If Cisco, the global technology focus around 2000, represents the glory of the Internet for 10 years, then this year's fire of NVIDIA may represent the AI ​​era just opened. Cisco provides the infrastructure for the Internet, and NVIDIA provides the driving force for AI. Since the 1970s, AI has been ups and downs. In the early days, it has been a high level in the academic world. It has never entered the people's side. Nowadays, with the help of algorithms, chips and data, artificial intelligence is booming, especially in the industry. The perception of the masses is also two days: some people are worried that artificial intelligence will threaten humanity in the future. Some people think that many of the current manifestations of artificial intelligence are "artificial mental retardation", which is still not worth mentioning.
Objectively speaking, at present, artificial intelligence is still in its infancy, and it is true that the products are mixed, and all kinds of pseudo-intelligence propaganda are falling, but the actual effect is disappointing. The artificial intelligence form is not necessarily a sci-fi style, humanoid, arm-and-legged robot, but is probably an input method, a takeaway application, a taxi software, or even a news push APP, etc., all of which are artificial intelligence. Artificial intelligence is not a specific industry, but a technology-driven, can penetrate into various industries, the so-called +AI. However, at present, the hot surface of artificial intelligence is still in the fields of automatic driving and face recognition.
It is said that about five years ago, there was a figure investigation team inside the public security system in China - a sub-sector specializing in image investigation in the Public Security Bureau. The first-tier city map investigation brigade may have a scale of 300-500 people, which is the larger of the entire public security bureau. Departments, along with the surge in the number of urban cameras, the number may expand; but with the application of AI technology in the field of security monitoring, the investment of graphics and surveillance may focus on technology and facilities in the future, and personnel are expected to decrease.
At present, the more mature applications in the field of image detection are “car license platesâ€, the license plate recognition is mature, data collision, trajectory analysis, intelligent control and other means have repeatedly made meritorious deeds. Many people complain that they have caught one license plate and that the trafficker has not caught a few. In fact, the license plate recognition technology is mature, the recognition rate below 160 miles is more than 90%; the second car is driving along the road, and the control is easy; finally, the rules of vehicle violation are easy to set. If the suspect is carrying a 440*140 sign on the road, it is actually very easy to catch. It's a pity that they don't take the usual roads and wear hoods with glasses, so even if they appear occasionally under the camera of law and order, it is not easy to identify.
Back to the question of the public security classmates, this is a very good question, and it also represents the progress of face recognition in the field of actual combat. At present, more and more airports, ports, stations, and subways are beginning to deploy face recognition. In places where people are crowded, it is more appropriate to deploy a face recognition environment. The indoor light is good, the channel range is small, the camera angle is appropriate, and the target distance is Nearly, and both are high-level industries, the infrastructure is better (mostly HD, network cameras). For sudden or emergency investigations, in 1:N (static mode) recognition mode, or M (small): N (dynamic mode), then based on the current face recognition false positive rate (by 97%) and server calculation Volume, system load and false positive rate are acceptable, but the efficiency of the case and the valuable intelligence information provided are very large.
The advantage of face recognition for control is that it is not based on other features such as documents that are easy to copy or tamper with.
At present, there are more than 1,000 stations in the Beijing subway. The number of cameras in the early subway stations was small, and it gradually increased. The early days were analog cameras. Later, MPEG-2/4 encoding, and later H.264 HD, the current average of 100 cameras per station is of. In fact, in the pit stop, gates and other "fortress" to upgrade or deploy a face recognition camera, it is not possible to calculate by 100 cameras, so the total number of face recognition cameras is not so much, M: N mode The amount of computing power and the number of errors is not so horrible, and will not cause the face recognition system to crash.
In addition to street cameras, street cameras are more widely used in "wide-area, large-scale surveillance and patrol". With the current face recognition technology, outdoor light, angle, distance, and face pixels (80*80 pixels are ideal) The requirements are still relatively strict. Based on the significance and value of outdoor large-scale inspections, AI manufacturers have conducted targeted research and development based on this pain point. It is said that the 50-meter range recognizes human faces as objects. The starting point and market space of this product is very good, but it has not yet landed, no feedback, and the value of the monomer should be high.
Back to the security industry, AI-based security monitoring is by no means a few years of "intelligent analysis" era of group magic dance, the future can really dare to promote their own "AI + security" may be nothing more than three or five. AI research and development, the need for GPU clusters, the latest Tesla V100 high-performance computing card cost millions, AI R & D personnel annual salary of millions, AI algorithm needs hundreds of billions of video massive data "feeding", ask several security companies can afford in this way? Four core competitive barriers for security artificial intelligence: algorithmic capabilities, data capabilities, productization capabilities, and channel capabilities.
In terms of AI technology, some emerging algorithm companies have certain advantages, such as contempt, business soup, cloud and so on, but in the security industry accumulation and data, far less than the old security companies such as Haikang, Netpower, Dahua, Yushi , Kodak and other companies. AI from the laboratory scientists, to the industry, the algorithm is one, data capabilities, product capabilities and channel capabilities, but also a test. The AI ​​curtain has just opened, the security industry as the leading force, carrying AI heavy fire, the actual combat power to enhance the geometry, wait and see!
'
400W Carbon Fiber Heater,Carbon Fibre Patio Heater,Carbon Fibre Room Heater,Carbon Fibre Radiant Heater
Foshan Shunde Josintech Electrical Appliance Technology Co.,Ltd , https://www.josintech.com