 기초한 자동 생체 식별 | |
US10769261B2 (en) | User image verification | |
CN109446889B (zh) | 基于孪生匹配网络的物体追踪方法及装置 | |
CN111476268A (zh) | 翻拍识别模型训练、图像识别方法、装置、设备及介质 | |
CN104915972A (zh) | 图像处理装置、图像处理方法以及程序 | |
US20210383123A1 (en) | System and Method for Predicting Formation in Sports | |
WO2023124278A1 (zh) | 图像处理模型的训练方法、图像分类方法及装置 | |
WO2023116565A1 (zh) | 智能设计网络安全架构图的方法 | |
US20240087368A1 (en) | Companion animal life management system and method therefor | |
CN116258861A (zh) | 基于多标签学习的半监督语义分割方法以及分割装置 | |
Ashwinkumar et al. | Deep learning based approach for facilitating online proctoring using transfer learning | |
CN113743426A (zh) | 一种训练方法、装置、设备以及计算机可读存储介质 | |
CN113283388B (zh) | 活体人脸检测模型的训练方法、装置、设备及存储介质 | |
CN107369086A (zh) | 一种身份证打码系统及方法 | |
CN111860601B (zh) | 预测大型真菌种类的方法及装置 | |
Kumar et al. | Convolutional Neural Network based Automated Attendance System by using Facial Recognition Domain | |
Machado et al. | Improving face detection | |
CN113837236B (zh) | 图像中目标对象的识别方法、装置、终端设备及存储介质 | |
CN112199976A (zh) | 证件图片生成方法及装置 | |
Vivek et al. | A Way to Mark Attentance using Face Recognition using PL | |
CN112101185B (zh) | 一种训练皱纹检测模型的方法、电子设备及存储介质 |
Legal Events
Date | Code | Title | Description |
---|---|---|---|
121 | Ep: the epo has been informed by wipo that ep was designated in this application |
Ref document number: 19893671 Country of ref document: EP Kind code of ref document: A1 |
|
NENP | Non-entry into the national phase |
Ref country code: DE |
|
122 | Ep: pct application non-entry in european phase |
Ref document number: 19893671 Country of ref document: EP Kind code of ref document: A1 |
|
122 | Ep: pct application non-entry in european phase |
Ref document number: 19893671 Country of ref document: EP Kind code of ref document: A1 |
|
32PN | Ep: public notification in the ep bulletin as address of the adressee cannot be established |
Free format text: NOTING OF LOSS OF RIGHTS PURSUANT TO RULE 112(1) EPC (EPO FORM 1205A DATED 24.06.2022) |
|
122 | Ep: pct application non-entry in european phase |
Ref document number: 19893671 Country of ref document: EP Kind code of ref document: A1 |