Shulou(Shulou.com)06/02 Report--

As a general-purpose technology that can penetrate thousands of industries, AI can often make a technology that sounds "old" instantly coruscate a new imagination. For example, remote sensing is a technical term that Chinese people are not unfamiliar with.

The so-called remote sensing generally refers to the use of remote sensors to detect the radiation and reflection characteristics of electromagnetic waves of objects. Through far away from the target and non-contact way, to judge and identify the detection target. This technology is generally used in aerial platforms, such as satellites, aircraft, drones and so on.

100 years ago, the forerunner of remote sensing science was born in modern geography and surveying. In 1972, NASA launched the Earth Resources Technology Satellite ERTS-1 with remote sensors, which announced the official arrival of modern remote sensing technology.

This technology, which has helped humans understand the earth for decades, is today having an imaginative encounter with AI technology. But the combination of the two is not so easy. The breakthrough of end-to-side AI computing power and the deep learning of algorithm model are becoming the indispensable cornerstone of AI+ remote sensing industry and even the rising field of spatial intelligence.

For example, in June this year, the State key Laboratory of surveying and Mapping remote Sensing Information Engineering of Wuhan University and Huawei jointly held a competition for sparse representation and intelligent analysis of remote sensing images. In this competition, which aims to promote the development of theories and technologies related to "sparse representation and fusion of spatial information", contestants will use Huawei's new intelligent computing product Atlas 200 DK development kit to implement the inference calculation of the algorithm model on the corresponding remote sensing image test data set.

Today, we can take this opportunity to explore the value and demand of remote sensing encounter AI.

When remote sensing encounters AI, new opportunities are also new challenges.

Since the birth of modern remote sensing, with the help of the two bombs and one satellite project, China has been the core participant in remote sensing technology. From the 1970s to now, China has developed a perfect remote sensing related disciplines and diversified application fields. There are a large number of application practices and frontier explorations in the fields of environmental remote sensing, atmospheric remote sensing, resource remote sensing, marine remote sensing, geological remote sensing, agricultural remote sensing, forestry remote sensing and so on.

Today, the main opportunity and challenge that remote sensing technology faces is to make this technology universal, push it into various industries, and even let remote sensing into the vertical production cycle of industry, agriculture and forestry, and become a "front-line worker" in urban planning, disaster prevention and relief and other fields. There are many technical difficulties, but remote sensing science itself cannot be broken independently. For example, in the field of UAV remote sensing, the long cycle of image recognition and the need to consume a lot of time and manpower cost from data collection to application data has become one of the main restriction points for remote sensing to enter various industries.

This problem can be solved by AI.

We know that the current round of AI technology, represented by deep learning, has brought an important capability, that is, machine vision technology system. Among them, the capabilities of image recognition, image processing, dynamic recognition and image classification can just be used in remote sensing data. In an ideal state, automatic identification and reasoning from data to effective conclusion information can be realized.

At present, many AI companies and research institutions have participated in the fusion of AI+ remote sensing industry, and a large number of high-quality algorithms have been developed in the fields of automatic image processing, remote sensing data interpretation and so on. Last year, China's first remote sensing artificial intelligence application technology research center was established in Chongqing. At present, our country has launched the exploration layout of remote sensing + AI in the fields of agriculture, industry, road network, meteorology, water conservancy, construction and so on.

Generally speaking, the main understanding of "intelligent remote sensing" in the industry is that AI technology can provide active, real-time, automatic error correction image recognition and reasoning ability in the field of remote sensing.

For UAV remote sensing, which is closely related to vertical applications in various industries, AI can bring the following help:

1. Complete a large number of automatic identification work to realize the automatic processing of remote sensing data to available data.

2. Shorten the use process of remote sensing data, through automatic identification and automatic preprocessing, so that remote sensing results can be read and used in real time, thus making real-time remote sensing + operation linkage possible.

3. The algorithm is used to restore the image data to reduce the influence of environment and weather. UAV remote sensing has to face some influencing factors, such as complex terrain, rain and fog weather, which can be partially eliminated by the reduction of specific algorithms.

4. Reduce labor consumption, improve the efficiency of surveying and mapping, and achieve early warning remote sensing in some fields. For example, in Yunnan-Tibet, Qinghai-Tibet and other plateau highway sections in China, AI+ remote sensing technology has been used to actively detect debris flows and landslides.

However, Rome was not built in a day. Although the combination of remote sensing and AI can solve a lot of problems and build a new imagination. However, in the real combination, there will still be many industrial obstacles.

For example, the AI algorithm mentioned above can improve the ability of real-time automatic recognition of remote sensing images, but in practical application. Remote sensing satellites and drones are still unable to achieve the real-time effect of uploading data to the cloud, processing it in the data center and then transmitting it back. The behavior of uploading to the cloud limits the feasibility of applying AI remote sensing technology in the industry. For example, it is not suitable for a large amount of data to be uploaded to the cloud in many industries related to the national economy and people's livelihood.

