2018年8月13日星期一

HKIE Seminar - AI Technologies for Precision Forestry and Fire Detection

The Hong Kong Institution of Engineers (HKIE) organized the technical seminar named “AI Technologies for Precision Forestry and Fire Detection” on 13 Aug 2018. AI and Robotics are hot topics and this seminar showed how AI and Robotics technologies helping our world in green economy.  My colleague - Mr. Curie Lee, guest speaker - Mr. Rex Sham and I took a photo for memory.


In the beginning, organizers took a group photo.


Mr. Rex Sham (Co-Founder and Chief Science Officer, Insight Robotics) was the guest speaker and his topic was “Insight Robotics – AI Technology Solutions for a Better World”.  Firstly, he introduced his company which founded in 2011 and their services included Wildfire Detection, Forestry Pest & Disease Detection and Forestry Resources Survey.  Their team had 40+ staff in which 40% were R&D.  They are the first company to export Wildfire Detection Robots from China and received 15M+ USD Private Investment.


The following diagram briefed their services by using Robots, Drones and AI to help their client by making sense of data and reduced risks and prevented losses. 


Rex discussed that wildfire caused 30% of carbon emission in which 92% of wildfire caused by human activities.  He said it was difficult to identify the location of fire source under Haze, Smog, Air pollution and Smoke.  Moreover, human monitoring of wildfire was boring and expensive. 


Moreover, using towers were difficult to identify the location of fires. The following diagram demonstrated the wrong estimation of fire location.


Using Satellites, but its resolution was not enough to point out the location. Specification of MODIS (Terra & Aqua) were:
-          Best of the World
-          250m x 250m per Pixel
-          Best-case 900m2 fire Size (4 pixels)
-          6 hours revisit internval
-          705 km above ground
-          Fire product is 1km x 1km per Pixel


And then Rex showed the case on December 8, 2011 which participate and provide exclusive fireproof technical support in Zengcheng City Guangdong Province fire drill event – Guangdong Provincial Forest Fire “3-Dimensional” Practice (2011128日視野機器人有限公司參與廣東省在增城舉辦的<廣東省森林消防立體滅火實戰演習>及獨家提供防火技術支持).


After that Rex shared some application cases in different areas and countries.  


Another extended service is Drones Data Analytics on high speed high resolution mapping for finding small Wildfire and building the local terrain map.


Finally, he demonstrated their High Resolution Ultra Large Area Image Stitching in Indonesia.  They had capability for tree counting and tree health record for every single tree. He told us to use human marking for 6 M trees within 4 months.  


Q&A Session
I asked questions how to identify the different species of tree.  Rex said it was different and need to train by professionals.  They planned to develop open platform for international cooperation.  In their coding, 20% is open source and 80% is close source.  He planned to increase the open source.


In here, I introduced my study on GIS related on forest classification.  My conference paper named “A Study on the Characteristic of Bulk and Canopy Deposition Fluxes in Hong Kong employing GIS Technique” in POLMET 2000 Kuala Lumpur (The 6th International Conference on Pollution in Metropolitan Cities). 

Systeme Pour l’Observation de la Terre (SPOT) was employed and which had two identical high-resolution visible (HRV) imaging instruments, each of which acquires images of a strip 60km wide.  The combined field of view is 117km with 3km overlap.  It acquires image data in two bands in the visible part (green and red) and one in the near-infrared.


After employing Hong Kong Ecological Database with classification data on vegetation coverage, the element fluxes were estimated into three classes, which are city area, forest and grassland.  However, classification of remote sensing image were Forest Area (Green), Glass Land (Brown), Artificial Area (Red) and Unclassified Area (Black).  

Reference:
Insight Robotics Ltd - https://www.insightrobotics.com/en/

沒有留言:

發佈留言