CN114140873A - Gait recognition method based on convolutional neural network multi-level features - Google Patents
Gait recognition method based on convolutional neural network multi-level features Download PDFInfo
- Publication number
- CN114140873A CN114140873A CN202111317122.XA CN202111317122A CN114140873A CN 114140873 A CN114140873 A CN 114140873A CN 202111317122 A CN202111317122 A CN 202111317122A CN 114140873 A CN114140873 A CN 114140873A
- Authority
- CN
- China
- Prior art keywords
- layer
- features
- gait
- neural network
- convolutional neural
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Pending
Links
- 230000005021 gait Effects 0.000 title claims abstract description 65
- 238000000034 method Methods 0.000 title claims abstract description 60
- 238000013527 convolutional neural network Methods 0.000 title claims abstract description 27
- 238000000605 extraction Methods 0.000 claims abstract description 10
- 238000007781 pre-processing Methods 0.000 claims abstract description 4
- 238000012549 training Methods 0.000 claims description 35
- 230000008569 process Effects 0.000 claims description 13
- 238000011176 pooling Methods 0.000 claims description 9
- 230000006870 function Effects 0.000 claims description 8
- 239000000284 extract Substances 0.000 claims description 7
- 235000019580 granularity Nutrition 0.000 claims description 6
- ORILYTVJVMAKLC-UHFFFAOYSA-N Adamantane Natural products C1C(C2)CC3CC1CC2C3 ORILYTVJVMAKLC-UHFFFAOYSA-N 0.000 claims description 3
- 238000013507 mapping Methods 0.000 claims description 3
- 230000000877 morphologic effect Effects 0.000 claims description 3
- 238000010586 diagram Methods 0.000 abstract description 2
- 238000013135 deep learning Methods 0.000 description 2
- 238000013461 design Methods 0.000 description 2
- 230000004075 alteration Effects 0.000 description 1
- 238000013459 approach Methods 0.000 description 1
- 230000009286 beneficial effect Effects 0.000 description 1
- 238000011161 development Methods 0.000 description 1
- 230000000694 effects Effects 0.000 description 1
- 210000000887 face Anatomy 0.000 description 1
- 230000003993 interaction Effects 0.000 description 1
- 210000000554 iris Anatomy 0.000 description 1
- 238000012986 modification Methods 0.000 description 1
- 230000004048 modification Effects 0.000 description 1
- 238000012544 monitoring process Methods 0.000 description 1
Images
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/22—Matching criteria, e.g. proximity measures
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/25—Fusion techniques
- G06F18/253—Fusion techniques of extracted features
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/04—Architecture, e.g. interconnection topology
- G06N3/045—Combinations of networks
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/08—Learning methods
- G06N3/084—Backpropagation, e.g. using gradient descent
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Physics & Mathematics (AREA)
- Data Mining & Analysis (AREA)
- Evolutionary Computation (AREA)
- Life Sciences & Earth Sciences (AREA)
- Artificial Intelligence (AREA)
- General Physics & Mathematics (AREA)
- General Engineering & Computer Science (AREA)
- General Health & Medical Sciences (AREA)
- Software Systems (AREA)
- Molecular Biology (AREA)
- Computing Systems (AREA)
- Biophysics (AREA)
- Biomedical Technology (AREA)
- Mathematical Physics (AREA)
- Computational Linguistics (AREA)
- Health & Medical Sciences (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Bioinformatics & Computational Biology (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Evolutionary Biology (AREA)
- Image Analysis (AREA)
Abstract
A gait recognition method based on multi-level features of a convolutional neural network comprises the following steps: reading an input gait image sequence, and preprocessing the image sequence; extracting features of different levels from the image sequence through the feature extraction branches with different convolution sizes; fusing the extracted features of different layers in different modes to obtain and store the features finally used for gait recognition; calculating loss by using a loss function according to the characteristic diagram and the label, and updating the convolutional neural network model by adopting a back propagation algorithm until a preset condition is met, and generating a final recognition model; and inputting the gait target to be recognized into the final recognition model, matching the output result of the model with the stored characteristic values, and taking the result with the highest similarity in the stored characteristic values as the recognition result. The invention enhances the online data of the input gait outline map, increases the diversity of the input data and improves the robustness of the algorithm to the pedestrian outline under the actual situation.