To this end, the best solution is to allow end-to-side devices such as UAV and remote sensor to perform AI operation, data local preprocessing, and real-time close to the production process. However, the industrial reality is that at present, the vast majority of enterprises in the AI+ remote sensing industry mainly solve the algorithm problem. However, the algorithm problem needs effective network environment and computing environment to guarantee. Just like the best home appliances, it is useless without electricity.

So we can see that in the remote sensing system, the hardware level, especially the side-to-side AI computing power supply becomes very important.

On the other hand, although AI reduces the manpower consumption needed to identify remote sensing data, the complex hardware environment is likely to bring more AI technical personnel consumption. In the current industrial environment, the cost of AI hardware talents is greater. Therefore, in the hardware side to ensure the remote sensing system, the compatibility of AI capabilities and environmental availability is also an important issue.

Fortunately, side-to-side AI computing solutions have emerged in the industry. For example, Huawei's Atlas intelligent computing products have brought a surprising breakthrough to the remote sensing industry: let AI computing power fly in the air.

The AI computing power of flying

In the logic above, we describe the current situation: if remote sensing wants to enter various industries, it is necessary to strengthen the application of UAV remote sensing. The core problem is to make the UAV itself have relatively abundant AI computing power, so that image processing, image recognition, environment recognition and other related operations take place directly in the equipment. On the one hand, this avoids the security problems that may be caused by data uploading to the cloud, but also speeds up the processing speed and shortens the business process.

All in all, the most effective solution to this problem at present is to let the AI computing power fly and install AI chips in the UAV itself.

The acceleration ability of AI flying in the sky is not easy, which requires abundant AI computing power, multi-link visual data processing ability, and environmental adaptability in UAV scenarios.

Back to the competition held by the State key Laboratory of surveying and Mapping and remote Sensing Information Engineering of Wuhan University, the contestants will use the Atlas 200 DK AI developer kit to complete the reasoning calculation of the data algorithm model. Just imagine, if not only Huawei Atlas intelligent computing platform is introduced on the algorithm side, but also Huawei Atlas 200acceleration module (Atlas 200AI acceleration module) is installed on UAV equipment, which is an embedded AI module, which is mainly used for end-to-side AI acceleration of cameras, drones, robots and other hardware. In the field of video analysis, it can handle real-time analysis of 16-channel high-definition video. The matching Atlas 200 DK AI developer kit can build the development environment in 30 minutes, providing computing performance as high as 16TOPS INT8.) Remote sensing will get powerful AI computing power in both end-survey and algorithm side, which will greatly improve the work efficiency of remote sensing surveying and mapping.

Atlas 200can also be combined with Huawei's full-stack and full-scene AI capabilities and Atlas serialization products to achieve better compatibility in cloud integration and multi-device computing, effectively dealing with complex solution deployment and cloud integration applications that may be required by AI+ remote sensing.

AI computing power flying in the sky is only the tip of the intelligent remote sensing screen. Looking further ahead, we can see that the model of machine vision and the connection of everything represented by Atlas 200is effectively creating breakthrough opportunities for various industries. The intelligent iteration of all things inevitably starts with a chip.

The "core" of all things begins.

Taken together, when the Atlas 200s enter the UAV remote sensing space, we can see a breakthrough in several application scenarios:

1. The ability of real-time image recognition and image preprocessing is enhanced. Let the real-time analysis of agricultural disasters, disaster-resistant dispatching, as well as road conditions, mountain detection and early warning and other real-time application efficiency.

2. Based on the security guarantee of local computing, improve the terminal processing capacity, so that the remote sensing work of electric power, mines and power systems can better apply AI capabilities for remote sensing mapping, and improve the industrial application ability of remote sensing technology.

3. The high performance-to-price ratio and mainstream computing power of Atlas series products can better support the landing of real industrial projects, help the industry to complete low-cost, low-threshold deployment of AI remote sensing, and can avoid out-of-stock in large-scale deployment, build a more perfect and stable supply chain, so that agriculture and other long-tail industries that require extensive deployment of AI remote sensing can be applied.

Generally speaking, Atlas enables intelligent remote sensing to enter into thousands of industries on the hardware channel, so that intelligent remote sensing technology can achieve real-time, industrial grade, low threshold and adapt to the needs of various industries to a certain extent.

At the same time, it should be noted that the synthesis of these three points is to let the machine vision ability enter the industry with high efficiency, high security and low cost. At present, this capability is not only needed in the field of remote sensing, in a large number of industries that need to apply visual recognition, image and video processing, end-to-side AI acceleration is an essential industrial base.

For example, intelligent cameras in public places and traffic scenes, door face recognition solutions in intelligent parks, vision systems of navigation robots, and so on, all need to be fast and easy to obtain, which can efficiently support the end-to-side AI computing power of industrial applications.

The wisdom of all things should begin with the fact that all things have a "core". Atlas 200let AI calculate to fly in the air, remote sensing landing thousands of stories, may be regarded as a "formula". It is reused into today's countless industrial ports to form the true blueprint of Pratt & Whitney AI.

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.