CQU Data Tool Eases “Mango Madness”

CQUniversity-developed FruitMaps is a data tool that takes real-time data from multiple sources and displays them visually to provide a simple, free, online decision support tool adapted for use by farmers to assess fruit maturity and assist harvesting planning.

The data collections that underpin the tool are stored on QRIScloud, QCIF’s cloud computing service.

FruitMaps’ pilot users, farm managers in Queensland and the Northern Territory, have welcomed the data tool to help minimise crop loss and plan for resources, such as labour, at peak times.

FruitMaps is currently being developed for other uses, such as assessing soil quality, and quality control of coral trout exported to China, and for use with other crops important to Australia’s economy, such as citrus, bananas and avocadoes.

Full article

See the 18 March 2019 update

A CQUniversity-developed data tool is helping Australian mango growers determine a number of harvesting factors ahead of time: crop size, when to pick, and how many labourers to hire and cartons to purchase.

FruitMaps, developed by Professor Kerry Walsh and postdoctoral research fellow Zhenglin Wang, takes real-time data from multiple sources and displays them visually to provide a simple, free, online decision support tool adapted for use by farmers to assess fruit maturity and assist harvesting planning. The data collections that underpin the tool, which is in its pilot stage, are stored on QRIScloud, QCIF’s cloud computing service.

Mangoes are harvested within a short window each year in growing regions in Australia’s north, a time known to farmers as “mango madness” given the hectic schedule.

When do you call the harvest start? Too early and the fruit will not be an optimum eating quality. Too late and fruit will either drop off the tree and be damaged or not travel well.

FruitMaps was created to ease the pressure of “mango madness” as a decision support system to aid growers in planning their harvest more than a month in advance. The tool began life three years ago.

“We developed the basics of the system three growing seasons’ ago, we fleshed it out in the second season and we’re still expanding its capability in the third season, which is upon us now,” Prof. Walsh said.

FruitMaps’ pilot users are farm managers in Queensland and the Northern Territory.

Senior agronomist Adam Kent of Simpsons Farms in Goodwood, Queensland, near Bundaberg, has recently started using FruitMaps in preparation for the farm’s next mango harvest.

The farm has about 22,000 mango trees. Forecasting harvest date and crop yield and thus determining when and how many fruit pickers to hire will be Mr Kent’s primary use for FruitMaps.

“If you get your harvest right, crop loss is very minimal. If you get it wrong, it can be quite high,” Mr Kent said. “Our harvest is after Christmas in January. Labour resources at Christmas time and New Year’s is incredibly hard to manage.

“I see the value in forecasting harvest date to ensure resources at peak time are available. And that’s just scratching the surface: Version two of FruitMaps is just up and running and we’ll get more out of it by putting more data in.”

Mr Kent has recently logged the flowering data of Simpsons Farms’ mango trees into FruitMaps and has already mapped the property’s orchard blocks into the app: “It’s very easy to use,” he said.

FruitMaps was born as an accessory to the ‘F-750 Produce Quality Meter’, developed to non-invasively assess fruit maturity in the field and to assist with the decision to harvest.

F-750 is a field-friendly spectrometer that can assess fruit dry matter content, which is a measure of its starch and sugar content. The technology has been taken from CQUniversity research and development into commercial production by Felix Instruments – Applied Food Science, an American company that designs agricultural devices.

Kyle Hikalea of Felix Instruments has also led the shift from R Studio to a JavaScript and Python Django coding environment for the FruitMaps project.

F-750 data was originally collected in-field, then transferred from the device to a farm Excel workbook to allow calculation of averages and trends. FruitMaps was built to allow easy visualisation of this data, graphically on a farm map, and within tables and graphs (e.g. Figure 2, below).

“But we quickly found that FruitMaps was a foundation stone for all our work, and now we funnel temperature data (assessing fruit maturity by heat sums), data from in-field imaging units on level of flowering, fruit load and fruit sizing, all displayed and interpreted in the FruitMaps application,” said Prof. Walsh (e.g. Figure 3, below).

The tool also has a variant useful for research and is currently being used within the work of several CQUniversity (CQU) postdoctoral fellows and research higher degree students.

The FruitMaps team first heard about QRIScloud through QCIF’s CQU-based eResearch Analyst Jason Bell, who Prof. Walsh described as “very enthusiastic, very informative and very supportive.”

Mr Bell helped the project team with all aspects of QRIScloud, including setting up and configuring servers, and providing technical troubleshooting advice.

