Home > News content

Liu Jun, former vice president of Baidu, founded wide bench technology, crowdsourcing pure vision scheme to join high precision map war

via:博客园     time:2018/3/8 7:27:15     readed:535

The day before, the high precision map company wide stool technology announced the completion of several hundred million yuan A round of financing. This round of financing by IDG capital investment, capital, Lancang capital and other investment, Huaxing Alpha as the sole financial adviser to the financing.

Since then, the wide stool has officially announced its entry into the field of automatic driving high precision map, and joins the high-precision map in the mass production war.

In order to have a more comprehensive understanding of the way to start the business of wide stool technology, Xin Zhi visited Liu Jun, founder and CEO, and explored the layout of wide stool technology in high-precision map.


* the founder of the wide stool technology and CEO

Let us first take a look at who is Liu Jun.

Liu Jun was the vice president of the Google China Institute of engineering and the chief engineer of the global engineering technology and the chief scientist of the people's search.

In May 2010, Liu Jun founded the first socialized search engine cloud network.

In September 16, 2013, Liu Jun formally joined Baidu as vice president of the Baidu Inc and was responsible for the operation of the Baidu strategic and Technical Committee.

During his job in Baidu, Liu Jun, as general manager of Baidu LBS business department, said that the market share of the Baidu map was 70%.

In 2015, the internal structure of Baidu was adjusted, and the LBS division was divided into Baidu takeout and operation help. Liu Jun is the chairman of the Baidu takeout.

In 2018, again, in the public view, Liu Jun became the CEO of the wide stool technology, and what kind of company was the wide stool technology?

Liu Jun told the new intelligence drive, we are a technology venture with a high level of high precision maps, which mainly provide the underlying data services for the auto driving industry. The core members have more than 40 people, mostly from Internet companies such as Google, BAT, Gao De and so on.

Liu Jun said that the strategic layout was clearly defined from the first day of the business:Do not build cars, do not do car networking operating system; do only high-precision maps, named

As everyone knows, the current domestic large listed companies take the lead in the layout of high precision automatic driving map market, Baidu, Tencent, Gao De, NavInfo in the past few months, have announced a strategy for the development of high precision map for the autopilot application. Moreover, the market of high precision map is a market with very fierce competition and very high threshold.

Breakthrough point of automatic driving pain point

At present, autopilot is bringing great changes to traffic travel, but the first problem facing autopilot is complex traffic environment. As the basic link of autopilot, map service has become one of the pain points in the industry.

Liu Jun, who has a lot of experience in maps, chooses to cut into the high precision map, and aims to make a large-scale high-precision map company.

Liu Jun said that the high-precision map matching sensor is the most effective way to deal with complex traffic environment. The company will focus on Intelligent high-precision map system development, large-scale production and other aspects, and will cooperate with partners to explore a variety of cooperation modes.

We only need to do high-precision map, from strategic choice, we hope to cooperate with everyone in the auto driving industry chain, and form partnerships without competition, Liu Jun said.

According to Lei Feng's new intellectual drive, high precision maps are divided into three parts: recognition, measurement and networking.

It is worth noticing that there are great difficulties in networking. From the perspective of talent reserve, talents in building 3D maps are scarce. China may not exceed 100.

In addition, there is a big challenge for high precision maps.Making the first map is only 10% of the work, and the rest of the 90% energy is constantly updating the map.

Therefore, in the process of realizing automatic driving, promoting the high-accuracy of map data, effective and low-cost data collection, automatic processing and real-time updating of map data have become the strategic direction for all the big businessmen to choose first.

Just last month, BMW and Daimler agreed to work together to make HERE high - precision real-time maps. HERE high precision real-time map is a cloud based service with high accuracy and can be updated continuously through multiple data sources.

At the beginning of last year, the joint Tencent shares NavInfo HERE map, will jointly set up for automatic driving of high precision map services. In addition, they will also develop and provide location technology and services for the Chinese market. At the same time, 2017 years, Gao de completed 320 thousand kilometers of high-precision map data collection, and also built an automated online collection process based on crowdsourcing, big data processing and cloud computing platform.