Description
Technical Field
The invention relates to the field of computer vision and deep learning, in particular to a gait recognition method based on multilayer characteristics of a convolutional neural network.
Background
Gait is a physiological and biological feature that describes a person's walking pattern. Compared with the biological characteristics of human faces, fingerprints, irises and the like, the gait recognition method has the advantages of non-contact, long distance, difficulty in camouflage, no need of active matching and the like, so that the gait recognition technology is widely applied to the fields of access control systems, safety monitoring, human-computer interaction and the like.
With the development of deep learning and convolutional neural network technology, gait recognition technology also makes great progress. The mainstream gait recognition method at present is generally divided into a template-based method and a sequence-based method, and recognition is carried out through a pedestrian gait contour map. Although the gait recognition method based on the sequence, such as GaitSet and GaitPart algorithms, can achieve good effects on public data sets, the GaitSet algorithm is based on the whole contour map for recognition and does not catch local features, the GaitPart algorithm can extract the local features, the extracted feature levels are single, and the recognition accuracy of the two algorithms under actual situations is not enough.
Disclosure of Invention
In view of the above, the present invention has been developed to provide a gait recognition method based on multi-level features of a convolutional neural network that overcomes or at least partially solves the above problems.
In order to solve the technical problem, the embodiment of the application discloses the following technical scheme:
a gait recognition method based on multi-level features of a convolutional neural network comprises the following steps:
s100, reading an input gait image sequence, and preprocessing the image sequence;
s200, extracting features of different levels of an image sequence through feature extraction branches with different convolution sizes;
s300, fusing the extracted features of different layers in different modes to obtain and store features finally used for gait recognition;
s400, calculating loss by using a loss function according to the feature map and the label, and updating the convolutional neural network model by adopting a back propagation algorithm until a preset condition is met, and generating a final recognition model;
s500, inputting the gait target to be recognized into a final recognition model, matching the output result of the model with the stored characteristic values, and taking the result with the highest similarity in the stored characteristic values as a recognition result.
Further, the specific method of S100 is: downloading a CASIA-B open source gait recognition data set, judging the length of a gait sequence in the data set, taking 30 middle images if the length of the sequence is more than 150, and discarding the sequence if the length of the sequence is less than 30; the input gait sequence is cut into 64 x 64 size according to the principle of the edge and the center line in the height and width directions.
Further, the specific method of S100 further includes: randomly performing morphological opening operation or closing operation on the image according to a certain proportion of ratio, wherein the kernel size of the opening and closing operation is set to be 3, the proportion ratio is set to be 0.2 during training, and the proportion ratio is set to be 0 during reasoning; the input sequence number of each time is P x K, wherein P is the number of pedestrians in each batch during training or reasoning, and K is the gait sequence number of each selected pedestrian under different viewing angles, dressing, backpacks or walking conditions.
Further, the specific method of S200 includes: extracting features of different granularities of an input image sequence by adopting three convolutions of 1x1, 3x3 and 5x5, wherein the number of input channels of the three convolutions is 1, the number of output channels of the three convolutions is 32, the step length is 2, and n image sequences with the length of 64 x 64 and the length of s-30 are input and extracted by the features of different granularities; and splicing the 3 feature maps with the size of 32 × 32 in a channel dimension through a Concat layer, and finally obtaining the feature maps containing different levels of information.
Furthermore, the Block layer is composed of two convolution layers with convolution size of 3 and step size of 1 and a maximum pooling with step size of 2, the number of input channels of the first convolution layer is 96, the number of output channels is 64, the number of input channels of the second convolution layer is 64, the number of output channels is 128, and the output of the feature map extracted by the Block layer and containing different layers of information is n × 30 × 128 × 16.