“QRIScloud has allowed for a central collection point, including images and crop maturity data. This site allows CQU researchers and collaborators, including farm managers, to upload data from remote locations, and to access data from anywhere,” Prof. Walsh said.

“The spatial display (measurements superimposed on farm maps, organised by orchard blocks) has allowed for new lines of thought on the agronomy underlying the spatial variability, and on management of the variability. This understanding led to the addition of analytics to the application, e.g. forecast of optimum harvest time.

“QRIScloud has provided access to large storage (currently a 1 TB data share), a system that is accessible outside the university network, and thus the development of a ‘pilot Web system’ at a low cost compared to hosting it on a commercial site.

“The QRIScloud resource has enabled us to keep enacting our dreaming of relevant features, with each new mango madness season bringing a new burst of grower and collaborator inputs and features.”

FruitMaps will continue to support an increased repertoire of activity, from automated image processing pipelines to tree crop yield mapping. The system also acts as a store of data, to be mined later.

Future potential uses include as a tool to capture information about tree cultivars and ages per orchard to guide industry groups in planning industry growth.

FruitMaps will also branch out to explore use with other crops important to Australia’s economy, such as citrus and bananas, and implementation is already underway for avocadoes.

Jason Bell continues to support FruitMaps and provides advice on the future direction of the project’s IT aspects.

FruitMaps has been supported by Horticulture Innovation Australia and the Australian Government Department of Agriculture and Water Resources under the Rural Research and Development (R&D) for Profit program (ST15005).

Update: 18 March 2019 — “Farming is technology, farming is exciting”

CQU Professor Kerry Walsh and his FruitMaps team have worked with the Queensland Government Department of Agriculture and Fisheries (DAF) over the last summer, assessing the crop load efficiency of its “Small Tree High Productivity Initiative” field trial site at the Walkamin Research Facility, near Mareeba.

The Walkamin facility, about an hour’s drive inland from Cairns, is leading and innovating in a wide range of scientific research and development. One long-term trial involves a large mango orchard planting systems experiment comparing planting densities, cultivars and canopy training systems. The high-density orchard systems are showing great potential for improved productivity and profitability.

FruitMaps was used to record images and machine-vision fruit load estimates of all trees in the mango orchard planting systems trial. The tool has assisted researchers to assess the various orchard system designs for their suitability for future machine-vision automation. Fruit Maps provides an early estimation of crop loads for farmers and researchers.

Collaboration with DAF orchard systems researchers will continue in future years providing an accessible record of orchard progress.

The FruitMaps team has also started to help others transform the data tool to other uses. They are currently working with CarbonLink to help farmers map and calculate soil carbon change to improve their land’s soil quality and thus improve food production and quality.

The FruitMaps team is also beginning a project with the Australian Reef Fish Trading Company (ARFTC) on the quality control of coral trout exported to China.

Meanwhile, ABC’s Landline TV program, which deals with rural Australian issues, featured FruitMaps in its “Mango Dreaming” report on 9 February 2019.

Mango producer Martina Matzner, the largest grower in the Northern Territory with about 50,000 trees, and a key developer of the Calypso variety of mango, showed ABC reporter Kristy O’Brien how she uses FruitMaps and said the tool is “taking the risk out of picking immature fruit.” FruitMaps, alongside other technology, also helps determine quite precise yield predictions.

“Once you know what your yield is, and you know it early enough, it comes automatically — you know how many staff you need, you know when you’re going to harvest, you know how many trucks you need, you know how many cardboard boxes you need to order, how many trailers go in there, when the fruit will be at the market.”

“Farming is technology, farming is exciting,” said Ms Matzner.

As Ms O’Brien noted: “It’s cutting edge technology that has huge impacts. The mango industry has expensive inputs and relies on a lot of labour. Knowing exactly when fruit will reach perfect ripeness reduces extremely costly guesswork.”

According to Ms O’Brien, Australia’s mango industry grew by 71 per cent from 2013 to 2017. “The push is on to secure more international markets and continue to see Australian households eat more mangoes.”

Figure 2. Example of association of in-field F-750 dry matter readings to orchard blocks, colour coded in terms of meeting a harvest criterion (block in red is not yet ready for harvest). (Image: CQUniversity.)

Figure 3. Example output of a field imaging pass with image processing to extract numbers of fruit per tree side. Each dot represents a tree side. (Image: CQUniversity.)

Figure 4. Field imaging rig operating in the Northern Territory. Images are passed via satellite link for processing (flower or fruit count) and displayed on FruitMaps. (Image: CQUniversity.)