40th Baidu technology salon notes

The topic of this technology salon is LBS technology. Although lbs applications have been gaining popularity in the industry in the past two years, I feel that there is still a certain distance from a wide range of applications.

The lecturers in this lecture are Zhang chuanming, Baidu positioning service director, Li Zhiwei, CTO of Momo, and Professor Deng Zhongliang of beiyou, it covers three aspects: technical implementation principle, practical application experience, and academic development frontier.

Zhang chuanming's lecture analyzed the current situation and prospects of Mobile Phone Positioning Technology, as well as the work done by Baidu Positioning Service API in solving the current weakness of positioning technology and some advanced practices in the industry to solve this problem:

The current positioning technology mainly consists of two aspects: Positioning Based on GPS Positioning navigation satellite technology and positioning based on mobile phone base stations and WiFi. Standard GPS Positioning methods have shortcomings in power consumption, geographic conditions, Canyon Effect, signal reflection, and indoor location failure. At the same time, mobile phone base stations and WiFi are also lacking in all aspects, baidu's lbs service mainly obtains relatively accurate positioning results through comprehensive analysis and processing of signal sources during various sensing periods. zhang chuanming's work not only uses EZ algorithm, maximum likelihood-based location estimation, rotation positioning, and other new analysis methods for signal source information, in addition, KNN and other data excavator learning methods are used for various signal sources to improve the overall utilization efficiency of signal sources.

Momo CTO focuses on two problems: one is the location offset of different coordinate systems in China, especially the random offset of the national coordinate system, they used their own sampling summary to correct the offset. the second problem is the rapid space retrieval technology. In this regard, MongoDB used by Momo adopts geohash, which will convert the longitude (longitude) and latitude (latitude) the value is converted to a base32-encoded string mona1 index. The Binary Index stored in the latitude and longitude dimensions is relatively rough and slow to pass. In this regard, the posygis spatial database on PostgreSQL supports the R-tree-over-gist tree index data structure. Geohash provides a better speed.

Professor Deng Zhongliang of beiyou is a master in academic circles. He has participated in many national basic projects and construction projects in this field. As an academic leader, he plays a huge role in chess, I would like to do something for the whole industry at the national level. At the same time, due to its official identity, it can better obtain and integrate various location information resources, it provides positioning information services at various levels for military, public institutions, and commercial use.