CN109359580B - Footprint identification and gait detection method and device based on deep learning - Google Patents
Footprint identification and gait detection method and device based on deep learning Download PDFInfo
- Publication number
- CN109359580B CN109359580B CN201811191596.2A CN201811191596A CN109359580B CN 109359580 B CN109359580 B CN 109359580B CN 201811191596 A CN201811191596 A CN 201811191596A CN 109359580 B CN109359580 B CN 109359580B
- Authority
- CN
- China
- Prior art keywords
- neural network
- human body
- footprint
- gait
- network
- 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.)
- Active
Links
Images
Classifications
-
- 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
- G06V40/20—Movements or behaviour, e.g. gesture recognition
- G06V40/23—Recognition of whole body movements, e.g. for sport training
- G06V40/25—Recognition of walking or running movements, e.g. gait recognition
-
- 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
- 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
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Physics & Mathematics (AREA)
- Data Mining & Analysis (AREA)
- General Physics & Mathematics (AREA)
- Life Sciences & Earth Sciences (AREA)
- Artificial Intelligence (AREA)
- General Health & Medical Sciences (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Health & Medical Sciences (AREA)
- Evolutionary Computation (AREA)
- General Engineering & Computer Science (AREA)
- Biomedical Technology (AREA)
- Computing Systems (AREA)
- Bioinformatics & Computational Biology (AREA)
- Biophysics (AREA)
- Computational Linguistics (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Molecular Biology (AREA)
- Evolutionary Biology (AREA)
- Mathematical Physics (AREA)
- Software Systems (AREA)
- Psychiatry (AREA)
- Social Psychology (AREA)
- Human Computer Interaction (AREA)
- Multimedia (AREA)
- Image Analysis (AREA)
- Measurement Of The Respiration, Hearing Ability, Form, And Blood Characteristics Of Living Organisms (AREA)
Abstract
The application discloses a footprint identification and gait detection method based on deep learning and a device thereof, which comprises the following steps: step S100: identifying a human body footprint image to be identified and outputting basic identity characteristic information of a footprint owner; step S200: inputting the basic identity characteristic information of the footprint owner into a fourth neural network trained by adopting a human body gait effective video data set, and outputting the gait information of the basic identity characteristic of the human body; step S300: and detecting a human body image matched with the gait information in the video to be detected according to the gait information of the basic identity characteristics of the human body. According to the method, the footprint identification and gait detection neural network is constructed, the training process of the network is completed offline, the mode of artificial visual identification is replaced, and the efficiency and the accuracy are improved.
Description
Technical Field
The application relates to a footprint identification and gait detection method and device based on deep learning, and belongs to the field of image identification.
Background
At present, in the criminal case detection process, fingerprint information of a case sending site is firstly extracted by a social public security organ, but if a criminal suspect wears a shield such as gloves to make a case, valuable fingerprint information cannot be extracted from the case sending site.
The footprint information of the person has the identity characteristic information of the human body due to the habit of the person to grow for a long time or leg diseases and the like. The footprint as an identity characteristic of a human body plays an important role in the process of detecting criminal cases in the society at present,
at present, most of public security organs need to employ experts with many years of footprint identification experience for manual visual identification, but the experts are fewer nationwide, and the culture of the footprint identification experts is time-consuming and labor-consuming. This is due to several reasons that footprint identification is not widespread nationwide.
Disclosure of Invention
According to one aspect of the application, a footprint identification and gait detection method based on deep learning is provided. According to the method, the gait and the footprints of various crowds are analyzed and identified by the neural network through a deep learning method, so that the identification accuracy is improved, meanwhile, the dependence of the footprint gait identification on experts is reduced, and the popularization and the application of the footprint identification are facilitated.
The footprint identification and gait detection method based on deep learning comprises the following steps:
step S100: identifying a human body footprint image to be identified and outputting basic identity characteristic information of a footprint owner;
step S200: inputting the basic identity characteristic information of the footprint owner into a fourth neural network trained by adopting a human body gait effective video data set, and outputting the gait information of the human body basic identity characteristic;
step S300: and detecting a human body image matched with the gait information in the video to be detected according to the gait information of the basic identity characteristics of the human body.
