Hunting cloud network (micro signal: [ilieyun]) in Shanghai reported in April 26th (
Computer vision is one of the important research areas in artificial intelligence, is the core of artificial intelligence system to interact with the outside world, the visual computing is becoming increasingly popular due primarily to the company for the depth distribution of VR/AR and machine learning, although more mentioned in artificial intelligence, machine learning, but only through continuous sample identification exercises and structural data feedback training can make the machine more and more intelligent.
Machine vision is trying to build from image acquisition to multidimensional image data analysis of structured data, and ultimately achieve the computer can like person as through visual observation to understand the world, have the ability to self adapt to the environment. To achieve this goal, the first is to image acquisition, the second is to image understanding. Different from the ordinary camera, the depth of the camera there is color, space and geometric sense. Now the ordinary camera can only get the color information and data, our common image is a collection of RGB information.
Such as borrow photos is the distant objects, flattened projection on a plane, but the human eye can judge the state of the picture according to the features of the objects, but will feature erase, the human eye is not able to distinguish. There is no geometric information in all the pictures and video information, so the core problem of visual recognition is the 3D information acquisition of the object. Depth sensor (depth camera) is designed to simultaneously obtain RGB information and XYZ spatial information.
Currently on the market mainstream depth detection technology for the structure of light technology line, the recognition distance, but easily affected by the sun, is the most mainstream of Machine Vision Engineering method. Generally speaking, the structured light system only needs a single camera, which is used to calibrate the space and to calculate the 3D space position by the coding of the laser spot array. For example, Microsoft with a sense of the game X-BOX and fame Kinect camera using the PrimeSense technology program, which is the first commercial structure of light technology to try. With the PrimesSense acquisition by apple and Intel realsense depth camera was also known, but forced to mix of X86 processors enable many customers complain incessantly. The other is a pure binocular program, super high precision recognition, but the recognition range is small, dual camera at the same time the difficulty of positioning the target object is larger, the representative of the product is LeapMotion.
Figure Yang technology innovation from the independent research and development of binocular active technology. By superimposing "structure light + binocular camera" technique to achieve depth calculation. Active binocular (Stereo Active) technology is a direction of the evolution of the depth camera technology. Figure Yang technology using the advantages of the scheme is in structured light system superimposed dual camera system has better depth map quality, better environment adaptability, and avoid the primesense patent. Although the Yang diagram depth camera in the algorithm, hardware cost and data complexity have increased, but the cost of Zheping say figure Yang more value in the main commercial applications for the technology brings the dividends. For example, the structure of the optical system in the room under the interference of light can be independent of the binocular camera, to ensure the integrity of the operation of the device.
In the specific business model, Yang technology is divided into three steps, the first is the sensor hardware products, based on can realize the environment awareness, sense, modeling, behavior recognition, in the business of domain diagram of Yang aimed at the intelligent security, environmental perception and AR/VR and robotics. According to Zhejiang revealed that the current figure has deep pupil and Yang Ling and another security of listed companies to carry out the project cooperation.
Followed by the product and service form, based on the depth of information on the visual application of software development, such as 3D modeling and SLAM, etc.. But for Zhejiang Ping Yang admitted that the figure want to do in the future is the general technology platform. Figure Yang client klaeng dark pupil had experienced PrimesSense was acquired after the stop supply crisis. "Israel's technological innovation, while strong that Israel will not be able to become an effective industrial providers, figure Yang wanted to become the master of aspects of the core technology platform technology companies, industry users provide stable and simple depth information interface and build the hardware, algorithms, and software open source platform."


Copyright belongs to the author.
Please contact the commercial hunting cloud network authorized, non commercial reproduced please indicate the source.
Author: Emily
Link: http://www.lieyunwang.com/archives/169947
Title: [first] map Yang Technology completed 15 million yuan of pre-A round of financing: