From Hobart, to London, to Dhaka: using cameras and AI to build an automatic litter detection system

Shutterstock

Arianna Olivelli, CSIRO and Uwe Rosebrock, CSIRO

It’s estimated about two million tonnes of plastics enter the oceans from rivers each year. But our waterways aren’t just conveyor belts transporting land waste to the oceans: they also capture and retain litter.

Currently, the most common method for monitoring litter relies on humans conducting on-ground visual counts. This process is labour-intensive and makes it difficult to monitor many locations simultaneously or over extended periods.

As part of CSIRO’s research to end plastic waste, we’ve been developing an efficient and scalable environmental monitoring system using artificial intelligence (AI).

The system, which is part of a larger pilot with the City of Hobart, uses AI-based image recognition to track litter in waterways.

Global insights help build a reliable model

The technology is underpinned by two branches of AI: computer vision and deep learning. Computer vision involves training computers to understand and interpret images and videos, whereas deep learning imitates how our brains process data.

Drawing on these capabilities, we worked in partnership with Microsoft (using its Azure cloud computing services) to develop an automated system for monitoring river litter.

We have been detecting and classifying items floating on the surface of Hobart’s stormwater channels, the River Thames in the UK and the Buriganga River in Bangladesh.

We’ve remotely analysed the amount of litter, the type of litter and how this changes across locations.

CSIRO research scientist Chris Wilcox setting up a fixed camera to monitor litter in Hobart.

Major damage from food packaging and bottles

Our work relies heavily on two applications of computer vision. These are “object detection” and “image classification”.

Object detection specifies the location of a particular object in an image and assigns it a label. Image classification assigns one or more labels to the image as a whole.

Before either of these models can be applied reliably, however, they have to be trained, tested and validated using a large number of labelled images. For this, we drew from our footage of river litter collected from Hobart, London and Dhaka.

Our dataset now contains more than 6,100 images with 14,500 individual items. The items are labelled across more than 30 categories including plastic bottles, packaging, beverage cans, paper and plastic cups.

Our data revealed food packaging, beverage bottles and cups were by far the most frequently spotted litter items across all three countries.

The Buriganga river flows by Dhaka. It’s one of Bangladesh’s most polluted rivers due to the ongoing dumping of industrial waste (such as from leather tanneries) and human waste.
Shutterstock

Fake images aren’t always harmful

To build a well-performing machine learning model, we needed a balanced set of training images featuring all item categories — even if certain categories are more frequent in real life.

Introducing synthetic (computer generated) images to our dataset was a game changer.

These images were generated by Microsoft’s synthetics team based in Seattle. They rendered various objects and superimposed them over backgrounds obtained from our field photos.

Once the digital objects were created, the superimposition process was automatic. Thus, the team managed to produce thousands of synthetic pictures over just a few weeks, rapidly expanding our training dataset.

In this synthetic image, the transparent cup, face mask and aerosol container are digital renderings superimposed over an original photo taken by one of our cameras.

How are objects identified?

There are a few steps by which our system identifies litter objects in photos. First, the photos are all scored against a single-label (“trash”) object detector. This identifies items of litter in the frame and stores their coordinates as annotations.

These coordinates are then used to isolate the items and score them against an image classifier which includes all the litter categories.

Finally, the model presents the category it thinks the item most likely belongs to, along with a suggested probability for how accurate this guess is.

Here’s an example of the system detecting a water bottle and packaging as trash, and then placing both items into their respective categories. Probabilities are provided for the likely accuracy of the system’s guess regarding an item’s classification.

An AI-driven approach to litter management allows a quicker response than a manual system. But when it comes to litter, the major challenge lies in creating a model that can account for millions of different shapes, colours and sizes.

---

Read more:
As cities grow, the Internet of Things can help us get on top of the waste crisis

---

We wanted to build a flexible model that could be transferred to new locations and across different river settings, including smaller streams (such as Hobart’s stormwater system) and large urban rivers (such as the River Thames or the Buriganga River).

This way, rather than building new models for each location, we only have to deploy more cameras. Data retrieved could help identify litter hot spots, implement better waste-related policies and improve waste management methods to make them safer, smarter and relatively cheaper.

Keeping an eye on Hobart’s litter

We’ve also been collaborating with the City of Hobart to develop an autonomous sensor network to monitor gross pollutant traps, such as floating barriers or litter socks.

These structures, integrated into Hobart’s stormwater drainage system, are supposed to prevent solid waste such as cans, bottles, tree branches and leaves from reaching the estuary and ocean.

We currently have a network of sensors and six cameras installed under bridges tracking litter in the traps. The system can inform an operator when a trap requires emptying, or other maintenance.

Once in full use, the technology will provide almost real-time monitoring of litter around Hobart — assisting efforts to reduce environmental harm caused by stagnant, and potentially hazardous, waste lost to the environment.

---

Read more:
How sensors and big data can help cut food wastage

---

Arianna Olivelli, Research Affiliate, CSIRO and Uwe Rosebrock, Senior Software Engineer, CSIRO

This article is republished from The Conversation under a Creative Commons license. Read the original article.

Advertisement

The world’s best fire management system is in northern Australia, and it’s led by Indigenous land managers

Rohan Fisher, Charles Darwin University and Jon Altman, Australian National University

The tropical savannas of northern Australia are among the most fire-prone regions in the world. On average, they account for 70% of the area affected by fire each year in Australia.

But effective fire management over the past 20 years has reduced the annual average area burned – an area larger than Tasmania. The extent of this achievement is staggering, almost incomprehensible in a southern Australia context after the summer’s devastating bushfires.

---

Read more:
I made bushfire maps from satellite data, and found a glaring gap in Australia’s preparedness

---

The success in northern Australia is the result of sustained and arduous on-ground work by a range of landowners and managers. Of greatest significance is the fire management from Indigenous community-based ranger groups, which has led to one of the most significant greenhouse gas emissions reduction practices in Australia.

As Willie Rioli, a Tiwi Islander and Indigenous Carbon Industry Network steering committee member recently said:

Fire is a tool and it’s something people should see as part of the Australian landscape. By using fire at the right time of year, in the right places with the right people, we have a good chance to help country and climate.

Importantly, people need to listen to science – the success of our industry has been from a collaboration between our traditional knowledge and modern science and this cooperation has made our work the most innovative and successful in the world.

A tinder-dry season

The 2019 fire season was especially challenging in the north (as it was in the south), following years of low rainfall across the Kimberly and Top-End. Northern Australia endured tinder-dry conditions, severe fire weather in the late dry season, and a very late onset of wet-season relief.

Despite these severe conditions, extensive fuel management and fire suppression activities over several years meant northern Australia didn’t see the scale of destruction experienced in the south.

A comparison of two years with severe fire weather conditions. Extensive early dry season mitigation burns in 2019 reduced the the total fire-affected areas.

This is a huge success for biodiversity conservation under worsening, longer-term fire conditions induced by climate change. Indigenous land managers are even extending their knowledge of savanna burning to southern Africa.

Burn early in the dry season

The broad principles of northern Australia fire management are to burn early in the dry season when fires can be readily managed; and suppress, where possible, the ignition of uncontrolled fires – often from non-human sources such as lightning – in the late dry season.

Traditional Indigenous fire management involves deploying “cool” (low intensity) and patchy burning early in the dry season to reduce grass fuel. This creates firebreaks in the landscape that help stop larger and far more severe fires late in the dry season.

Relatively safe ‘cool’ burns can create firebreaks.
Author provided

Essentially, burning early in the dry season accords with tradition, while suppressing fires that ignite late in the dry season is a post-colonial practice.

Savannah burning is different to burn-offs in South East Australia, partly because grass fuel reduction burns are more effective – it’s rare to have high-intensity fires spreading from tree to tree. What’s more, these areas are sparsely populated, with less infrastructure, so there are fewer risks.

---

Read more:
The burn legacy: why the science on hazard reduction is contested

---

Satellite monitoring over the last 15 years shows the scale of change. We can compare the average area burnt across the tropical savannas over seven years from 2000 (2000–2006) with the last seven years (2013–2019). Since 2013, active fire management has been much more extensive.

The comparison reveals a reduction of late dry season wildfires over an area of 115,000 square kilometres and of all fires by 88,000 square kilometres.

How fire has changed in northern Australia.
Author provided

Combining traditional knowledge with western science

The primary goals of Indigenous savanna burning projects remain to support cultural reproduction, on-country living and “healthy country” outcomes.

Savanna burning is highly symbiotic with biodiversity conservation and landscape management, which is the core business of rangers.

Ensuring these gains are sustainable requires a significant amount of difficult on-ground work in remote and challenging circumstances. It involves not only Indigenous rangers, but also pastoralists, park rangers and private conservation groups. These emerging networks have helped build new savanna burning knowledge and innovative technologies.

---

Read more:
Our land is burning, and western science does not have all the answers

---

While customary knowledge underpins much of this work, the vast spatial extent of today’s savanna burning requires helicopters, remote sensing and satellite mapping. In other words, traditional burning is reconfigured to combine with western scientific knowledge and new tools.

For Indigenous rangers, burning from helicopters using incendiaries is augmented by ground-based operations, including on-foot burns that support more nuanced cultural engagement with country.

On-ground burns are particularly important for protecting sacred sites, built infrastructure and areas of high conservation value such as groves of monsoonal forest.

Who pays for it?

A more active savanna burning regime over the last seven years has led to a reduction in greenhouse gas emissions of more than seven million tonnes of carbon dioxide equivalent.

---

Read more:
Savanna burning: carbon pays for conservation in northern Australia

---

This is around 10% of the total emission reductions accredited by the Australian government through carbon credits units under Carbon Farming Initiative Act. Under the act, one Australian carbon credit unit is earned for each tonne of carbon dioxide equivalent that a project stores or avoids.

By selling these carbon credits units either to the government or on a private commercial market, land managers have created a A$20 million a year savanna burning industry.

How Indigenous Australians and others across Australia’s north are reducing emissions.

What can the rest of Australia learn?

Savanna fire management is not directly translatable to southern Australia, where the climate is more temperate, the vegetation is different and the landscape is more densely populated. Still, there are lessons to be learnt.

A big reason for the success of fire management in the north savannas is because of the collaboration with scientists and Indigenous land managers, built on respect for the sophistication of traditional knowledge.

This is augmented by broad networks of fire managers across the complex cross-cultural landscape of northern Australia. Climate change will increasingly impact fire management across Australia, but at least in the north there is a growing capacity to face the challenge.

Rohan Fisher, Information Technology for Development Researcher, Charles Darwin University and Jon Altman, Emeritus professor, School of Regulation and Global Governance, ANU, Australian National University

This article is republished from The Conversation under a Creative Commons license. Read the original article.

Climate change must be part of Australia’s electricity system review

Tony Wood, Grattan Institute

On Friday, Australia’s federal and state energy ministers met for an extraordinary meeting following the complete loss of power in South Australia on September 28. The COAG Energy Council announced a wide-ranging independent review to provide advice to governments on a coordinated, national reform blueprint. The review will be chaired by Australia’s Chief Scientist, Dr Alan Finkel.

Dr Finkel has been challenged with steering Australia’s energy system around some big potholes while keeping his eye on the horizon. And all in about six months.

The review will consider work already being done around maintaining the security, reliability and affordability of electricity as delivered by the National Electricity market (NEM) (which covers all states except Western Australia and the Northern Territory).

The state-wide blackout became a political opportunity for Australia’s politicians. Yet it is certainly too early and hopefully wrong to say if this is just a reactive response.