Preferably, the step of "identifying" in said step S100: the method comprises the steps of collecting a human body footprint effective image data set, constructing a first neural network according to a preset network structure, training the first neural network by adopting the human body footprint effective image data set, obtaining a second neural network, inputting a human body footprint image to be identified into the second neural network, and outputting basic identity characteristic information of a footprint owner.
Preferably, the training of the fourth neural network comprises the steps of: the method comprises the steps of collecting a human body gait effective video data set, constructing a third neural network according to a preset network structure, and training the third neural network by adopting the human body gait effective video data set to obtain a fourth neural network.
Preferably, the step S100 includes the following steps:
step S110: collecting human body footprint images and storing the human body footprint images into a human body footprint image data set in a gathering manner;
step S120: sorting the effective images in the human body footprint image dataset;
step S130: and after the labels are added to the effective images, storing the labeled images into the human body footprint effective image data set.
Preferably, the step S100 includes the following steps:
step S140: inputting the effective data set of the human body footprint image into the first neural network, and then carrying out forward propagation step by step and backward propagation step by step to obtain a training image neural network;
step S150: testing whether the accuracy of the recognition result of the training image neural network reaches a threshold value;
step S160: if the judgment result is yes, outputting the training image neural network as a second neural network, and if the judgment result is no, repeating the steps S140-S150.
Preferably, the step S200 includes the following steps:
step S210: collecting human gait videos and storing the human gait videos as a human gait video data set in a gathering manner;
step S220: sorting the effective videos in the human body gait video data set;
step S230: and after the effective videos are added with the tags, storing the tagged videos as the human gait effective video data set.
Preferably, the step S200 includes the steps of:
step S240: inputting the human body gait effective video data set into the third neural network, and then carrying out forward propagation step by step and backward propagation step by step to obtain a training video neural network;
step S250: testing whether the accuracy of the recognition result of the training video neural network reaches a threshold value;
step S260: if the judgment result is yes, outputting the training video neural network as a fourth neural network, and if the judgment result is no, repeating the steps S240-S250.
Yet another aspect of the present application provides a footprint identification and gait detection device based on deep learning for use in the method as described above, comprising:
a network construction module: the first neural network and the third neural network are constructed according to a preset network structure;
a network training module: the first neural network is trained by adopting the human body footprint effective image data set to obtain a second neural network, and the third neural network is trained by adopting the human body gait effective video data set to obtain a fourth neural network.
A network module: the second neural network is used for recognizing the human body footprint image to be recognized and outputting the basic identity characteristic information of the footprint owner to the fourth neural network, and the fourth neural network outputs the gait information of the basic identity characteristic of the human body;
a detection module: the method is used for detecting the human body image matched with the gait information in the video to be detected according to the gait information of the basic identity characteristics of the human body.
Preferably, the network construction module comprises a first network construction module and a second network construction module, and the first network construction module is used for constructing a first neural network according to a preset network structure; the second network construction module is used for constructing a third neural network according to a preset network structure.
Preferably, the network training module comprises: the first network training module is used for training the first neural network by adopting a human body footprint effective image data set to obtain a second neural network; the second network training module is used for training the third neural network by adopting the human body gait effective video data set to obtain a fourth neural network.
Benefits of the present application include, but are not limited to:
(1) according to the footprint identification and gait detection method and device based on deep learning, the footprint identification and gait detection neural network is constructed, the training process of the network is completed on line, the mode of artificial visual identification is replaced, and the efficiency and accuracy of human body footprint and gait identification are improved.
Drawings
FIG. 1 is a schematic flow chart of a footprint identification and gait detection method based on deep learning in the preferred embodiment of the present application;
FIG. 2 is a schematic flow chart of a footprint identification method based on deep learning in the preferred embodiment of the present application;
FIG. 3 is a schematic block diagram of the process of footprint identification based on deep learning in another preferred embodiment of the present application;
FIG. 4 is a schematic block diagram of a flow chart of a gait detection method based on deep learning in the preferred embodiment of the application;
FIG. 5 is a schematic block diagram of a flow chart of a gait detection method based on deep learning in another preferred embodiment of the present application;
FIG. 6 is a schematic structural diagram of a neural network constructed in the footprint identification and gait detection method based on deep learning in the preferred embodiment of the present application;
fig. 7 is a schematic structural diagram of a footprint identification and gait detection device based on deep learning in the preferred embodiment of the present application.
