Vision System for Gait Recognition

When we're in a dangerous environment and face recognition is required for customs clearance,

Do you hesitate at the moment when you need to take off your mask

When dangerous people just around us,

Do you feel insecure

When the rioters, masked with caps and masks, fled,

Do we have a sense of justice to bring the mob to justice

Here he comes, no matter how you pretend, I know you




 

Research Background:

Face recognition and fingerprint recognition are widely used in biometrics, such as security and consumer electronics, but gait and iris recognition are rapidly emerging as "new" technologies that can identify people by how they walk.In layman's terms, I can recognize you without looking at your face.

Biometrics USES the body's inherent physical features (facial features, DNA, etc.) or behavioral features (gait, fencing habits, etc.) to be recognized by computers.From the point of the market performance of various identification technology, normal physiological characteristics recognition because of its characteristics, make its have certain defects in the process of application, for example, fingerprint and iris recognition has its own specific technology, it is difficult to integrate seamlessly into a number of factors, in particular, for ease of use, security and accuracy, the widely applied biological recognition technology is difficult to achieve both ways at the same time, high precision requires a high degree of the iris recognition technology, and a simple 2 d face and fingerprint identification may be not accurate enough.

The appearance of gait recognition technology makes up for the shortcomings of current fingerprint recognition and face recognition technology.Firstly, broad recognition range is the most obvious characteristic of gait recognition. In theory, as long as the camera's pixel is high enough, it can be identified at an infinite distance. The current gait recognition system of Milky Way water drop has been applied to the recognition from 50 meters away.In addition, the target recognition system does not need to coordinate with the gait recognition system to carry out specific actions, only need to walk through naturally, and can be identified from any Angle;Finally, different body shape, behavior habits and other characteristics determine the gait has a good discernibility.

All in all, gait recognition is a kind of control method, can greatly improve the work efficiency of a particular industry, and, as a new technology, gait recognition not only can be independently character identity recognition, and just can make up for the defect of fingerprints or face recognition technology, now has become a strong power in the field of biometric factors.

Research significance:

Gait recognition because of its "non-contact", make up for the defect of the other biometric technologies, enrich the application range of biometric market, the public security organ and the urban management department, the finance regulatory sensitive sites, major conferences and events security, smart home and criminal police detachments departments have important meaning, can greatly improve the working efficiency of the particular industry.For "Hong Kong riots" events, for example, the mob as masked, and can't identification by face to apply fingerprint identification is the play of the lips, the mob by wearing a mask or masked to avoid being caught by the police, and let the police, the gait recognition is one of the most damaging, but people in the invisible knowledge.

Main Research Contents:

Because the traditional gait recognition method relies heavily on the complex manual definition of features, and the designed features may contain various deficiencies and related problems, the gait features cannot be fully or accurately expressed.So this article will use the deep learning technology, build a network model of end-to-end convolution model to make independent, flexible learning math expression of the gait characteristics, at the same time, due to the gait recognition is extremely vulnerable to the interference of external factors, such as walking backpack, background changes, etc., aiming at this problem, this paper will be based on skeleton image gait feature extraction and use a variety of pooling method to optimize feature, USES triplet loss function to optimize the network parameters at the same time, in order to improve recognition accuracy and robustness.Specifically, the main research contents of this paper are as follows:

(1) Video acquisition and preprocessing.Ordinary camera is used to capture the gait video of pedestrians, and the video is converted into skeleton image set.The normalization method is used to preprocess the image set so that the position and size of different moving objects are the same.

(2) Gait feature extraction.The input of the network in this paper is a set of skeletal image sequences of gait. First, the frame-level features are extracted from each skeleton image.Second, all the frame level features aggregated into a feature vector (including a use of pooling method and using residual structure to improve the performance of sets pooling feature), in order to establish the connection between the multi-scale characteristic skeleton diagram (global features of a single frame image can complete representation of the gait), this article will extract the characteristics of different levels all input follow-up collection features;Finally, in order to establish the overall connection of the skeleton sequence, this paper synthesizes the global characteristics through the full connection network.Because the global gait sample is nearly infinite for the gait recognition system, the number of samples used for training is small, and the variation among the samples is small, this paper suggests that using the Triplet loss function to measure and optimize the network may be the best choice.

(3) Similarity calculation.The similarity between gait features is calculated and the gait prediction results are given.



★☆★☆★☆★☆★☆★☆★☆★☆★☆★☆★☆★☆★☆★☆★☆

About us
At-a-Glance

XR is a high-tech enterprise in China focusing on pushing the development of Industry 4.0 in various manufacturing industries(such as automotive,light industry, furniture,daily use...)
We are proud of our strength in providing our clients with the one-stop service line from R&D to the after-sale service that enable our clients to be automated,digitized and smart in their manufacturing sector.

Our vision is to grasp the future of manufacturing with core competitiveness.
Our mission is to make manufacturing simple.

We are aim to provide our clients with turnkey solutions for the intelligent integrated production system based on our core technical team and continuously creating value for our clients.

Our services
Customized service
We have a practised & professional R&D team fulfilled with vitality and innovative spirit.
90% of our engineers were graduated from the most famous technology universities and have abroad further study experience which allow XR has the abundant strength to offer the services as below:
1.  OEM/ODM
2.  Non-standard service
3.  Turnkey solutions(Just provide us with a simple product,we can offer you perfect solutions to the whole production line)


Lead time customized
Your efficiency is one of the most crucial points that we pay close attention to.
Shorten our lead time to satisfy your requirement is our duty.


Fast responses
We have the appointed team just works for your project which allows us to solve your questions on time.
7*24 hotline customer service


Excellence Quality Control
We have CMM system in house for inspection, every product have to be inspected before shipment.
1 year warranty and life-time technology support.





★☆★☆★☆★☆★☆★☆★☆★☆★☆★☆★☆★☆★☆★☆★☆



Core clients & partners





★☆★☆★☆★☆★☆★☆★☆★☆★☆★☆★☆★☆★☆★☆★☆

Certifications







★☆★☆★☆★☆★☆★☆★☆★☆★☆★☆★☆★☆★☆★☆★☆
 
Contact us
 
Contact person: Rachel


★☆★☆★☆★☆★☆★☆★☆★☆★☆★☆★☆★☆★☆★☆★☆

FAQ


Q:Are you a manufacture?
A:XR is not only a OEM manufacture, we have a compete R&D team that can offer turnkey solutions for our customers.  

Q:How can we be sure about  you are worth to be trusted?
A:You are more than welcome to visit our factory any time! It only takes 40mins drive from Guangzhou International Airport to our factory , We are pleased to help with the transportation and accommodation.

Q:How can we know the products we are going to receive are in good terms?
A:We will sent you the CMM inspection report, photos&videos once the product had been completed. Shipment will not be proceed unless you tell us to.

Q:What's your payment terms?
A:20% deposit in advance(contract signed)
    30% when the design is approved
    100% paid before shipment
    We prefer TT.

Q:Do you have any discount?
A:Please don't hesitate to contact us for further information.