Further, the specific method of S300 includes: fusing the multi-level features extracted in the S200 in the directions of W and H of the feature map by adopting a Block _ W layer, a Block _ H layer, an HPM layer, a VPM layer and a Concat2 layer; the Block _ W layer and the Block _ H layer are formed by 2 layers of convolutions with the convolution size of 3 and the step length of 1, the number of input channels and the number of output channels of the first layer of convolution are respectively 128 and 256, and the number of input channels and the number of output channels of the second layer of convolution are respectively 256 and 256.
Further, the specific method of S300 further includes: the Block _ W layer equally divides the features output by the feature extraction layer into n equal parts in the W direction, local features are respectively extracted from the n strip-shaped feature graphs divided in the W direction by using convolution operation, the features concat are aggregated, and strip-shaped local information in the W direction is extracted by adopting global maximum set pooling; the Block _ H layer equally divides the features output by the feature extraction layer into n equal divisions in the H direction, respectively extracts local features from the n H-direction segmented strip feature graphs by using convolution operation, then aggregates the features concat, and extracts the H-direction strip local information by adopting global maximum set pooling; mapping the h-direction pooled features to a discrimination space by the HPM layer by using an independent full-connection layer; the VPM layer maps the pooled features in each w direction to a discrimination space by using an independent full connection layer; the Concat2 layer contains a Concat operation to splice the output feature maps of the HPM layer and the VPM layer together in the channel dimension to form a feature map containing local features in both the w direction and the h direction.
Further, in S400, Treplitloss and centerlos are used for joint training, and the loss function is designed as follows:
LOSS＝Ltri+β·LCen
wherein the joint loss is represented by a ternary loss LtriAnd central loss LCenAnd two lost weight balance factors beta; ternary loss LtriIn the formula, N is the number of samples in a training batch,
for a certain sample chosen at random in the training,
the positive samples corresponding to the training samples are,
the negative sample corresponding to the training sample is the negative sample,
the in-class euclidean distance between the randomly selected sample and the similar sample,
the Euclidean distance between the randomly selected sample and the different samples, alpha is the minimum difference value between the inter-class distance and the intra-class distance corresponding to the sample, and the center loss LCenIn xiC is the class center of all samples, and is obtained by the training process.
Further, the training process is optimized by adopting an Adam algorithm, the learning rate is set to be 0.0001, the momentum is 0.9, and the training is stopped if the loss of the training process is less than 0.0000001 or the iteration number is more than 100000.
The technical scheme provided by the embodiment of the invention has the beneficial effects that at least:
the gait recognition method based on the multilayer characteristics of the convolutional neural network disclosed by the invention has the advantages that the online data enhancement is carried out on the input gait outline graph, the diversity of the input data is increased, and the robustness of the algorithm to the pedestrian outline under the actual situation is improved; the method adopts convolution of different receptive fields to extract gait information and respectively expresses the gait by using different local characteristics in the height h and width w directions, and adopts ternary loss and central loss in the algorithm training process, thereby increasing the distance between different classes, reducing the distance between the same classes and improving the accuracy of gait recognition.
The technical solution of the present invention is further described in detail by the accompanying drawings and embodiments.
Drawings
The accompanying drawings, which are included to provide a further understanding of the invention and are incorporated in and constitute a part of this specification, illustrate embodiments of the invention and together with the description serve to explain the principles of the invention and not to limit the invention. In the drawings:
fig. 1 is a flowchart of a gait recognition method based on a convolutional neural network multi-level feature according to embodiment 1 of the present invention;
fig. 2 is a structural diagram of a convolutional neural network in a gait recognition method based on multi-level features of the convolutional neural network in embodiment 1 of the present invention.
Detailed Description
Exemplary embodiments of the present disclosure will be described in more detail below with reference to the accompanying drawings. While exemplary embodiments of the present disclosure are shown in the drawings, it should be understood that the present disclosure may be embodied in various forms and should not be limited to the embodiments set forth herein. Rather, these embodiments are provided so that this disclosure will be thorough and complete, and will fully convey the scope of the disclosure to those skilled in the art.
In order to solve the problems in the prior art, the embodiment of the invention provides a gait recognition method based on multilayer characteristics of a convolutional neural network.
Example 1
The embodiment discloses a gait recognition method based on multi-level features of a convolutional neural network, as shown in fig. 1, comprising the following steps:
s100, reading an input gait image sequence, and preprocessing the image sequence;
in this embodiment, the specific method of S100 is: downloading a CASIA-B open source gait recognition data set, judging the length of a gait sequence in the data set, taking 30 middle images if the length of the sequence is more than 150, and discarding the sequence if the length of the sequence is less than 30; the input gait sequence is cut into 64 x 64 size according to the principle of the edge and the center line in the height and width directions.
In some preferred embodiments, the specific method of S100 further comprises: randomly performing morphological opening operation or closing operation on the image according to a certain proportion of ratio, wherein the kernel size of the opening and closing operation is set to be 3, the proportion ratio is set to be 0.2 during training, and the proportion ratio is set to be 0 during reasoning; the input sequence number of each time is P x K, wherein P is the number of pedestrians in each batch during training or reasoning, and K is the gait sequence number of each selected pedestrian under different viewing angles, dressing, backpacks or walking conditions. As an example P may be set to 8 and K may be set to 16.
S200, extracting features of different levels of an image sequence through feature extraction branches with different convolution sizes; in this embodiment, as shown in fig. 2, the specific method of S200 includes: extracting features of different granularities from an input image sequence by adopting three convolutions of 1x1, 3x3 and 5x5, wherein the number of input channels of the three convolutions is 1, the number of output channels is 32, the step size is 2, and a feature map of n image sequences with the length of 64 x 64 and the length of s-30 is n x 30 x 32 after extracting the features of the different granularities; splicing 3 characteristic graphs with the size of 32 × 32 in a channel dimension through a Concat layer, and outputting n × 30 × 96 × 32; the Block layer is composed of two convolution layers with convolution size of 3 and step size of 1 and a maximum pooling with step size of 2, the number of input channels of the first convolution layer is 96, the number of output channels is 64, the number of input channels of the second convolution layer is 64, the number of output channels is 128, and the feature graphs containing different levels of information extracted by the Block layer are output as n 30 x 128 x 16.
S300, fusing the extracted features of different layers in different modes to obtain and store features finally used for gait recognition; specifically, in this embodiment, a Block _ W layer, a Block _ H layer, an HPM layer, a VPM layer, and a Concat2 layer are used to fuse the multi-level features extracted in S200 in the W and H directions of the feature map; the Block _ W layer and the Block _ H layer are formed by 2 layers of convolutions with the convolution size of 3 and the step length of 1, the number of input channels and the number of output channels of the first layer of convolution are respectively 128 and 256, and the number of input channels and the number of output channels of the second layer of convolution are respectively 256 and 256.
The Block _ W layer equally divides the features output by the feature extraction layer into n equal parts in the W direction, local features are respectively extracted from the n strip-shaped feature graphs divided in the W direction by using convolution operation, the features concat are aggregated, and strip-shaped local information in the W direction is extracted by adopting global maximum set pooling; the Block _ H layer equally divides the features output by the feature extraction layer into n equal divisions in the H direction, respectively extracts local features from the n H-direction segmented strip feature graphs by using convolution operation, then aggregates the features concat, and extracts the H-direction strip local information by adopting global maximum set pooling; mapping the h-direction pooled features to a discrimination space by the HPM layer by using an independent full-connection layer; the VPM layer maps the pooled features in each w direction to a discrimination space by using an independent full connection layer; the Concat2 layer contains a Concat operation to splice the output feature maps of the HPM layer and the VPM layer together in the channel dimension to form a feature map containing local features in both the w direction and the h direction.
S400, calculating loss by using a loss function according to the feature map and the label, and updating the convolutional neural network model by adopting a back propagation algorithm until a preset condition is met, and generating a final recognition model;
in this embodiment, Treplitloss and Centerlos are used for joint training, and the loss function is designed as follows:
LOSS＝Ltri+β·LCen
wherein the combined loss is threeLoss of element LtriAnd central loss LCenAnd two lost weight balance factors beta; ternary loss LtriIn the formula, N is the number of samples in a training batch,
for a certain sample chosen at random in the training,
the positive samples corresponding to the training samples are,
the negative sample corresponding to the training sample is the negative sample,
the in-class euclidean distance between the randomly selected sample and the similar sample,
the Euclidean distance between the randomly selected sample and the different samples, alpha is the minimum difference value between the inter-class distance and the intra-class distance corresponding to the sample, and the center loss LCenIn xiC is the class center of all samples, and is obtained by the training process. And optimizing by adopting an Adam algorithm in the training process, setting the learning rate to be 0.0001 and the momentum to be 0.9, and stopping training if the loss of the training process is less than 0.0000001 or the iteration times is more than 100000.
S500, inputting the gait target to be recognized into a final recognition model, matching the output result of the model with the stored characteristic values, and taking the result with the highest similarity in the stored characteristic values as a recognition result. When gait recognition is carried out, firstly establishing a query set, namely storing the gait characteristic values of all targets; and for a gait target to be identified, calculating the characteristic value and the Euclidean distance between the gait target and the stored target one by one according to S100-S300, and taking the result closest to the Euclidean distance of the stored target as an identification result.
The gait recognition method based on the multilayer characteristics of the convolutional neural network disclosed by the embodiment performs online data enhancement on an input gait contour map, increases the diversity of input data, and improves the robustness of an algorithm to a pedestrian contour under an actual scene; the method adopts convolution of different receptive fields to extract gait information and respectively expresses the gait by using different local characteristics in the height h and width w directions, and adopts ternary loss and central loss in the algorithm training process, thereby increasing the distance between different classes, reducing the distance between the same classes and improving the accuracy of gait recognition.
It should be understood that the specific order or hierarchy of steps in the processes disclosed is an example of exemplary approaches. Based upon design preferences, it is understood that the specific order or hierarchy of steps in the processes may be rearranged without departing from the scope of the present disclosure. The accompanying method claims present elements of the various steps in a sample order, and are not intended to be limited to the specific order or hierarchy presented.
In the foregoing detailed description, various features are grouped together in a single embodiment for the purpose of streamlining the disclosure. This method of disclosure is not to be interpreted as reflecting an intention that the claimed embodiments of the subject matter require more features than are expressly recited in each claim. Rather, as the following claims reflect, invention lies in less than all features of a single disclosed embodiment. Thus, the following claims are hereby expressly incorporated into the detailed description, with each claim standing on its own as a separate preferred embodiment of the invention.
Those of skill would further appreciate that the various illustrative logical blocks, modules, circuits, and algorithm steps described in connection with the embodiments disclosed herein may be implemented as electronic hardware, computer software, or combinations of both. To clearly illustrate this interchangeability of hardware and software, various illustrative components, blocks, modules, circuits, and steps have been described above generally in terms of their functionality. Whether such functionality is implemented as hardware or software depends upon the particular application and design constraints imposed on the overall system. Skilled artisans may implement the described functionality in varying ways for each particular application, but such implementation decisions should not be interpreted as causing a departure from the scope of the present disclosure.
The steps of a method or algorithm described in connection with the embodiments disclosed herein may be embodied directly in hardware, in a software module executed by a processor, or in a combination of the two. A software module may reside in RAM memory, flash memory, ROM memory, EPROM memory, EEPROM memory, registers, hard disk, a removable disk, a CD-ROM, or any other form of storage medium known in the art. An exemplary storage medium is coupled to the processor such the processor can read information from, and write information to, the storage medium. Of course, the storage medium may also be integral to the processor. The processor and the storage medium may reside in an ASIC. The ASIC may reside in a user terminal. Of course, the processor and the storage medium may reside as discrete components in a user terminal.
For a software implementation, the techniques described herein may be implemented with modules (e.g., procedures, functions, and so on) that perform the functions described herein. The software codes may be stored in memory units and executed by processors. The memory unit may be implemented within the processor or external to the processor, in which case it can be communicatively coupled to the processor via various means as is known in the art.