List of parts and reference numerals:
| reference numerals | Name of component |
| 10 | Network construction module |
| 11 | First network building block |
| 12 | Second network construction module |
| 20 | Network training module |
| 21 | First network training module |
| 22 | Second network training module |
| 30 | Network module |
| 31 | First neural network module |
| 32 | Second neural network module |
| 40 | Image acquisition module |
| 60 | Detection module |
Detailed Description
The present application will be described in detail with reference to examples, but the present application is not limited to these examples.
Referring to fig. 1, the application provides a footprint identification and gait detection method based on deep learning, which comprises the following steps:
step S100: identifying a human body footprint image to be identified and outputting basic identity characteristic information of a footprint owner;
step S200: inputting the basic identity characteristic information of the footprint owner into a fourth neural network trained by adopting a human body gait effective video data set, and outputting the gait information of the human body basic identity characteristic;
step S300: and detecting a human body image matched with the gait information in the video to be detected according to the gait information of the basic identity characteristics of the human body.
The footprint identification in step S100 may be performed according to various existing methods.
Preferably, the step of "identifying" in said step S100: the method comprises the steps of collecting a human body footprint effective image data set, constructing a first neural network according to a preset network structure, training the first neural network by adopting the human body footprint effective image data set, obtaining a second neural network, inputting a human body footprint image to be identified into the second neural network, and outputting basic identity characteristic information of a footprint owner.
And the neural network is adopted to identify the footprint images, so that the identification efficiency and accuracy can be improved.
Preferably, the training of the fourth neural network comprises the steps of: the method comprises the steps of collecting a human body gait effective video data set, constructing a third neural network according to a preset network structure, and training the third neural network by adopting the human body gait effective video data set to obtain a fourth neural network.
The fourth neural network is trained, so that the gait video identification accuracy can be improved.
Referring to fig. 1, 2 and 4, the application provides a footprint identification and gait detection method based on deep learning, which comprises the following steps:
step S100: collecting a human body footprint effective image data set, constructing a first neural network according to a preset network structure, training the first neural network by adopting the human body footprint effective image data set to obtain a second neural network, inputting a human body footprint image to be identified into the second neural network, and outputting basic identity characteristic information of a footprint owner;
step S200: collecting a human body gait effective video data set, constructing a third neural network according to a preset network structure, and training the third neural network by adopting the human body gait effective video data set to obtain a fourth neural network;
step S300: inputting the basic identity characteristic information of the footprint owner into the fourth neural network, outputting the gait information of the basic identity characteristic of the human body, and detecting a human body image matched with the gait information in a video to be detected.
According to the method, a second neural network capable of recognizing the footprint image and a fourth neural network capable of recognizing the gait video are obtained through deep learning and training respectively, and the second neural network and the fourth neural network are used together to realize detection and recognition of the footprint image to the gait at one time, so that the recognition efficiency and the accuracy of a detection result are improved. The information contained in the footprint image is fully utilized, the footprint of the human body to be recognized is combined with the gait for detection, and the occurrence of detection errors is avoided.
Referring to fig. 3, preferably, the step S100 includes the following steps:
step S110: collecting human body footprint images and storing the human body footprint images into a human body footprint image data set in a gathering manner; specifically, the method can be used for collecting public security organization data, web crawlers, video screenshot labels and the like;
step S120: sorting the effective images in the human body footprint image dataset;
step S130: and after the labels are added to the effective images, storing the labeled images into the human body footprint effective image data set.
The label is the basic identity characteristic information of the human body. The basic identity characteristic information of the human body comprises but is not limited to height, weight, age, gender, types of shoes worn, whether diseases exist in legs, whether specialties exist in soles and the like. The effective image in the application refers to a complete image capable of accurately reflecting the required information. For example an image containing a complete footprint.
Referring to fig. 6, the predetermined network structure includes an input layer, a hidden layer, and an output layer. The hidden layer comprises a data preprocessing layer, a convolution layer, an activation function layer, a pooling layer, a full-connection layer and an anti-overfitting layer. Wherein the over-fitting prevention layer adopts L2A regularization method. The first neural network and the third neural network are constructed according to the structure.
