Heterogeneous Computing Platform for Embedded Vision (ART/277CP)

Heterogeneous Computing Platform for Embedded Vision (ART/277CP)

Heterogeneous Computing Platform for Embedded Vision (ART/277CP)
29 / 04 / 2019 - 28 / 10 / 2020

Mr Louis Yat-cheung NGAI

1. Software engines with the following technologies: a) 3D image processing, SLAM and 3D mapping algorithms; b) Middleware level dimension measurement, positional tracking, spatial mapping with Android/Linux compatible APIs; c) A heterogeneous computing engine supports visual and depth sensing. 2. Reference design of demo applications for business engagement: a) “Indoor navigation” or “Enterprise AR” applications with the above software engines and a self-developed embedded system prototype. 3. Contract Service Deliverables for Customer 1 and Customer 2: a) AR software applications with SLAM and 3D mapping algorithms. (For Customer 1); b) 3D imaging industrial application software. (For Customer 2); c) Optimized software modules for final release package. (For Customers 1); d) Document of algorithms with UAT case, application programming interface(API) and user manual. (For Customers 1 and 2).

AIphotonics Limited (CS) [Sponsor]
CITIC Telecom International CPC Limited (CS) [Sponsor]
CITIC Telecom International CPC Limited (TL) [Sponsor]
FRONTIS Company Limited [Sponsor]

Abstract: Augmented Reality(AR) has become one of the top technological trends in the next decade. While most industry players focus on the AR headset hardware design and Apps development for gaming and infotainment, the application of Enterprise AR is still in the early stage. Many important aspects have not been systematically explored, such as seamless merging of digital objects with the reality to produce new environment and visualization for operational efficiency improvement, construction of a dense 3-dimensional mapping for the surroundings around the device, and accurate measurement of the real objects dimensions, etc.

In this proposal, we aim to develop a highly customized and optimized embedded vision platform for addressing the above-mentioned challenges; moreover, it helps enable advanced AR as well as smart industrial applications. We will leverage on heterogeneous computing and 3D imaging technologies for developing technical features; including simultaneous localization and mapping (SLAM) algorithm for embedded system, dense 3D mapping for Enterprise AR and robot applications, as well as depth and visual sensor fusion for smart industrial applications, etc.