What’s in the review?

The review has three timeframes. The immediate priority will be to systematically assemble existing processes and work programs initiated over the last year by the energy council and identify any major gaps in the context of energy security and reliability in the NEM. Some of these processes, such as a review of market governance arrangements, have been completed but not fully actioned.

Others have only recently been announced. These include analysis of the impact of federal, state and territory carbon policies on energy markets and the reviews of the South Australian blackout. They will not all be complete by the December council meeting.

The review is expected to deliver a blueprint via a final report early in the new year. It is likely to include specific actions, both physical and financial, that respond to recent events such as South Australia’s price shock in July and blackout in September. These two issues should not be conflated. To do so, would risk solving neither.

The council has highlighted the significant transition underway in the Australian electricity market. The drivers include “rapid technological change, the increasing penetration of renewable energy, a more decentralised generation system, withdrawal of traditional baseload generation and changing consumer demand”. The blueprint will address all of these issues in a comprehensive and coordinated way not previously a feature of the council’s output.

There is much uncertainty to how some of these drivers will evolve over the next two decades. To be really effective, the blueprint will need to consider a range of plausible long-term scenarios but focus on near-term options that can be adapted to evolving developments on all fronts.

The Chief Scientist will, amongst other things, bring to the review his knowledge of current and likely future developments in energy technologies. This will be important in considering policy, legislative and rule changes that favour the adoption of technologies that could address both low-emissions and reliability but are otherwise technology-neutral.

The federal energy minister, Josh Frydenberg, and his state and territory counterparts are to be applauded for the speed and cohesiveness they have shown in instigating the review. This follows a similar approach that permeated their August meeting where considerable progress was made on key energy market reforms across several fronts.

Get climate policy right

There are two critical areas of concern. The fundamental driver behind the issues listed in the review’s terms of reference is climate change and the policy response to it.

The federal government is committed to a 2017 review of its domestic climate change policies against its 2030 emissions reduction target.

State and territory governments have announced or implemented their own climate change and renewable energy policies. It is not surprising that states such as Victoria remain committed to these policies even though they are open to criticism for being uncoordinated at a national level, or failing to consider implications for system reliability and security.

Primary responsibility must rest with the federal government to deliver a credible scalable climate policy. Much can then flow from there, including agreement from states and territories to truly act in the spirit of national coordination to which they committed.

Greenhouse gas reduction is best achieved by putting a price on emissions through one of several options that have been canvassed in 2016 and in a form that acts with the electricity market and not outside it. The review’s terms of reference are silent on this issue, and yet recognise that the nature and structure of climate change policy have critical implications for the NEM.

Wind and solar power are intermittent. Their integration into the generation mix while maintaining reliability is best achieved by valuing flexibility either through the NEM or via complementary policies or regulations. The review is oddly silent on this issue. It is to be hoped that this is unintended and will be picked up in the course of the review.

There were high expectations for Friday’s council meeting. A state-wide blackout does that. These expectations have been delegated to the review which the council must support and drive to outcomes.

Minister Frydenberg has strongly and repeatedly emphasised that the government will not compromise energy reliability and security in the transition to a low emissions future. Failure on this front will not be forgiven.

Tony Wood, Program Director, Energy, Grattan Institute

This article was originally published on The Conversation. Read the original article.

Kangaroo Detection System for Cars

Estuarine Crocodiles: Crocodile Society

The link below is to an article that looks at life within the Estuarine Crocodile social system.

For more visit:
http://www.australiangeographic.com.au/journal/alpha-boss-crocs-rule-crocodile-society.htm

Asia: Dams to Cause Environmental Disaster – Mekong River

The link below is to an article reporting on the construction of some 78 dams along the Mekong River system in south-east Asia, raising major concerns for the health of the river system and its fish population.

For more visit:
http://www.scidev.net/en/agriculture-and-environment/news/dams-a-potential-catastrophe-for-mekong-fisheries.html