WiMi Hologram Cloud Inc (WIMI)

[StockpickerCAPC]

WiMi Develops Holographic Vision-Based HV-SLAM For Passive Navigation System.

Source
https://www.moomoo.com/community/feed/109674638802949

January 12, 2023

With the recent development of intelligence and algorithms, SLAM (simultaneous localization and mapping) technology has become one of the focus of market attention. SLAM or CML (Concurrent Mapping and Localization), instant localization and map construction or concurrent map building and localization, is considered to be a key technology for robots or some intelligent devices to perform autonomous mobility. WiMi Hologram Cloud, Inc. (NASDAQ:WIMI) has been investing in the R&D of SLAM for hologram vision 3D map construction, HV-SLAM. In general, intelligent machines such as robots and drones that need to move and cruise usually also use this technology, such as satellite navigation control. However, user may encounter many technical problems, such as signal loss, or a 3D complex structured environment. Positioning will be needed because it can sense the surroundings in an unknown environment through sensors for the positioning of the machine and thus the navigation.

WiMi's HV-SLAM is a technology for passive navigation and self-cruising positioning. HV-SLAM is crucial for the mobility and interaction of intelligent devices such as robots, drones or UAVs, because it represents the basis of this capability: knowing the location, surroundings, how to act autonomously next. It can be said that any intelligent body with some mobility has some form of SLAM system.

WiMi's HV-SLAM obtains images from a depth camera. The camera contains three core devices:
- a laser projector,
- an optical diffraction element (DOE), and
- an infrared camera, which are used to form a 3D holographic map of the system.
It allows the device to better intelligently determine its course of action and how to act. For example, when robots come to an unfamiliar environment, to quickly familiarize with the environment and complete the task, then robots will need to do the following things.

Feature extraction:
acquire information such as the surrounding environment with the sensor and record the feature data.
Map construction: according to the information acquired by the sensor, the environmental features are constructed in the system in the form of a 3D holographic map.

Dynamic calibration adjustment:
during the movement, new feature landmarks are continuously acquired and the 3D holographic map model in the system is corrected.

Trajectory annotation:
determining its position based on the feature landmarks acquired during the previous period of movement.

Loop closure detection:
check whether the loops can be matched and return safely.

The algorithm of WiMi’s HV-SLAM constructs a 3D holographic map of the world in real time and simultaneously tracks the position and orientation of the camera (head-mounted on a handheld or augmented reality device or mounted on a robot). Combining convolutional neural networks and deep learning to achieve self-cruising positioning passive navigation functions, HV-SLAM focuses on geometry and space, while deep learning is applied to perceive and recognize objects to give suggestions. Computer vision techniques are applied to visual SLAM that includes detection, description and matching of salient features, image recognition and retrieval, etc. to significantly improve efficiency. The HV-SLAM system can help the machine understand the world in front of it geometrically, i.e., build associations in the local coordinate system, while the deep learning system can help the machine perform classification reasoning, i.e., build associations between different object instances.

Of course WiMi’s HV-SLAM also has some algorithms that in reality may need to be considered and continuously optimized and upgraded. Going forward, multi-sensor fusion, optimized data association and loopback detection, integration with front-end heterogeneous processors, improved robustness and repositioning accuracy are the next development direction of SLAM technology, but these will be gradually resolved with the development of consumer stimulus and industry chain.