What has been described above includes examples of one or more embodiments. It is, of course, not possible to describe every conceivable combination of components or methodologies for purposes of describing the aforementioned embodiments, but one of ordinary skill in the art may recognize that many further combinations and permutations of various embodiments are possible. Accordingly, the embodiments described herein are intended to embrace all such alterations, modifications and variations that fall within the scope of the appended claims. Furthermore, to the extent that the term "includes" is used in either the detailed description or the claims, such term is intended to be inclusive in a manner similar to the term "comprising" as "comprising" is interpreted when employed as a transitional word in a claim. Furthermore, any use of the term "or" in the specification of the claims is intended to mean a "non-exclusive or".
Claims (9)
1. A gait recognition method based on multilayer characteristics of a convolutional neural network is characterized by comprising the following steps:
s100, reading an input gait image sequence, and preprocessing the image sequence;
s200, extracting features of different levels of an image sequence through feature extraction branches with different convolution sizes;
s300, fusing the extracted features of different layers in different modes to obtain and store features finally used for gait recognition;
s400, calculating loss by using a loss function according to the feature map and the label, and updating the convolutional neural network model by adopting a back propagation algorithm until a preset condition is met, and generating a final recognition model;
s500, inputting the gait target to be recognized into a final recognition model, matching the output result of the model with the stored characteristic values, and taking the result with the highest similarity in the stored characteristic values as a recognition result.
2. The gait recognition method based on the multi-level features of the convolutional neural network as claimed in claim 1, wherein the specific method of S100 is as follows: downloading a CASIA-B open source gait recognition data set, judging the length of a gait sequence in the data set, and if the sequence length is more than 150, extracting 30 images in the middle; if the sequence length is less than 30, discarding the sequence; the input gait sequence is cut into 64 x 64 size according to the principle of the edge and the center line in the height and width directions.
3. The gait recognition method based on the multi-level features of the convolutional neural network as claimed in claim 1, wherein the specific method of S100 further comprises: randomly performing morphological opening operation or closing operation on the image according to a certain proportion of ratio, wherein the kernel size of the opening and closing operation is set to be 3, the proportion ratio is set to be 0.2 during training, and the proportion ratio is set to be 0 during reasoning; the input sequence number of each time is P x K, wherein P is the number of pedestrians in each batch during training or reasoning, and K is the gait sequence number of each selected pedestrian under different viewing angles, dressing, backpacks or walking conditions.
4. The gait recognition method based on the multi-level features of the convolutional neural network as claimed in claim 1, wherein the specific method of S200 comprises: extracting features of different granularities of an input image sequence by adopting three convolutions of 1x1, 3x3 and 5x5, wherein the number of input channels of the three convolutions is 1, the number of output channels of the three convolutions is 32, the step length is 2, and n image sequences with the length of 64 x 64 and the length of s-30 are input and extracted by the features of different granularities; and splicing the 3 feature maps with the size of 32 × 32 in a channel dimension through a Concat layer, and finally obtaining the feature maps containing different levels of information.
5. A gait recognition method based on the multilevel characteristics of the convolutional neural network as claimed in claim 4, characterized in that the Block layer is composed of two convolutional layers with convolution size of 3 and step size of 1 and a maximum pooling with step size of 2, the number of input channels of the first layer of convolution is 96, the number of output channels is 64, the number of input channels of the second layer of convolution is 64, the number of output channels is 128, and the output of the feature map containing different levels of information extracted by the Block layer is n 30 x 128 x 16.
6. The gait recognition method based on the multi-level features of the convolutional neural network as claimed in claim 1, wherein the specific method of S300 comprises: fusing the multi-level features extracted in the S200 in the directions of W and H of the feature map by adopting a Block _ W layer, a Block _ H layer, an HPM layer, a VPM layer and a Concat2 layer; the Block _ W layer and the Block _ H layer are formed by 2 layers of convolutions with the convolution size of 3 and the step length of 1, the number of input channels and the number of output channels of the first layer of convolution are respectively 128 and 256, and the number of input channels and the number of output channels of the second layer of convolution are respectively 256 and 256.
