Land classification has long been a key staple of the Earth Observation industry, and at the end of January, an exciting paper was published on how satellite data could be used to support work on anti-slavery.
The paper â€˜Earth Observation and Machine Learning to Meet Sustainable Development Goal 8.7: Mapping Sites Associated with Slavery from Spaceâ€™ by Foody et al was published in the journal Remote Sensing.
It describes the researchers, from the University of Nottingham and the Chinese Academy of Sciences, work using machine learning techniques to try and identify brick kilns in the Rajasthan region of India, part of the so-called Brick Belt which stretches across India, Pakistan, Bangladesh and Nepal. Itâ€™s estimated that up to sixty-eight percent of the 4.4 to 5.2 million manual labours working in the brick kilns in this belt are enslaved in bonded or forced labour, and nineteen percent of the workers are under the age of 18.
The team focused on identifying Bullâ€™s Trench Brick Kilnâ€™s, these are oval or circular in shape and are named after the British engineer W.Bull who devised these kilns at the start of the 20th Century based on an earlier design used in Germany. Unlike a lot of satellite-based classifications which focuses on using spectral signatures, this work utilised the distinctive shape of the brick kilns for the classification. Although, they vary in size the kiln shape is distinctive and it often has a tall chimney in the middle.
The team used images from Google Earth and applied a machine learning classifier founded on region-based convolution neural networks (R-CNN) focusing on the kiln shape as the key identifying feature. An initial classifier, the Faster R-CNN, was trained to identify brick kilns, however, this over-estimated the number of kilns in the region. This was refined with classification using a CNN and finally, a known training dataset was used that had been obtained by previous human visual interpretations.
The outputs of this work produced an overall accuracy of 94.94%, picking out 169 of the 178 known kilns in the area of interest in Rajasthan, which shows that this sort of analysis could be useful in supporting antislavery work.
Secondly, last week, the Kenya Forest Service reported that they were going to undertake a pilot project to use Earth Observation to monitor tropical forests; in particular the remote and harder to reach areas. Using satellites will allow them to make assessments faster and more accurately than previously, which were done through ground-surveys. The pilot project aims to map just over two thousand kilometres of Kwale County, which is located on the south coast of Kenya and borders Tanzania. Within the Shimba Hills National Reserve is one of the largest coastal forests in East Africa.
Working in the industry weâ€™re passionate about Earth observation and satellite images, however, these examples show just how powerful a tool we have and how it can change the world!