The newly established wide stool near understand technology, real-time update is used to focus on automatic driving of high precision map location, so at the beginning, Liu Jun said most of the energy input to update the map, from the increase of test vehicles, more processing huge amount of data and the cloud service, and semantic high precision and wide map further upgrade the development direction of stool.

As a result, the map dealers have selected the layout of the map updating, hoping to gain the initiative in the scramble for high-precision maps.

Catch up with the high precision map enterprise in the market

Because of such fierce competition, the wide bench has begun to cooperate with many car factories and automatic driving solutions businesses after the initial development of products. At present, there has been a willingness to cooperate with other enterprises invested by IDG capital. The following will be released in the car with L3 level automatic driving function, using the wide stool and high-precision map scheme.

Mr Lau said:

Crowdsourcing, most of the high-precision map data collection is equipped with cameras, lidar, microwave radars, etc., but the high cost of equipment makes some technology companies slow progress.

The wide stool uses a pure vision technology to replace the laser radar, so that the collection of data is truly a commercial development model that can be replaced by the updated map.

Liu Jun explained that the large amount of data can be exchanged for high precision, by the same angle fast continuous shooting multiple photos, the formation of a large number of data, then use image processing algorithms to identify the independent research and development, the rapid processing of large amounts of image data. This contains a lot of calculation, and the amount of calculation is beyond imagination.

After the first step of recognition, there are more complex follow-up measurements and a network from two-dimensional to three-dimensional maps, and there are difficulties waiting for breakthroughs.

Although there are many technical barriers to high precision maps at present, some good results have been achieved, Liu Jundao.


* the practical road map of the wide stool

In fact, we use crowdsourcing to collect high vision map makers with pure vision, not only a wide stool company, but also a high-precision map company LVL5 from Tesla Andrew Kouri and Erik Reed.

The same reason, LVL5 does not use expensive laser radar and millimeter wave radar, but using a mobile camera to map.

However,Liu Jun stressed that it was used in surveying and mapping, and there were some differences from the LVL5's mobile camera.Lei Feng net new intelligence drive to know, wide stool technology high precision map accuracy of about 20 centimeters, and LVL5 map of 30-40 cm.

According to the new intellectual Lei Feng network drive understanding, high precision maps and map industry is still in a decentralized state, data acquisition characteristics of each enterprise have some differences, the industry at present there is no universal standard, and the company will not share any data, because they think that this is the core of important information.

For example, DeepMap CEO James Wu previously mentioned:

In this regard, Liu Jun said that it would provide data sharing based on high precision map positioning and must be done.

The appeal of the qualification of Surveying and mapping

Compared to most start-ups, autopilot giants have their own surveying and mapping systems. For example, Google has Google Maps, Google Earth, Google Street View, and navigation application Waze. Its mapping capability is also considered a key advantage for Waymo, and Waymo has completed an automatic driving test of more than 4 million miles.

In China, it is necessary to obtain the qualification of Surveying and mapping before it can be measured. It is reported that the "Regulations for surveying and mapping qualification management" will be divided into four grades: A, B, C and D, and those who need to do the surveying and mapping data collection and provide relevant information will be required to apply for the qualification certificate of Surveying and mapping to engage in the activities of the corresponding qualification range.

Li Shufu, the chairman of Geely, expressed the appeal of Surveying and mapping qualification. Under the premise of guaranteeing the information security of the state, we must open the mapping qualification to some automatic driving technology development enterprises conditionally and reduce the barrier of automatic driving technology development.

In this regard,Mr Lau stressed:Active applications are being made for other licences related to maps.

Finally, the high precision map of wide stool technology has already covered the provinces and cities in China. Liu Jun said that most of the roads in Beijing have been tested and completed since the establishment of the company.

Liu Jun further explained that when the company enters a new city, it needs to rebuild the map. Each city has a different driving rule, which means that the map needs to be constantly adjusted, and we have to adjust the software when we get into each area.

In a word, in 2018, the wide bench will first test and collect data from the domestic expressway, then further expand to other roads, and gradually complete the high-precision map version of the wide bench, and will also seek to cooperate with a large number of car factories and automatic driving enterprises.

China IT News APP

Download China IT News APP

Please rate this news

The average score will be displayed after you score.

Post comment

Do not see clearly? Click for a new code.

User comments