7. The method for gait recognition based on multi-level features of convolutional neural network as claimed in claim 6, wherein the specific method of S300 further comprises: the Block _ W layer equally divides the features output by the feature extraction layer into n equal parts in the W direction, local features are respectively extracted from the n strip-shaped feature graphs divided in the W direction by using convolution operation, the features concat are aggregated, and strip-shaped local information in the W direction is extracted by adopting global maximum set pooling; the Block _ H layer equally divides the features output by the feature extraction layer into n equal divisions in the H direction, respectively extracts local features from the n H-direction segmented strip feature graphs by using convolution operation, then aggregates the features concat, and extracts the H-direction strip local information by adopting global maximum set pooling; mapping the h-direction pooled features to a discrimination space by the HPM layer by using an independent full-connection layer; the VPM layer maps the pooled features in each w direction to a discrimination space by using an independent full connection layer; the Concat2 layer contains a Concat operation to splice the output feature maps of the HPM layer and the VPM layer together in the channel dimension to form a feature map containing local features in both the w direction and the h direction.
8. The gait recognition method based on multilevel features of convolutional neural network as claimed in claim 1, wherein in S400, joint training is performed by using Treplitloss and cenerloss, and the loss function is designed as follows:
LOSS＝Ltri+β·LCen
wherein the joint loss is represented by a ternary loss LtriAnd central loss LCenAnd two lost weight balance factors beta; ternary loss LtriIn the formula, N is the number of samples in a training batch,
for a certain sample chosen at random in the training,
the positive samples corresponding to the training samples are,
the negative sample corresponding to the training sample is the negative sample,
the in-class euclidean distance between the randomly selected sample and the similar sample,
the Euclidean distance between the randomly selected sample and the different samples, alpha is the minimum difference value between the inter-class distance and the intra-class distance corresponding to the sample, and the center loss LCenIn xiC is the class center of all samples, and is obtained by the training process.
9. The gait recognition method based on the multilevel features of the convolutional neural network as claimed in claim 8, wherein the training process is optimized by using Adam algorithm, the learning rate is set to 0.0001, the momentum is 0.9, and the training is stopped if the loss of the training process is less than 0.0000001 or the iteration number is more than 100000 times.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202111317122.XA CN114140873A (en) | 2021-11-09 | 2021-11-09 | Gait recognition method based on convolutional neural network multi-level features |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202111317122.XA CN114140873A (en) | 2021-11-09 | 2021-11-09 | Gait recognition method based on convolutional neural network multi-level features |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN114140873A true CN114140873A (en) | 2022-03-04 |
Family
ID=80392761
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202111317122.XA Pending CN114140873A (en) | 2021-11-09 | 2021-11-09 | Gait recognition method based on convolutional neural network multi-level features |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN114140873A (en) |
Cited By (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN114627424A (en) * | 2022-03-25 | 2022-06-14 | 合肥工业大学 | Gait recognition method and system based on visual angle transformation |
Citations (8)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20170243058A1 (en) * | 2014-10-28 | 2017-08-24 | Watrix Technology | Gait recognition method based on deep learning |
| CN110110668A (en) * | 2019-05-08 | 2019-08-09 | 湘潭大学 | A kind of gait recognition method based on feedback weight convolutional neural networks and capsule neural network |
| CN110969087A (en) * | 2019-10-31 | 2020-04-07 | 浙江省北大信息技术高等研究院 | Gait recognition method and system |
| US20200320284A1 (en) * | 2018-04-12 | 2020-10-08 | Tencent Technology (Shenzhen) Company Ltd | Media processing method, related apparatus, and storage medium |
| CN111985332A (en) * | 2020-07-20 | 2020-11-24 | 浙江工业大学 | Gait recognition method for improving loss function based on deep learning |
| CN112818755A (en) * | 2021-01-13 | 2021-05-18 | 华中科技大学 | Gait recognition method based on active learning |
