In addition to Weibo, there is also WeChat
Please pay attention
WeChat public account
Shulou
2025-01-15 Update From: SLTechnology News&Howtos shulou NAV: SLTechnology News&Howtos > Internet Technology >
Share
Shulou(Shulou.com)06/03 Report--
From the ZTE incident in 2018, the Huawei incident in 2019 to the AI software ban just issued in 2020, the localized international situation has constantly focused our attention on those "technologically sensitive" areas.
This time, the bans are tied together by "AI" and "remote sensing".
Friends who are concerned about current affairs may already know that, according to US federal government documents, the export of certain types of geospatial imaging software is restricted, of which drones, self-driving and other areas are the most affected.
In the words of James Lewis, a technical expert at the US Center for International Strategic Studies, this is to "stop US companies from helping China make better artificial intelligence products to help its military power."
Wait? How can artificial intelligence and space remote sensing technology have anything to do with the military?
Not long ago, the United States demonstrated to the world the "new game" of AI remote sensing-the local supreme commander Sulejmani was accurately "beheaded", and the mission was not even carried out by a human, but by a Reaper drone. Operators thousands of miles away dropped four hellfires, boom--
Of course, in China and even in most countries, AI and remote sensing are more related to agriculture, surveying and mapping, exploration, mapping software and other applications. So will the ban on software exports bring a "semiconductor" crisis to this burgeoning field?
Making Swords into ploughs: the relationship between AI and remote Sensing
The word "remote sensing" must be no stranger to most of them. Whether they are workers who are active in the ocean, deep mountains and space, or ordinary people who start navigation and drones in cities every day, they can not do without the support of remote sensing technology.
The remote sensing monitoring all over the sea, land and air, and the resulting "spatial big data" also provides a good job for artificial intelligence.
In order to give you a more intuitive sense of the technical capabilities of "AI+ remote sensing", we might as well quote the recent achievements of the Reaper to interpret it.
Generally speaking, the application of AI in the field of remote sensing mainly plays three core roles:
1. High-intensity and real-time multi-source heterogeneous data processing capability.
With the process of industrial informatization in the past few decades, the field of remote sensing has basically completed digitization and visualization. This also leads to more diverse sources of remote sensing data, for example, when performing operations, it is often necessary to collect data from cameras, provide spatial information from Leo and high-orbit satellites, and distinguish a variety of light sources such as visible light and infrared spectrum. recognize different objects in a complex environment, and so on. The addition of artificial intelligence algorithm can help remote sensing system to improve a series of work such as data integration, analysis and decision-making, and complete the task more efficiently.
For example, the operator of the Reaper drone monitored the ground through the high-definition thermal imaging images transmitted by the satellite, and sent the pictures back to the ground station in real time through the satellite. After confirming that the target was in place, the operator quickly pressed the launch button and hit 100% of the car body. If the computing power of the UAV is not real-time, accurate and fast, it may not be able to complete the "harvest" in an instant.
(simulation test vehicle exercise of Reaper)
2. high obstacle avoidance and automatic operation ability to deal with complex weather and environment
When it comes to UAV, we know that autopilot technology depends on microwave radar observation, combined with satellite information for navigation and obstacle avoidance, but in the actual environment, we often encounter a variety of unexpected situations. For example, satellites are unlikely to observe weather phenomena within an hour, which will lead to bad weather such as wind and rain during UAV missions. Once the cloud coverage of the satellite image research area is more than 10%, it is difficult to extract high-quality data from it, or flying in complex areas such as forests, how to ensure that rapid changes are identified and evaluated?
All these need the support of artificial intelligence. For example, select certain characteristic parameters, train the weather prediction model, let artificial intelligence command and complete avoidance in advance; and, for example, classify and identify physical features through the model, and then recognize the development and changes of the environment. Achieve a higher level of automation in space navigation, obstacle avoidance and other aspects.
In fact, in the Reaper drone, we can see its ability to clear the air route, surveillance and reconnaissance, noise elimination and so on. Its high-definition thermal imaging camera and laser rangefinder / projector in its pod, coupled with an intelligent flight system, ensure that it can fly at an altitude of 25000 miles day and night. It can still accurately reach the operating area hundreds of miles away and carry out the mission smoothly.
3. Improve the ability of environmental monitoring, judgment and early warning in areas where manpower is difficult to reach.
Another value of artificial intelligence in the field of remote sensing is that by combining with sensors such as drones and unmanned vehicles, it can give terminal devices an intelligent "brain" to replace human beings to complete some previously impossible tasks.
This is not much reflected in the Reaper drones, since operators are supervised online in 24-hour shifts. However, the deployment of satellite remote sensing technology equipped with AI in advance in some inaccessible areas can greatly reduce manual operation and achieve emergency observation and early warning of major natural disasters.
For example, the professional disaster detection team of China Resources Satellite has provided 3203 emergency service data to more than 20 user units, including the Ministry of Land and Resources, the China Seismological Bureau, and the China Meteorological Administration, with an average emergency response time of no more than 1.5 hours. At a time when people are worried about months of wildfires in Australia, artificial intelligence may provide another way to prevent disasters in this sparsely populated country in the future.
Speaking of which, we may have get to the power of "AI+ remote sensing", which can be turned into smoke to destroy a city, or into showers to benefit life on earth. Casting swords into ploughs may be the right way to open technology.
AI embargo: will China's remote sensing industry "earthquake"?
After talking about the relationship between AI and remote sensing technology, it is time to talk about how much "blow" will the US embargo on AI software play to China's remote sensing industry.
