UASs for Plant biosecurity and Plant Pest Detec6on and Surveillance: Challenges and Opportuni6es

Transcription

1 UASs for Plant biosecurity and Plant Pest Detec6on and Surveillance: Challenges and Opportuni6es Mcfadyen, A 1,2, Gonzalez, F 1,2, Campbell, D 1,2, Eagling, D 2, Gaston, K 3 1,2 Australian Research Centre for Aerospace AutomaAon (ARCAA), Queensland University of Technology (QUT), 2 Plant Biosecurity CRC 3 Director of Environment & Sustainability InsAtute, University of Exeter Professor of Biodiversity & ConservaAon, University of Exeter UAV mee'ng at the University of Exeter UAV conference July 2014 Two Videos of our UAV with Spore trap and a mulaspectral camera hpps:// Agricultural%20UAV%20Examples%20.mp4

2 arcaa Australian Research Centre for Aerospace AutomaAon (ARCAA) ARCAA conducts research into all aspects of aviaaon automaaon, with a paracular research focus on autonomous technologies which support the more efficient and safer ualisaaon of airspace, and the development of autonomous aircraz and on- board sensor systems for a wide range of commercial applicaaons. Enabling Technologies: sense and avoid, automated emergency force landing Example Applica6ons: Power line inspecaon with Highly Automated AircraZ Gas sensing environmental monitoring Precision agriculture- plant pest detecaon with UAS

3 Evalua6ng Unmanned AircraH Systems for Deployment in Plant Biosecurity (5055) hpp:// Scope: plant pest ; no weed detecaon (PBCRC does not looks at Weeds)

4 Outline 1. Problem being addressed 2. RegulaAons & Advisories 3. UAS Placorm design/selecaon consideraaons 4. Sensors design/selecaon consideraaons 5. Data Processing consideraaons 6. Technology Survey; Matching UAS to sensors 7. Challenges and OpportuniAes UAS OperaAons, aerial placorms, sensors and data 8. Conclusions

5 Problem being addressed Early and cost effec6ve detec6on and monitoring of harmful plant pest or disease incursions is Ame consuming but extremely important for plant biosecurity. We explored the feasibility (e.g. technical, economic and legal) of deploying UAV s in a plant biosecurity context. (a) Observation 1 (b) Observation 2 (c) Observation 3 Example pre-infection (a), mild infection (b), and severe infection (c) of cereal crops from disease.

6 Target Crops & Pests 4% 17% 19% Land Coverage (Ha) 60% Wheat Barley Oats Other Value ($) Wheat 0% 21% Barley 61% 18% Oats Other 4% 18% 29% Loss ($) 49% Yellow Spot Stripe Rust Aphids Mites Estimated relative land coverage (green), annual value (grey) and financial loss (red) of cereal crops.

7 Target Crops & Pests Over AUD878Million loss due to fungal disease was observed in recent years. Classified as high or extreme risks to the Australian grain industry with yellow spot and stripe rust causing in excess of AUD300Million in annual loss. Stripe Rust (b) Yellow Spot (c) Barley Yellow Dwarf Virus (d) Stem Rust

8 Target Crops & Pests Over AUD272Million in loss was recently observed due to combined effect of invertebrates including Red- Legged Earth Mites/Blue Oat Mites and various Aphids. (a) Russian Wheat Aphid (b) Other Wheat Aphids (c) Oat Mite (d) Oat Mite Damage (e) Oat Mite Damage

9 Who will use this research? Growers Agricultural consultants Federal agencies Research Peers

10 Examples of UAS Pest DetecBon David G. Schmale III, Benjamin R. Dingus and Charles Reinholtz (2008). Development and applica6on of an autonomous unmanned aerial vehicle for precise aerobiological sampling above agricultural fields. Journal of Field RoboAcs, 25(3), pp ST autonomous UAV with the aerobiological sampling devices mounted in front of the leading edge of the wings. Sensor Used: Aerobiological Sampling Devices UAV Used: Senior Telemaster Autopilot: MicroPilot MP2028g Video Link hpps:// v=jfdcks8upfu Aerobiological samples collected 100m above crop fields with an autonomous UAV. Colonies of Fusarium (lez) were rendered from FSM and microbial assemblages (right) were rendered from PDA.

11 Examples of UAS Pest DetecBon Gonzalez, Felipe, Narayan, Pritesh P., Castro, Marcos P.G., & Zeller, Les (2011) Development of an autonomous unmanned aerial system to collect 6me- stamped samples from the atmosphere and localize poten6al pathogen sources. Journal of Field RoboAcs, 28(6), pp Spore trap installaaon. The spore trap was fiped and secured on- board the UAS placorm and the spore trap servo sensor driving tape was integrated with the on- board autopilot system. Sensor Used: Volumetric Spore Trap UAV Used: Flamingo UAV Autopilot: Microilot 2128g Geo- Referencing Captured Samples. It can be seen that the sample can be traced to the region indicated in the map overlayed on GPS coordinate data.

12 Examples of UAS Pest DetecBon Crea6ng a more RESilient QUeensland Unmanned AircraH for Emergency Response and Biosecurity- Sensing and Placorm AutomaAon for Miconia ApplicaAon Major Research Ac6vi6es Safe low- alatude flight in difficult terrain, a surveying system for the detecaon of the Miconia system. Video Link hpp://csironewsblog.com/2012/06/27/top- aerospace- experts- to- send- unmanned- aircraz- to- the- resqu/

13 Examples of Airborne based MulB- Spectral Imaging for Pest DetecBon Airborne/satellite or ground based imagery has been used for disease detection of cereal crops. A project in Colorado (USA) conducted field experiments to measure the effects of aphid infestations on cereal crops using an airborne multi-spectral camera[2]. Band 1 Band 2 Band 3 Basic principle for multi-spectral imaging. Course spectral partitions. Cereal crop infected by Russian aphids. Water stress detection at 2m resolution. Yellow and red indicate stressed regions.

14 Examples of Airborne based MulB- Spectral Imaging for Pest DetecBon In Oklahoma multi-spectral imaging was able to accurately discriminate between plant stress caused by Russian Aphid and other factors using airborne imagery (from manned aircraft) [1]. Using a UAS at lower altitude is expected to increase the spatial resolution, and therefore improve detail and thus accuracy of such results. (a) Raw Image (b) Aphid damage (c) Other damage

15 RegulaBons & Advisories Expected UAS Classes Subclass A B C D W e i g h t < >20 (kg) S p e e d (knots) Approval Requireme nts M : Mandatory P: Potential Online Approval (P) No additional documenta tion (P) Online Approval (P) Ops Manual (M) Maintenance Manual (P) Airworthiness (P) Operational Risk Assessment (P) Ops Manual (M) Maintenance Manual (P) Airworthiness (P) Operational Risk Assessment (P) Ops Manual (M) Maintenance Manual (M) Flight Manual (M) Airworthiness (M) Operational Risk Assessment (M)

16 RegulaBons & Advisories Item UAV controllers approval/ Unmanned pilot's licence) UOC approval process (worst case scenario) Renewals at the anniversary points, not including additional aircraft or information. Approximate Cost (AUD) (Initial licences for 1 year. The indication is then to have 3 yearly renewals, but this has yet to be confirmed)

17 UAS PlaHorm design/selecbon considerabons Approximate Flight Times Time (min) Overlap (%) 15kts 40kts 60kts Approximate flight times for coverage of a 1km 2 field. Each line represents a nominal operating speed (knots). The overlap represents the approximate percentage of ground that is covered twice. This is required for some sensors (imaging) in order to obtain useable data.

18 Sensors design/selecbon considerabons Small visual, (electro- opac), mula- spectral and hyper- spectral cameras are currently available for onboard implementaaon. It may be that only a paracular region of the EM spectrum is required for reliable disease discriminaaon. Wavelength Near Infrared Radio Microwave Infrared Visible Ultraviolet X-ray Gamma ray Approximate spectral imaging range for plant biosecurity

19 Sensors design/selecbon considerabons Sensor & Lens Parameters Camera Dimensions (mm): 6.66 x 5.32 Pixel Size (microns): 5.2 Focal Length (mm): 9.6 Altitude (m) (ft) Ground Resolution (m/pixel) FOV (m) (~ 400) x (~ 700) x (~ 1200) x (~ 3000) x 508 Low spatial resolution Ik+2 Ik+2 Ik+2 Ik+2 High spatial resolution (a) (b) In (a) potential image overlap issues (rose) and missed ground coverage (red) over cropped area (green) is depicted. In (b) spatial resolution difference with vehicle altitude is shown.