Preferably, referring to fig. 3, the step S100 includes the following steps:
step S140: inputting the effective data set of the human body footprint image into the first neural network, and then carrying out forward propagation step by step and backward propagation step by step to obtain a training image neural network;
step S150: testing whether the accuracy of the recognition result of the training image neural network reaches a threshold value;
step S160: if the judgment result is yes, outputting the training image neural network as a second neural network, and if the judgment result is no, repeating the steps S140-S150.
Referring to fig. 3, when in use, the human body footprint image to be recognized only needs to be input into the second neural network, so that the basic identity characteristic information of the footprint owner in the image can be obtained, including but not limited to height, weight, age, sex, whether diseases exist in legs, whether speciality exists in soles and the like.
Preferably, the step S200 includes the following steps:
step S210: collecting human gait videos and storing the human gait videos as a human gait video data set in a gathering manner; specifically, the method can be used for collecting public security organization data, web crawlers, video screenshot labels and the like;
step S220: sorting the effective videos in the human body gait video data set;
step S230: and after the effective videos are added with the tags, storing the tagged videos as the human gait effective video data set.
The label is the basic identity characteristic information of the human body. The basic identity characteristic information of the human body comprises but is not limited to height, weight, age, gender, types of shoes worn, whether diseases exist in legs, whether specialties exist in soles and the like.
Referring to fig. 5, preferably, the step S200 includes the steps of:
step S240: inputting the human body gait effective video data set into the third neural network, and then carrying out forward propagation step by step and backward propagation step by step to obtain a training video neural network;
step S250: testing whether the accuracy of the recognition result of the training video neural network reaches a threshold value;
step S260: if the judgment result is yes, outputting the training video neural network as a fourth neural network, and if the judgment result is no, repeating the steps S240-S250.
Referring to fig. 5, when in use, the basic identity characteristics of the human body are input into the fourth neural network, so that the basic gait information of the identity characteristic owner can be obtained.
Referring to fig. 7, another aspect of the present application also provides an apparatus for the above method, comprising:
the network construction module 10: the first neural network and the third neural network are constructed according to a preset network structure;
the network training module 20: the first neural network is trained by adopting the human body footprint effective image data set to obtain a second neural network, and the third neural network is trained by adopting the human body gait effective video data set to obtain a fourth neural network.
The network module 30: the second neural network is used for recognizing the human body footprint image to be recognized and outputting the basic identity characteristic information of the footprint owner to the fourth neural network, and the fourth neural network outputs the gait information of the basic identity characteristic of the human body;
the detection module 60: the method is used for detecting the human body image matched with the gait information in the video to be detected according to the gait information of the basic identity characteristics of the human body.
Preferably, the method comprises the following steps: the image acquisition module 40: the method is used for collecting the human body footprint image to be recognized.
Preferably, the network construction module 10 includes a first network construction module 11 and a second network construction module 12, where the first network construction module 11 is configured to construct a first neural network according to a preset network structure; the second network construction module 12 is configured to construct a third neural network according to a preset network structure.
Preferably, the network training module 20 includes: a first network training module 21 and a second network training module 22, wherein the first network training module 21 is configured to train the first neural network by using a human body footprint effective image data set to obtain a second neural network; the second network training module 22 is configured to train the third neural network with the human gait effective video data set to obtain a fourth neural network.
The first neural network for footprint identification and the third neural network for gait detection are constructed and trained in an offline mode, the first neural network for footprint identification and the third neural network for gait detection are constructed in the network construction module 10, and the first neural network for footprint identification and the third neural network for gait detection are trained in the network training module 20. And training by respectively using the human body footprint image effective data set and the human body gait video effective data set to respectively obtain a second neural network for footprint identification and a fourth neural network for gait detection.
And if a new human body footprint image and/or a human body gait video appear, updating the new human body footprint image and/or the human body gait video to a human body footprint image effective data set and/or a human body gait video effective data set, and retraining the network by using the updated human body footprint image effective data set and/or the human body gait video effective data set to obtain a new second neural network and/or a new fourth neural network.
The present application will be described in detail with reference to specific embodiments.
Embodiment 1 footprint identification and gait detection method based on deep learning
Referring to fig. 2 and 4, the footprint identification and gait detection method based on deep learning in the embodiment includes the following steps:
1. and collecting the human body footprint effective image data set for screening the collected human body footprint images so as to reserve an image data set which meets the network training requirement and is called an effective image data set.
2. The first neural network is constructed and trained in an offline construction and training mode, so that the timeliness of system application can be guaranteed.
3. And finally outputting the basic identity characteristic information of the owner of the provided human body footprint image by the footprint identification method.
4. The collected human body gait effective video data set is used for screening the collected human body gait effective video so as to reserve a video data set which meets the network training requirement and is called as an effective video data set.
5. The third neural network is constructed and trained in an offline construction and training mode, so that the timeliness of system application can be guaranteed.
6. The gait detection method outputs the gait information of the basic identity characteristic information of the human body, and simultaneously detects the person matched with the gait information in the provided video.
Embodiment 2 footprint identification and gait detection method based on deep learning
Referring to fig. 3 and 5, the footprint identification and gait detection method based on deep learning in the embodiment includes the following steps:
1. firstly, a human body footprint image data set is acquired by collecting public security organization data, web crawlers, video screenshot labels and other modes.
2. And eliminating the data which do not meet the network training requirement in the data set, and sorting out effective images.
3. The characteristics of the valid data are in one-to-one correspondence with the tags, and the identity characteristic information comprises but is not limited to height, weight, age, gender, types of shoes worn, whether the legs have diseases or not, whether the soles have particularity and the like.
4. Storing the processed effective data as a human body footprint image effective data set, and meanwhile, according to the following steps of 6: 2: 2, the effective data set of the human body footprint image is divided into a training set, a cross validation set and a test set according to the proportion, so that the subsequent use is facilitated.
5. And constructing a first neural network according to the structural schematic diagram of the neural network constructed by the method shown in FIG. 6, training the first neural network by using a training set of the human body footprint valid data set, and selecting a model by using an inspection verification set.
6. Testing the trained new first neural network on the test set, and if the accuracy reaches a set threshold, taking the first neural network as a second neural network; if the accuracy rate does not reach the set threshold value, the network training is repeated until the accuracy rate reaches the set threshold value.
7. The footprint image to be identified is input to the second neural network, outputting the footprint owner's basic identity characteristics including, but not limited to, height, weight, age, gender, type of shoes worn, whether there is a disease in the legs, whether there is a specificity in the soles of the feet, etc.
8. Firstly, a human gait video data set is collected by collecting public security organization data, web crawlers, video annotation and other modes.
9. And eliminating the data which do not meet the network training requirement in the data set, and sorting out effective videos.
10. The characteristics of the valid data are in one-to-one correspondence with the tags, and the identity characteristic information comprises but is not limited to height, weight, age, gender, types of shoes worn, whether the legs have diseases or not, whether the soles have particularity and the like.
11. Storing the processed effective data as an effective data set of human gait video, and meanwhile, according to the following steps of 6: 2: 2, the effective data set of the human body footprint image is divided into a training set, a cross validation set and a test set according to the proportion, so that the subsequent use is facilitated.
12. And constructing a third neural network according to the structural schematic diagram of the neural network constructed in the figure 6, training the third neural network by using a training set of a human gait effective data set, and selecting a model by using a checking and verifying set.
13. Testing the trained new third neural network on the test set, and if the accuracy reaches a set threshold, taking the new third neural network as a fourth neural network; if the accuracy rate does not reach the set threshold value, the network training is repeated until the accuracy rate reaches the set threshold value.
14. And inputting the basic human identity characteristics to be detected (including but not limited to height, weight, age, sex, type of shoes worn, whether the legs are diseased or not, whether the soles have particularity or not and the like) into a fourth neural network, and outputting basic gait information of the human identity characteristic owner.
15. A person matching gait information is detected in a video provided.
Embodiment 3 footprint identification and gait detection method based on deep learning
Referring to fig. 1, the footprint identification and gait detection method based on deep learning in the present embodiment includes the following steps:
1. the construction and training of the network are consistent with those of examples 1-2.
2. The input of the second neural network is a human body footprint image needing to be identified through external input, the output of the second neural network is basic identity characteristic information (including but not limited to height, weight, age, sex, type of shoes worn, whether diseases exist in legs, whether specialties exist in soles and the like) of a footprint owner, and meanwhile the basic identity characteristic information is input into the fourth neural network.
3. The fourth neural network inputs the footprint owner basic identity characteristic information (including but not limited to height, weight, age, sex, type of shoes worn, whether the legs are ill, whether the soles are special, and the like) output by the second neural network, and the output of the fourth neural network is the identity owner basic gait information.
Example 4 method of constructing neural network
Referring to fig. 6, the neural network construction method in the above embodiments includes the following steps:
1. the network structure comprises an input layer, a hidden layer and an output layer. The hidden layer comprises a data preprocessing layer, a convolution layer, an activation function layer, a pooling layer, a full-connection layer and an anti-overfitting layer. Wherein the over-fitting prevention adopts an L2 regularization method.
2. The network adopts forward propagation to process data, and adopts backward propagation to continuously update network parameters, wherein an L2 regularization method is used in the backward propagation process to prevent overfitting of the network.
3. The data preprocessing layer performs data preprocessing on the effective data set, including data normalization processing and the like.
4. After data preprocessing, transmitting training set data into a first convolution layer for convolution operation; the first volume of base layer output is transmitted into a first activation function layer, and an activation function is used for operation processing; the first activation function layer output is passed into the first pooling layer, where the data is pooled using maximum pooling or average pooling.
5. After the first pooling layer is processed, the data are transmitted into a second convolution layer for convolution operation; the output of the base layer of the second volume is transmitted into a second activation function layer, and the activation function is used for operation processing; the second activation function layer output is transmitted into a second pooling layer, and data is pooled using maximum pooling or average pooling.
6. The data processed by the second pooling layer is transmitted into a third convolution layer for convolution operation; the output of the base layer of the third volume is transmitted into a third activation function layer, and the activation function is used for operation processing; and the output of the third activation function layer is transmitted into a third pooling layer, and the data is subjected to pooling processing by using maximum pooling or average pooling.
7. And continuously performing convolution, activation and pooling operations until the Nth pooling layer.
8. And the Nth pooling layer data is transmitted into the first full-connection layer, and the first full-connection layer output is transmitted into the second full-connection layer until the Nth pooling layer data is transmitted to the Mth full-connection layer.
9. And transmitting the cross validation set into the model to obtain an optimal model, wherein the optimal model comprises the determination of the parameters N and M.
10. Testing the trained network model by using a test set, and if the accuracy reaches a set threshold, accepting the network model; if the accuracy rate does not reach the set threshold value, the steps are repeated to carry out network training.
Embodiment 5 footprint identification and gait detection device based on deep learning
Referring to fig. 7, the footprint identification and gait detection device based on deep learning includes:
the system comprises an image acquisition module 40, a network construction module 10, a network training module 20, a network module 30 and a detection module 60. In addition, the human body footprint image effective data set and the human body gait video effective data set are both stored in a local computer.
Both the network building module 10 and the network training module 20 use offline startup because the operation of the two modules requires a valid data set and the network building and training process takes a long time, which does not meet the real-time requirement of online use.
The network constructing module 10 includes a first network constructing module 11 and a second network constructing module 12, which are respectively used for constructing a first neural network and a third neural network. The network training module 20 includes a first network training module 21 and a second network training module 22, which are respectively used for training the first neural network and the third neural network and respectively outputting the second neural network and the fourth neural network.
The network module 30 includes a first neural network module 31 and a second neural network module 32, which respectively correspond to the trained second neural network and the trained fourth neural network.
The image acquisition module 40 processes the footprint image to be identified, and transmits the processed data to the second neural network, and the second neural network outputs the basic identity characteristic information (including but not limited to height, weight, age, sex, type of shoes worn, whether the legs have diseases or not, whether the soles have particularity or not, and the like) of the footprint owner through calculation; the basic identity characteristic information is used as input and transmitted into a fourth neural network, and the fourth neural network outputs basic gait information of an identity characteristic owner through calculation.
The basic gait information is transmitted to the detection module 60, and the detection module 60 detects the person matched with the basic gait information in the provided external video to be detected.
Although the present application has been described with reference to a preferred embodiment, it should be understood that various changes, substitutions and alterations can be made herein without departing from the spirit and scope of the application.
Claims (10)
1. A footprint identification and gait detection method based on deep learning is characterized by comprising the following steps:
step S100: identifying a human body footprint image to be identified and outputting basic identity characteristic information of a footprint owner;
step S200: inputting the basic identity characteristic information of the footprint owner into a fourth neural network trained by adopting a human body gait effective video data set, and outputting the gait information of the human body basic identity characteristic;
step S300: and detecting a human body image matched with the gait information in the video to be detected according to the gait information of the basic identity characteristics of the human body.
2. The deep learning based footprint identification and gait detection method according to claim 1, wherein the step of "identifying" in the step S100 is: the method comprises the steps of collecting a human body footprint effective image data set, constructing a first neural network according to a preset network structure, training the first neural network by adopting the human body footprint effective image data set, obtaining a second neural network, inputting a human body footprint image to be identified into the second neural network, and outputting basic identity characteristic information of a footprint owner.
3. The deep learning based footprint identification and gait detection method according to claim 1, characterized in that the training of the fourth neural network comprises the steps of: the method comprises the steps of collecting a human body gait effective video data set, constructing a third neural network according to a preset network structure, and training the third neural network by adopting the human body gait effective video data set to obtain a fourth neural network.
4. The deep learning based footprint identification and gait detection method according to claim 2, wherein the step S100 comprises the steps of:
step S110: collecting human body footprint images and storing the human body footprint images into a human body footprint image data set in a gathering manner;
step S120: sorting the effective images in the human body footprint image dataset;
step S130: and after the labels are added to the effective images, storing the labeled images into the human body footprint effective image data set.
5. The deep learning based footprint identification and gait detection method according to claim 2, wherein the step S100 comprises the steps of:
step S140: inputting the effective image data set of the human body footprint image into the first neural network, and then carrying out forward propagation step by step and backward propagation step by step to obtain a training image neural network;
step S150: testing whether the accuracy of the recognition result of the training image neural network reaches a threshold value;
step S160: if the judgment result is yes, outputting the training image neural network as a second neural network, and if the judgment result is no, repeating the steps S140-S150.
6. The deep learning based footprint identification and gait detection method according to claim 1, wherein the step S200 comprises the following steps:
step S210: collecting human gait videos and storing the human gait videos as a human gait video data set in a gathering manner;
step S220: sorting the effective videos in the human body gait video data set;
step S230: and after the effective videos are added with the tags, storing the tagged videos as the human gait effective video data set.
7. The deep learning based footprint identification and gait detection method according to claim 3, wherein the step S200 comprises the steps of:
step S240: inputting the human body gait effective video data set into the third neural network, and then carrying out forward propagation step by step and backward propagation step by step to obtain a training video neural network;
step S250: testing whether the accuracy of the recognition result of the training video neural network reaches a threshold value;
step S260: if the judgment result is yes, outputting the training video neural network as a fourth neural network, and if the judgment result is no, repeating the steps S240-S250.
8. A deep learning based footprint identification and gait detection device for use in a method according to any of claims 1 to 7, comprising:
a network construction module: the first neural network and the third neural network are constructed according to a preset network structure;
a network training module: for training the first neural network with a human footprint valid image dataset, obtaining a second neural network,
the human body gait effective video data set is used for training the third neural network to obtain a fourth neural network;
a network module: the second neural network is used for recognizing the human body footprint image to be recognized and outputting the basic identity characteristic information of the footprint owner to the fourth neural network, and the fourth neural network outputs the gait information of the basic identity characteristic of the human body;
a detection module: the method is used for detecting the human body image matched with the gait information in the video to be detected according to the gait information of the basic identity characteristics of the human body.
9. The deep learning based footprint identification and gait detection device according to claim 8, wherein the network building module comprises a first network building module and a second network building module,
the first network construction module is used for constructing a first neural network according to a preset network structure;
the second network construction module is used for constructing a third neural network according to a preset network structure.
10. The deep learning based footprint identification and gait detection device according to claim 8, wherein the network training module comprises: a first network training module and a second network training module,
the first network training module is used for training the first neural network by adopting a human body footprint effective image data set to obtain a second neural network;
the second network training module is used for training the third neural network by adopting the human body gait effective video data set to obtain a fourth neural network.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201811191596.2A CN109359580B (en) | 2018-10-12 | 2018-10-12 | Footprint identification and gait detection method and device based on deep learning |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201811191596.2A CN109359580B (en) | 2018-10-12 | 2018-10-12 | Footprint identification and gait detection method and device based on deep learning |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN109359580A CN109359580A (en) | 2019-02-19 |
| CN109359580B true CN109359580B (en) | 2021-05-11 |
Family
ID=65349151
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201811191596.2A Active CN109359580B (en) | 2018-10-12 | 2018-10-12 | Footprint identification and gait detection method and device based on deep learning |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN109359580B (en) |
Families Citing this family (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN110992397B (en) * | 2019-10-21 | 2023-07-21 | 浙江大华技术股份有限公司 | Personnel access track tracking method, system, computer equipment and storage medium |
| CN111177446B (en) * | 2019-12-12 | 2023-04-25 | 苏州科技大学 | Method for searching footprint image |
Citations (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN103093234A (en) * | 2012-12-21 | 2013-05-08 | 中国科学院合肥物质科学研究院 | Identity recognition method based on ground reactive force during walking |
| WO2013113769A1 (en) * | 2012-01-30 | 2013-08-08 | Lion Systems S.A. | Method and device for evaluating a contact area between a foot and a surface |
| CN103473539A (en) * | 2013-09-23 | 2013-12-25 | 智慧城市系统服务（中国）有限公司 | Gait recognition method and device |
Family Cites Families (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US8152694B2 (en) * | 2009-03-16 | 2012-04-10 | Robert Bosch Gmbh | Activity monitoring device and method |
| US9846805B2 (en) * | 2016-03-16 | 2017-12-19 | Sportsmedia Technology Corporation | Tracking of handheld sporting implements using computer vision |
-
2018
- 2018-10-12 CN CN201811191596.2A patent/CN109359580B/en active Active
Patent Citations (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| WO2013113769A1 (en) * | 2012-01-30 | 2013-08-08 | Lion Systems S.A. | Method and device for evaluating a contact area between a foot and a surface |
| CN103093234A (en) * | 2012-12-21 | 2013-05-08 | 中国科学院合肥物质科学研究院 | Identity recognition method based on ground reactive force during walking |
| CN103473539A (en) * | 2013-09-23 | 2013-12-25 | 智慧城市系统服务（中国）有限公司 | Gait recognition method and device |
Also Published As
| Publication number | Publication date |
|---|---|
| CN109359580A (en) | 2019-02-19 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN109685102B (en) | Chest focus image classification method, device, computer equipment and storage medium | |
| CN110516161B (en) | Recommendation method and device | |
| CN108806792B (en) | Deep learning face diagnosis system | |
| CN114727766A (en) | System for collecting and identifying skin conditions from images and expert knowledge | |
| CN107292330B (en) | Iterative label noise identification algorithm based on double information of supervised learning and semi-supervised learning | |
| JP2022031730A (en) | System and method for modeling probability distribution | |
| CN109346159B (en) | Case image classification method, device, computer equipment and storage medium | |
| CN112364174A (en) | Patient medical record similarity evaluation method and system based on knowledge graph | |
| CN105139237A (en) | Information push method and apparatus | |
| CN113723312B (en) | Rice disease identification method based on visual transducer | |
| CN109359580B (en) | Footprint identification and gait detection method and device based on deep learning | |
| CN111192660A (en) | Image report analysis method, equipment and computer storage medium | |
| CN110992988A (en) | Speech emotion recognition method and device based on domain confrontation | |
| Vajdi et al. | Human gait database for normal walk collected by smartphone accelerometer | |
| CN115762721A (en) | Medical image quality control method and system based on computer vision technology | |
| CN116312915A (en) | Method and system for standardized association of drug terms in electronic medical records | |
| CN114662477A (en) | Stop word list generating method and device based on traditional Chinese medicine conversation and storage medium | |
| CN113171082A (en) | Method and device for evaluating injury of anterior cruciate ligament | |
| CN108846327B (en) | Intelligent system and method for distinguishing pigmented nevus and melanoma | |
| CN116630062A (en) | Medical insurance fraud detection method, system and storage medium | |
| CN113887335B (en) | Fall risk real-time evaluation system and method based on multi-scale space-time hierarchical network | |
| Rachna et al. | A Robust Machine Learning Model for Breast Cancer Prediction | |
| CN114003684A (en) | Medical information relation prediction method and system based on open world assumption | |
| Suganthi et al. | Performance Analysis of Machine Learning Classification for Parkinson's Disease from Biomedical Audio Data | |
| Chen et al. | Giant panda face recognition based on PandaFaceNet |
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 | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |