However, the LIDAR can only create the depth image of the plane, and the cost has remained high for many years and has not been widely used.With the development of vision algorithms and visual computing processors, the robot sweeping ushered in a visual recognition technology And AI algorithm era.
Visual navigation refers to the use of the camera to scan the surrounding environment, and then combined with infrared sensors, the use of mathematical operations and geometry, trigonometry mapping of the room in order to navigate, and before and after the image according to the location of each landmark to determine the current Move the route and update and adjust the environment model it builds.
Currently, only Dyson and iRobot are currently on the market using Vslam, which launched 360Eye in September 2014 and the latter launched Roomba 980 in September 2015. There have also been many startups doing Vslam solutions in China in recent years, including SanKoBot was founded in February 2016. Unlike other competitors, SanKoBot's Vslam program includes a host of innovative technologies including Vslam vision, object recognition, image processing, face recognition, Both independent innovation and industrial landing capabilities, providing artificial intelligence algorithms and the overall program.
Based on Vslam open source algorithm, SanKoBot highly optimizes most of the performance, for example, to achieve a series of difficult indoor scenarios that can be applied to low light, weak texture, light, and even rapid light changes, breaking the Vslam technology is highly dependent In the meantime, the SKB HOME AI brain equipped with the "SKB HOME AI Brain" will identify, classify and record the objects in the image and plan the optimal cleaning path to achieve automatic obstacle avoidance and improve the cleaning efficiency.
Robotics based on vslam + identification technology to achieve indoor map and object detection
In October last year, SanKoBot unveiled the monocular vision module in Hong Kong to help the family sweep the robot industry to upgrade and iterate. As the first company in the world to realize the Vslam algorithm in the ordinary chip (R16 without acceleration chip) Also laid the position of industry leader.
Monocular vision positioning module and the use of the visual sweeping machine
In November, SanKoBot became the world's first real-time indoor object recognition detection application that can be run simultaneously with Vslam in the field of sweeping robots, which draws an indoor map while performing recognition detection and recording the object position on the map Based on the end of real-time detection and Vslam, without cloud-based assistance or acceleration chip is based on the same camera and monocular.
In December, the module with path planning algorithm + Vslam visual positioning is fully covered, and the whole machine is in mass production. Until now, SanKoBot has supplied such well-known enterprises as TCL and other domestic and foreign enterprises.
Robot-based real-time identification technology (monocular)
In less than two years since its founding, SanKoBot can achieve such a brilliant result, which can not be separated from the mature team.Founder and CEO Li Lei has many years of domestic and foreign TMT industry experience and chip industry experience, has internationally renowned Israeli image Chief of Cortica China, an early investor in Silicon Valley and Israeli startups; CUI Jiewei, a PhD in computational neuroscience from the University of Maryland, is a former Numenta Fellow at Silicon Valley AI Corporation and owns 2 The United States of America AI core patents, won the William Hodos Award; algorithm director Wang Zijing former director of Panasonic Institute of Singapore algorithm, with 16 years of computer vision and AI research and development, industrial floor experience, the design of its face detection, human detection algorithm Used in Panasonic surveillance cameras and digital cameras.In addition, the technical and management teams are from United Technologies, UBM, SANY, ZTE and IWC.
In Lei's opinion, SanKoBot's technology in the device to achieve a similar function of human eyes and part of the brain, cleaning robot is only the first step in landing.
As visual Vslam's technology comes to an end, SanKoBot announced in December last year that it has partnered with a well-known domestic company to apply its visual recognition technology to the brain of the city, for example, when the number of cameras in the city is about 500,000, It's a massive amount of information that a mere 8,000 tons of data has been generated in just two weeks, and finding information about a car in such a huge amount of data for a certain period of time may take as many as 400 people a day, using traditional methods Can be found.But through the technology provided by SanKoBot, the car's color, brand, pedestrian and other data in the recording time to be refined and classified, stored in the cloud index.In its search for keywords, time periods and other information within a few minutes You can retrieve the results, which will greatly enhance the efficiency of urban management.
In June last year, SanKoBot completed 11 million angel rounds of financing, the national investment leader, Wang Lijie pre-Angel, Shenzhen Netac and Ding Tsui, IKEA and other investment with the same month access to Shanghai Zhangjiang Lingang government support and millions of smart manufacturing special Funds currently Shankai smart is a new round of financing for R & D investment and product commercialization.