20 Data Processing considerabons Internet/Cloud UAV UAV Crowd Sourcing Farmer Ground Station Broader Community Biosecurity Community Plant Biosecurity Laboratory / Data Processing Centre Example information sharing and data flow using UAS technology for plant biosecurity

21 Technology Survey - UAS The report includes a technology survey which can be used to help match the appropriate sensor and placorm configuraaon to a specific biosecurity task. In many instances the sensors and placorms used could be used for other crops, pests and growing regions or climates than stated. Criteria UAS: Manufacturer Auto. Max Dim. Payload (kg) Cruise Speed Endurance Sensor Coverage Unit Cost 20 representaave UAS are included ; including Silvertone Flamingo, CyberEye II,Sensefly, Senior Telemaster, Quanta- G, CropCam, Sig Rascal, T- REX 600E, Vario XLV/XLC, MikroKopter OktoKopter, Custom Parasails

22 Criteria for Traps: Technology Survey- Sensors Appearance Data Rate Max Dim (m) Weight (kg) Pest/Disease Unit Cost Traps: 4 representaave Spore traps are included : USDA Spore Trap, VT Fungal Spore Trap DAFF QLD Fungal Spore, ARCAA ST Criteria for Vision Imaging, Mul6spectral and Hyperspectral sensors: Appearance Data Rate SpecRes. (nm) Pixel Res. FOV (deg) Max Dim (m) Weight (kg) Power (W) Pest/Disease Visual Imaging : 5 representaave sensors were evaluated including : Pentax OpAo Line, PointGrey CCD/ CMOS, PointGrey Spherical Mul6spectral: 7 representaave sensors were evaluated including: Tetracam mini- MCA, FluxData FD Series, PixelTeq Spectrocam Hyperspectral Sensor : 8 representaave sensors were evaluated including HyVista HySpex, Headwall Micro- Hyperspec, VTT UASI

23 Challenges and OpportuniBes The full report discuses 54 recommenda'ons They are categorized as : UAV OperaBons, Aerial plahorms, Sensors and data

24 Challenges and OpportuniBes in UAV OperaBons 1. Plaborm design/ choice should be primarily dictated by applica6on requirements and not as a result of an apempt to circumvent strict regulaaon. As such, experience and familiariza6on with the approval process should be sought, acquired and applied in conjuncaon to designing any UAS for plant biosecurity. 2. For rapid adopaon of UAS technology in biosecurity, small UAS under 2kg operaang under 400 within visual line of sight (VLOS) and away from populous areas (rural, remote and farm- based operaaons) should be employed.

25 Challenges and OpportuniBes in Aerial PlaHorms 1. Larger UAS should be used when flexibility and re- configurability is required. A range of sensors for various biosecurity tasks may be needed and/or new applicaaons may emerge. 2. For frequent and/or long- term opera6ons (high temporal resolu6on), mul6ple plaborms of the same type (fleet of UAS) and/or medium to high end products should be acquired. 3. Fixed wing plabormsà broad area (disanct, regional). Rotary wing aircrah à localised, close proximity coverage (plant, crop, ports)

26 Challenges and OpportuniBes in Flight Control, Guidance & CoordinaBon 1. OpportuniAes exists on directed dynamic re- configurable control and guidance, using sampled data 2. OpportuniAes exists on advanced control and guidance to improve image rec6fica6on and coverage whilst ensuring effec6ve sampling can be achieved each flight. This may require addiaonal onboard sensors. 3. OpportuniAes and challenges exist toward effec6ve mission/path planning targeted at ensuring unbiased sampling

27 Challenges and OpportuniBes in Sensors 1. Imaging: Research aimed at improving imaging sensor quality alone should not be undertaken in the plant biosecurity framework. Research should be directed toward using visual imagery as a complimentary sensor to improve pest detecaon, localisaaon and mapping, as well as augment UAS guidance and control. 2. Hyperspectral Imaging: Research should be directed at acquiring pest specific in- field data sets with UAS+sensor throughout the growing season. 3. Multispectral Imaging: Research should be directed at developing interchangeable filters for in- field tesang of mulaple pests or diseases of different crops. AddiAonally, pest specific in- field data sets with UAS +sensor should be acquired throughout the growing season

28 Challenges and OpportuniBes in Data Processing & Transmission 1. Research should be directed toward parallelizaaon of image processing and control tasks using appropriate hardware such as field programmable gate arrays (FPGA) and Graphical Processing Units (GPU). Internet/ Cloud Crowd Sourcing UAV Farmer UAV Ground Station 2. Research should focus on using UAS based data sets to help quan6fy the sensing and plaborm requirements specific to robust detec6on, localisa6on and discrimina6on of biosecurity threats. Biosecurity Community Broader Community Plant Biosecurity Laboratory / Data Processing Centre

29 Conclusions UAVs can and are revoluaonising Plant bio- security and plant pest surveillance Careful selecaon of Sensor- UAV match is needed for cost effecaveness and usability RegulaAons should not be a deteterrant For more details please see link to PBCRC report (70 pages) hpp:// We are inerested in colaboraaons: More InformaAon, please [felipe.gonzalez@qut.edu.au] Acknowledgements PBCRC supporang this research and Jacqueline (Jax) for arranging and se{ng up this web presentaaon Thank you QuesBons?