| WO2021115159A1 (en) * | 2019-12-09 | 2021-06-17 | 中兴通讯股份有限公司 | Character recognition network model training method, character recognition method, apparatuses, terminal, and computer storage medium therefor |
| US20210224524A1 (en) * | 2020-01-22 | 2021-07-22 | Board Of Trustees Of Michigan State University | Systems And Methods For Gait Recognition Via Disentangled Representation Learning |
-
2021
- 2021-11-09 CN CN202111317122.XA patent/CN114140873A/en active Pending
Patent Citations (8)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20170243058A1 (en) * | 2014-10-28 | 2017-08-24 | Watrix Technology | Gait recognition method based on deep learning |
| US20200320284A1 (en) * | 2018-04-12 | 2020-10-08 | Tencent Technology (Shenzhen) Company Ltd | Media processing method, related apparatus, and storage medium |
| CN110110668A (en) * | 2019-05-08 | 2019-08-09 | 湘潭大学 | A kind of gait recognition method based on feedback weight convolutional neural networks and capsule neural network |
| CN110969087A (en) * | 2019-10-31 | 2020-04-07 | 浙江省北大信息技术高等研究院 | Gait recognition method and system |
| WO2021115159A1 (en) * | 2019-12-09 | 2021-06-17 | 中兴通讯股份有限公司 | Character recognition network model training method, character recognition method, apparatuses, terminal, and computer storage medium therefor |
| US20210224524A1 (en) * | 2020-01-22 | 2021-07-22 | Board Of Trustees Of Michigan State University | Systems And Methods For Gait Recognition Via Disentangled Representation Learning |
| CN111985332A (en) * | 2020-07-20 | 2020-11-24 | 浙江工业大学 | Gait recognition method for improving loss function based on deep learning |
| CN112818755A (en) * | 2021-01-13 | 2021-05-18 | 华中科技大学 | Gait recognition method based on active learning |
Non-Patent Citations (1)
| Title |
|---|
| 罗金梅;罗建;李艳梅;赵旭;: "基于多特征融合CNN的人脸识别算法研究", 航空计算技术, no. 03, 25 May 2019 (2019-05-25) * |
Cited By (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN114627424A (en) * | 2022-03-25 | 2022-06-14 | 合肥工业大学 | Gait recognition method and system based on visual angle transformation |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN109784186B (en) | Pedestrian re-identification method and device, electronic equipment and computer-readable storage medium | |
| EP3690742A1 (en) | Method for auto-labeling training images for use in deep learning network to analyze images with high precision, and auto-labeling device using the same | |
| CN112446270B (en) | Training method of pedestrian re-recognition network, pedestrian re-recognition method and device | |
| CN112750140B (en) | Information mining-based disguised target image segmentation method | |
| CN110969087B (en) | Gait recognition method and system | |
| US8885943B2 (en) | Face detection method and apparatus | |
| CN110659589B (en) | Pedestrian re-identification method, system and device based on attitude and attention mechanism | |
| CN112288011B (en) | Image matching method based on self-attention deep neural network | |
| CN109472209B (en) | Image recognition method, device and storage medium | |
| CN109034086B (en) | Vehicle weight identification method, device and system | |
| CN107240087B (en) | Object detection system and method | |
| CN111476806B (en) | Image processing method, image processing device, computer equipment and storage medium | |
| CN111368672A (en) | Construction method and device for genetic disease facial recognition model | |
| US20220165048A1 (en) | Person re-identification device and method | |
| CN112733665B (en) | Face recognition method and system based on lightweight network structure design | |
| CN110991513A (en) | Image target recognition system and method with human-like continuous learning capability | |
| CN111401303A (en) | Cross-visual angle gait recognition method with separated identity and visual angle characteristics | |
| CN115578589B (en) | Unsupervised echocardiography section identification method | |
| CN113962281A (en) | Unmanned aerial vehicle target tracking method based on Siamese-RFB | |
| CN114140873A (en) | Gait recognition method based on convolutional neural network multi-level features | |
| CN110909582B (en) | Face recognition method and equipment | |
| CN114358205A (en) | Model training method, model training device, terminal device, and storage medium | |
| CN117636298A (en) | Vehicle re-identification method, system and storage medium based on multi-scale feature learning | |
| Roushdy | Detecting coins with different radii based on hough transform in noisy and deformed image | |
| CN114792401A (en) | Training method, device and equipment of behavior recognition model and storage medium |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination |