Reading as a god

Chapter 248 Mysterious Technology?
After Han Yanning was dismissed, Zhang Shan was ready to study again.

But at this moment, Zhang Shan suddenly remembered the prompt from the system——

"The host successfully triggers the 29th achievement of the system - passing on the torch, and a special gift package of mysterious technology is given to the host~"

Mysterious technology?
It's exciting to think about it~
Zhang Shan decisively chose to open the gift bag.

"The gift package is opening..."

"Congratulations to the host for obtaining an extraordinary biometric technology - gait recognition"

Zhang Shan:! ! !
Biometric technology is not a new concept. In the AI ​​era, more and more biometric features can be extracted and identified.

At present, there are many applications in the market, such as face recognition, fingerprint recognition, voiceprint recognition, etc.~
Although face recognition is not a bad street at this time point~
But it is almost on the eve of widespread commercial use.

Even iris recognition and palm vein recognition, which have a relatively high threshold, have been applied in individual scenarios.

However, due to the high threshold of gait recognition technology, it is almost hard to find a trace in the landing application.

Gait recognition is an emerging biometric recognition technology, which aims to identify people through their walking posture. Compared with other biometric technologies, gait recognition has the advantages of non-contact, long-distance and not easy to camouflage.In the field of intelligent video surveillance, it has more advantages than image recognition.

Gait refers to the way people walk, which is a complex behavioral characteristic.Criminals may dress themselves up so that not a single hair of their body falls off the scene of the crime, but there is one thing they can't control, and that's the way they walk.

Research by Professor Mark Nixon of the Department of Electronics and Computer Science at the University of Southampton in the United Kingdom has shown that everyone has a completely different walking posture, because people have different walking postures in terms of muscle strength, tendon and bone length, bone density, visual acuity, coordination, There are slight differences in experience, body weight, center of gravity, degree of muscle or bone damage, physiology, and individual walking "style".

It's very difficult for a person to disguise their gait, and whether criminals are wearing a mask and walking naturally toward a bank teller or fleeing a crime scene, their gait can give them away.

Human beings are very good at gait recognition, and have the experience of being able to identify familiar people based on their gait at a certain distance.

The input of gait recognition is a sequence of walking video images, so its data collection is similar to face recognition, and it is non-invasive and acceptable.

However, due to the large amount of data in sequence images, the computational complexity of gait recognition is relatively high, and it is also difficult to process.

Although a great deal of research work has been done on gait in biomechanics, research work on gait-based identification has just begun.

The main feature extracted in gait recognition is the movement of each joint of the human body.

It's easy to say, but difficult to implement~
At least so far, there are no commercial gait-based identification systems.

Of course, all this will now come to an end with the birth of this gift package.

Zhang Shan can feel the Versailles of the system. The construction of a mature gait recognition technology can measure its value with 4000 million yuan.

But the system chose to give this technology to Zhang Shan as a gift package worth 4000 million.

~~~
Although gait recognition has the advantages of non-contact, long-distance and not easy to camouflage~
But Zhang Shan can imagine the difficulty of commercial application of this technology!
But in the application, gait recognition needs to randomly sample a large amount of time-series gait data for model learning.Build a gait library, and then provide system solutions based on gait data.

When collecting data in the early stage of building a database, the positive recognition rate of gait recognition is low, and it is easily disturbed by gender, step length, rhythm, speed, etc.;

At the same time, camera angles, weather conditions, occluders, appendages, rough roads, and even clothing lighting will affect accuracy;
It is difficult to obtain information such as age and gender, and it is not universal, such as the disabled are not suitable for gait recognition technology.

Then, in the process of commercialization, most monitoring scenes are relatively complex, there may be multiple moving objects, and pedestrian images are easily affected by external factors such as weather and light, and change.

Taking the shadow of a moving target under sunlight conditions as an example, it may be connected to the detected target or may be separated from the target.If the former, the shadow distorts the object shape, making silhouette-based gait recognition methods unreliable; if the latter, the shadow may be mistaken for a wrong object in the scene.

Therefore, the challenge of gait detection in actual scenarios is not small.

The feature points extracted by gait recognition include two aspects. One is to extract static internal features, such as height, head shape, leg bones, joints, muscles and other physiological structures. These features are relatively stable. No major changes will occur;
The second aspect is to extract the dynamic characteristics of people, such as walking posture, arm swing, shaking head and shrugging, motor nerve sensitivity, etc. These dynamic characteristics are inseparable from physical attributes and will be relatively stable for a long time.However, the extraction and processing of these dynamic features are also more difficult.

Pedestrian re-identification relies on the overall posture of pedestrians for retrieval, mainly extracting static external features, such as clothing, backpacks, hairstyles, umbrellas, etc., and identifying people by analyzing pedestrians' clothing and posture.

However, these static external features are easy to change, such as changes in the overall appearance, changing clothes, hairstyles, hats, shoes, even camouflage posture, body shape, etc.

Therefore, it is also very troublesome when it comes to pedestrian re-identification~
Pedestrian re-identification needs to use a large number of image information of a single human body under different monitoring at the same time for model learning.For pedestrian recognition in a relatively short period of time in the monitoring scene, the accuracy is high, and the algorithm speed can reach the millisecond level.

In a real scene, the re-identification algorithm needs to achieve high-precision and fast recognition across time periods, scenes, and image acquisition devices with different imaging qualities, and the areas covered by these cameras do not overlap with each other, resulting in a lack of coherence Information.

Moreover, in different pictures, the posture, behavior and appearance of the characters will change greatly, and the lighting, background and occluders are different at different times and scenes, and there are often interferences from other characters with similar body shapes and clothes in the background; The resolution is also high and low, and the position of the characters in the picture is far and wide, all of which pose great challenges to the re-identification technology.

The changeable pose of pedestrians leads to the failure of the widely used alignment technology on human faces in re-identification.The difficulty of obtaining pedestrian data is far greater than that of face recognition data, and the complexity of pedestrian information is far greater than that of human faces. The combination of these two factors makes the research on re-identification algorithms more difficult and more important. .

Through the effective design of the algorithm, it is now the consensus of the industry to reduce the dependence on data to achieve a breakthrough in the re-identification effect.This not only puts higher requirements on the algorithm, but also requires more efficient chips to provide powerful computing power support. The lack of either of the two will affect the actual application value of pedestrian re-identification.

The improvement of the training dimension makes the applicability of gait recognition stronger, and its algorithm has more advantages in transfer learning.Other recognition methods may only need a few picture samples to start training the model, but for gait recognition, a large amount of video data is required to train the model.

Although he doesn't know much about this specific technology, Zhang Shan now has extraordinary programming skills.

The core of extraordinary programming ability is Zhang Shan's sensitivity to algorithms.

Zhang Shan can think of the complexity of the model training carried out by this technology. I am afraid that everyone used nearly a thousand pieces of video data~
Covering a variety of different clothing, multiple different scenes, multiple walking states, multiple different horizontal perspectives, multiple different vertical perspectives, and multiple attributes (including age, gender, ethnicity, height, nationality), where is it so easy to carry out Accurate humanoid segmentation.

Not to mention anything else, compared with other biometric recognition methods based on two-dimensional image training and learning, gait recognition is based on three-dimensional video training, and the amount of input data has increased by one dimension.

What does this mean? Accompanied by the change in dimension is a huge change in the level of understanding and the perspective of thinking.

……

But in any case, the system has successfully realized this technology, and even listed this technology as a novice technology in the system's value reference system~
Although in the eyes of the system, this technology is only a novice technology, but Zhang Shan does not dare to underestimate it.

After a new technology is born, there is always a question that always comes along, that is, what is it useful for?

The answer to this question often determines whether the technology is valuable and how valuable it is.

In the field of machine recognition, there are already various recognition methods such as face recognition, voice recognition, fingerprint recognition, and iris recognition.

However, compared with these recognition methods, many characteristics of gait recognition are irreplaceable by the above recognition methods and are unique.

Compared with other recognition methods, the recognition distance of gait recognition is longer, and the recognition distance can reach up to 50 meters under ordinary high-definition cameras; it belongs to uncontrolled recognition and does not require the active cooperation of the recognition object; A whole-body recognition technology, determined by body shape, head shape, muscle strength characteristics, motor nerve sensitivity, walking posture, etc., local changes will not greatly affect the recognition results.

These characteristics of gait recognition are destined to be unique in some application fields.

For example, relying on gait recognition to quickly retrieve target suspects.

According to the introduction of the system, Zhang Shan knew that using this technology to retrieve suspects based on gait information, the retrieval could be completed within 5 seconds at the fastest, which greatly changed the current situation of manual case handling before and saved a lot of manpower and time.

Moreover, the gait recognition technology is more "sharp-eyed" than ordinary investigators, and the efficiency and accuracy of recognizing people are higher.

In the field of medical rehabilitation: the recognition mechanism can be used to study the gait early warning mechanism to prevent the elderly from falling with medical institutions, and use gait recognition technology to monitor, analyze, organize and evaluate the long-term walking posture of the elderly, and compare multiple times and multiple states Under the influence of external factors on gait, the threshold of fall risk is calculated, and the risk level is divided.

In this way, a normalized gait information monitoring and management system for the elderly can be established to realize the need for early warning of falls and effectively reduce the probability of real falls for the elderly.In addition, other medical application scenarios such as leg postoperative rehabilitation are also being explored.

Not only in the medical field, but also in the pan-security field, this technology can show its talents.

Key security units including field oil wells, nuclear power plants, and large-scale water conservancy facilities can use gait recognition technology to prevent external intrusion. The solution is: first, the "gait" of the staff is pre-entered into the database, set as a white list, and marked as accessible , As long as external personnel who have not entered data enter the monitoring range, because there is no matching gait data in the system bottom database, the system will judge it as a "blacklist" and will call the police in time.

Of course, these are just the tip of the iceberg. The technology that the system gave Zhang Shan this time is not a simple algorithm or a mechanism.

It is a vast engineering system that integrates functions such as gait database building, gait recognition, gait retrieval, gait comparison, and large-scale tracking.

Even the construction plan system of the gait recognition has been tailored~
The gait database construction plan can be customized according to different needs to create a complete, fast and high-quality gait database construction plan.

It can complete the standardized collection, storage and management of the gait data of the target object, and can work with the soon-to-be-built face database and fingerprint database to form a multi-featured and holistic identification system in the field of public security.

With all these things, Zhang Shan can imagine that this technology will face success once it comes out~
The system is almost completely paved~
He even set up a technology company named Waterdrop in the name of Zhang Shan~
Good guy, the system will not be arranged by three bodies~
System: Don't compare me with that low-level galaxy~
Zhang Shan:? ? ?
Faced with the blow of dimensionality reduction, the existence of low dimensions can only accept the arrangement of fate.

There is no doubt that the gait technology is now on the market, which is completely a dimensionality reduction blow to other players in the security field.

What made Zhang Shan even more surprised was that after this technology was given to Zhang Shan, six or seven patents appeared in Zhang Shan's name.

There are even a few papers that I don’t know when they were published, and they are also included in Zhang Shan’s name~
Fortunately, these thesis topics are extremely common~
Without looking at these papers together, it is difficult to see the connection.

Even if you look at any one of them from the most acute perspective, you may find it unremarkable.

So far, Zhang Shan's hanging heart is finally in his stomach.

It is not scary to have a few papers appear inexplicably, after all, for a person with an 11-figure net worth like him.

There are many ways to water thesis.

But inexplicably, there are a lot of top-level papers, and Zhang Shan has a hundred mouths and can't tell clearly~
(End of this chapter)