At present, it seems that after going to the theatre and eating melons, everyone has gone back to their own homes and done what they should do. Neither did the "spare tire" be activated by the embargo like the semiconductor industry, nor did it condemn the inquiry at the top of its voice.
Is the effect of AI software on remote sensing not big enough?
First of all, the ban on software that automatically analyzes geospatial images will not bring about a chain reaction any time soon.
The main function of this kind of software is to train deep convolution neural network to automatically analyze geospatial images and point cloud geospatial images. For example, identifying targets such as vehicles and houses can reduce the pixel changes of the image during zooming, rotation and other operations.
(restrictions from the United States Bureau of Industry and Security)
Overall, the ban on AI software is not enough for everyone to fight.
The core reason, on the one hand, comes from the long tug-of-war caused by the embargo, and many interest groups are involved. For example, the ban will affect companies that use relevant software and develop aerial maps and 3D maps on the basis, because many of the finished software are based on platforms such as AWS and GCP that directly provide map processing API.
However, open source software on TensorFlow and pytorch is not affected, and relevant enterprises, platforms and communities can take positive actions to avoid this problem.
As the most direct example, in October 2018, the US Department of Commerce added eight Chinese enterprises, including Haikang, Dahua, iFLYTEK, Kuangshi, Shangtang and Yitu, to the entity list, stipulating that Nvidia, Intel and other companies are not allowed to sell chips to these companies. However, to this day, China is still an indispensable and important market for these enterprises.
Therefore, this is not the first time that the United States has restricted technology exports, nor can it be the last. Everyone is used to it, and their own industry has also developed. It can only be said that "Sichuan Jianguo" lives up to its reputation.
Of course, more importantly, the "blockade" of software may objectively cause a certain "delay" to the research and development of China's drones and self-driving enterprises. But first, there are many factors for space remote sensing software to integrate with artificial intelligence, such as the practical accumulation of band combination of high-resolution remote sensing images, exclusive chips and data sets for vertical fields, and so on, which will affect the accuracy and availability of AI in the field of remote sensing.
At the same time, China itself has strong early accumulation and R & D advantages in the field of AI algorithms, such as DJI DJI, a drone company with the best technology, Huawei, which makes efforts in side-side AI computing, Baidu, which has developed self-driving R & D capabilities, AI enterprises that have launched intelligent interpretation solutions for remote sensing images, and so on, and before the "soft power", the embargo is naturally less tense.
More importantly, the US AI software embargo may also "injure the enemy by 800 and self-damage by 1000." As we all know, the rapid development of artificial intelligence is inseparable from the open atmosphere and industrial environment, and software especially depends on the open source, trust and global communication environment.
Many American industry circles also need Chinese strength to participate in the co-construction, which is why, when the United States asked github to ban login to China, the other side even began to think of opening a subsidiary company in China.
After all, China, as a huge market for artificial intelligence research and development, applications, and industrial chain manufacturing, "invisible" to China also means "invisible" China, which will only accelerate the emergence of Chinese versions of Android, IOS, and GitHub.
This is why some netizens say that the embargo on spatial geography software, like the 2000 strong password export ban, is doomed to failure. Because-- you can stop a vendor with unique technology from providing prohibited hardware components, but you can't stop the spread of knowledge in the whole field!
Coincidentally, this time "knowledge" is in China, and there is no need for "export".
AI remote sensing: the next step is to enter the "no man's land".
Of course, this does not mean that China's intelligent remote sensing technology can rest easy.
Although a software embargo will not hurt the entire industry, on the whole, there is still much to be done in the field of remote sensing.
For example, AI reasoning algorithms in China's remote sensing field have made great progress, but proprietary reasoning chips are still limited, most of which still use general computing chips such as Nvidia, which leads to a certain efficiency gap in a specific domain architecture of DSA, and process accuracy also needs to be improved, thus affecting the research and development of cutting-edge remote sensing equipment such as Reaper drones.
In addition, the importance of data to intelligent remote sensing is mentioned above, and many remote sensing data in geographical space are provided cooperatively by satellite networks. This puts forward certain requirements for the aviation information industry to ensure the system construction of high spatial resolution, high time resolution, high spectral resolution and high score data, such as the recent formation of Beidou satellite network. it has important long-term significance for intelligent remote sensing. However, the remote sensing data with higher precision and complexity need to be processed with stronger computing power, which further requires the domestic semiconductor industry to continue to attack the difficulties.
In addition, remote sensing technology also depends on the overall technological upgrading of sensors, monitoring equipment, drones and other fields. For example, MQ-9 reapers have "almost no sound" during the flight, so that the target will not be detected in advance. At present, both the application of satellite remote sensing images and the technical threshold of artificial intelligence lack the help of relevant professionals, and the iterative speed of one-stop technology such as image acquisition, interpretation, analysis and training still needs to be improved.
All in all, when remote sensing changes from static to real-time, from fighting alone to deeply integrating with AI, we will see more innovations that benefit human society in the future. In the face of such a trend, what is needed is not a blindly optimistic "Ollie", but a concerted effort to open up the mountains and forests.
Welcome to subscribe "Shulou Technology Information " to get latest news, interesting things and hot topics in the IT industry, and controls the hottest and latest Internet news, technology news and IT industry trends.
Views: 0
*The comments in the above article only represent the author's personal views and do not represent the views and positions of this website. If you have more insights, please feel free to contribute and share.
Continue with the installation of the previous hadoop.First, install zookooper1. Decompress zookoope
"Every 5-10 years, there's a rare product, a really special, very unusual product that's the most un
© 2024 shulou.com SLNews company. All rights reserved.