Interview with Alison Ballance, Radio National New Zealand - click on the link below...
Labels: AUT Fox, Data collection, NIR camera, Pix4UAV, Skycam, Swampfox, vegetation stress camera
We are now making significant progress with our bush survey and ground truthing. However, each step forward has the potential to raise new problems, some obvious, but others requiring a bit of detective work.
A few weeks ago, we went back to our piece of bush near to the North Shore Model Aero Club strip and flew several transects with our vegetation stress camera. We launched in a 15-20kph wind (as a full-scale pilot I have real difficulty thinking in terms of kph and not knots!) and began the survey, but the aircraft reported wind of up to 30kph. This caused our transects to be 'bent' - the autopilot attempts to arrive at the waypoint, but in a cross wind, the track becomes significantly curved. No problem for the aircraft, but the images must overlap significantly for us to use Pix4UAV to synthesise the orthomosaic.
We decided to bring the aircraft back overhead the GCS and commenced a manual approach, using the GCS computer to bring the A/C down in 50ft increments and the hand controller to steer the Hawk. (Our trainer suggested that manual steering and a lower parachute deployment is a better option than a full auto approach in stronger conditions). In Heading Manual mode, the autopilot maintains airspeed and stabilises the A/C. It appears that shortly after entering Heading Manual mode, the pilot engaged Partial Control mode. This allows the pilot to control both roll and pitch, and airspeed. No problem here - the pilot is an accomplished model aircraft flyer. At an appropriate height the pilot cut the throttle and deployed the recovery parachute. The chute deployed properly and the A/C began its descent. However, close to the ground there was significant turbulence and the A/C began to pitch and roll violently. It landed safely, but the parachute began to drag, cartwheeling the A/C over the rough ground, breaking the main spar and one elevator, and damaging the vertical stabiliser. All this can (and has been ) fixed.
When we analysed the images from the vegetation stress camera, we found that we had more images than were recorded in the photo log file. This is a potential problem, as the images must be geo-referenced for entry into and processing by Pix4UAV.
A quick chat with Dr. Tim Brooks at Skycam revealed something more that we didn't know - when the aircraft is put into Partial Control mode, the camera can continue photographing, but the images are not recorded in the photo log. Problem solved - just match the photographs to the point in the photo log where the last image was recorded and discard the rest.
Our student, Hins, processed the images and we find that we collected a good set, but that there is a strip down the middle of the bush with no images; the result of the curved transects. One solution is to enter intermediate waypoints into the flight path to force the A/C to fly straighter transects.
Major lessons learned: when the aircraft is landing under parachute, the ground team need to get to the touchdown point quickly to secure the aircraft, and when the wind is up around 15 - 20kph and forecast to rise, it's probably better to stay in the UAV Ops HQ and write papers, or if desperate for a flying fix, practice with the flight simulator!
A few weeks ago, we went back to our piece of bush near to the North Shore Model Aero Club strip and flew several transects with our vegetation stress camera. We launched in a 15-20kph wind (as a full-scale pilot I have real difficulty thinking in terms of kph and not knots!) and began the survey, but the aircraft reported wind of up to 30kph. This caused our transects to be 'bent' - the autopilot attempts to arrive at the waypoint, but in a cross wind, the track becomes significantly curved. No problem for the aircraft, but the images must overlap significantly for us to use Pix4UAV to synthesise the orthomosaic.
We decided to bring the aircraft back overhead the GCS and commenced a manual approach, using the GCS computer to bring the A/C down in 50ft increments and the hand controller to steer the Hawk. (Our trainer suggested that manual steering and a lower parachute deployment is a better option than a full auto approach in stronger conditions). In Heading Manual mode, the autopilot maintains airspeed and stabilises the A/C. It appears that shortly after entering Heading Manual mode, the pilot engaged Partial Control mode. This allows the pilot to control both roll and pitch, and airspeed. No problem here - the pilot is an accomplished model aircraft flyer. At an appropriate height the pilot cut the throttle and deployed the recovery parachute. The chute deployed properly and the A/C began its descent. However, close to the ground there was significant turbulence and the A/C began to pitch and roll violently. It landed safely, but the parachute began to drag, cartwheeling the A/C over the rough ground, breaking the main spar and one elevator, and damaging the vertical stabiliser. All this can (and has been ) fixed.
When we analysed the images from the vegetation stress camera, we found that we had more images than were recorded in the photo log file. This is a potential problem, as the images must be geo-referenced for entry into and processing by Pix4UAV.
A quick chat with Dr. Tim Brooks at Skycam revealed something more that we didn't know - when the aircraft is put into Partial Control mode, the camera can continue photographing, but the images are not recorded in the photo log. Problem solved - just match the photographs to the point in the photo log where the last image was recorded and discard the rest.
Our student, Hins, processed the images and we find that we collected a good set, but that there is a strip down the middle of the bush with no images; the result of the curved transects. One solution is to enter intermediate waypoints into the flight path to force the A/C to fly straighter transects.
Major lessons learned: when the aircraft is landing under parachute, the ground team need to get to the touchdown point quickly to secure the aircraft, and when the wind is up around 15 - 20kph and forecast to rise, it's probably better to stay in the UAV Ops HQ and write papers, or if desperate for a flying fix, practice with the flight simulator!
Labels: AUT Hawk, Pix4UAV, vegetation stress camera
Prof. John B went to Palmerston North last week to conduct acceptance flights of AUT's new Swampfox UAV.
The Fox is totally New Zealand designed and built - the airframe was designed by Skycam UAV Ltd. in conjunction with the University of Queensland, the autopilot was designed by the New Zealand Defence Technology Agency, and the whole aircraft and systems were constructed Skycam UAV. The Fox is a flying wing configuration, rather than a conventional wings plus tailplane combination.
As usual for this time of year, the Manawatu weather was unreliable and rain was forecast for the whole week. However, carpe diem! Monday dawned cloudy and a bit windy, but Prof. John B and Dr. Tim Brooks of Skycam packed the control vehicle and headed out to Mangahao in the Wairarapa to fly in Danger Area 522, which allows Beyond Visible Line of Sight (BVLS) flights up to 2500ft AMSL (approximately 2000ft AGL).
After assembly, which is very straightforward, Tim demonstrated the launch procedure. The Fox can be operated from the same GCS as our Hawk, but the launch handling is a little different. After pre take-off checks are complete, the Fox is attached to the bungee catapult and the Operator enters the Take-off command into the GCS computer. In contrast to the Hawk, the motor does not immediately start, which is initially a bit disconcerting. This is actually a good thing, because the Handler holds the Fox by the trailing edges, which places a very large propeller right between the wrists! When the aircraft is released, the accelerometers in the autopilot sense the launch and the motor is turned on. Rate of climb is around 700fpm. Prof. John B did the second launch; the aircraft is 'bounced' on the catapult and released at the top of the bounce, so that the Fox is pointing about 15 degrees above horizontal.
We had fitted two Sony NEX5-N cameras into the sensor cassette and during the two flights totalling about 50 minutes, we took around 3200 high resolution photographs, half in RGB, the other half in nIR. These will be processed with Pix4UAV software and will be a little bonus for the farmer - we will provide the orthomosaics to the farmer. It appears that the areas of grass grub infestation show up very well in these images.
Recovery of the UAV was straightforward - after the Return command was entered, the Fox returned to the launch point; Landing Mode 4 entered commanded the aircraft to descend to 150ft AGL and fly upwind before releasing the parachute. The parachute allows the UAV to land upside down, thus protecting the sensor window and the winglets.
The remainder of the week was far too windy and wet for further flights, though we did go out again on Tuesday and did some ground testing of the video downlink on the Hawk.
The next time the Fox flies, it will probably be in Antarctica - the UAS is on its way to Christchurch now for transport to the ice.
The Fox is totally New Zealand designed and built - the airframe was designed by Skycam UAV Ltd. in conjunction with the University of Queensland, the autopilot was designed by the New Zealand Defence Technology Agency, and the whole aircraft and systems were constructed Skycam UAV. The Fox is a flying wing configuration, rather than a conventional wings plus tailplane combination.
Dr. Tim Brooks with the Swampfox
As usual for this time of year, the Manawatu weather was unreliable and rain was forecast for the whole week. However, carpe diem! Monday dawned cloudy and a bit windy, but Prof. John B and Dr. Tim Brooks of Skycam packed the control vehicle and headed out to Mangahao in the Wairarapa to fly in Danger Area 522, which allows Beyond Visible Line of Sight (BVLS) flights up to 2500ft AMSL (approximately 2000ft AGL).
After assembly, which is very straightforward, Tim demonstrated the launch procedure. The Fox can be operated from the same GCS as our Hawk, but the launch handling is a little different. After pre take-off checks are complete, the Fox is attached to the bungee catapult and the Operator enters the Take-off command into the GCS computer. In contrast to the Hawk, the motor does not immediately start, which is initially a bit disconcerting. This is actually a good thing, because the Handler holds the Fox by the trailing edges, which places a very large propeller right between the wrists! When the aircraft is released, the accelerometers in the autopilot sense the launch and the motor is turned on. Rate of climb is around 700fpm. Prof. John B did the second launch; the aircraft is 'bounced' on the catapult and released at the top of the bounce, so that the Fox is pointing about 15 degrees above horizontal.
We had fitted two Sony NEX5-N cameras into the sensor cassette and during the two flights totalling about 50 minutes, we took around 3200 high resolution photographs, half in RGB, the other half in nIR. These will be processed with Pix4UAV software and will be a little bonus for the farmer - we will provide the orthomosaics to the farmer. It appears that the areas of grass grub infestation show up very well in these images.
Recovery of the UAV was straightforward - after the Return command was entered, the Fox returned to the launch point; Landing Mode 4 entered commanded the aircraft to descend to 150ft AGL and fly upwind before releasing the parachute. The parachute allows the UAV to land upside down, thus protecting the sensor window and the winglets.
The remainder of the week was far too windy and wet for further flights, though we did go out again on Tuesday and did some ground testing of the video downlink on the Hawk.
The next time the Fox flies, it will probably be in Antarctica - the UAS is on its way to Christchurch now for transport to the ice.
Labels: Pix4UAV, Skycam, Swampfox, test flight, UAV
This week's sortie was totally successful from a data collection point of view.
Our first flight netted 706 RGB+nIR images and the second 471 nIR images. This will keep our research student busy for many hours! These images will now be taken into Pix4UAV and stitched into 3-D orthomosaics.
The students also walked around the area with our hand-held spectroradiometer, taking measurements of various plants and trees. We hope that these measurements can be correlated with the UAV images.
What was the little scare? The Hawk is landed by bringing it back to the launch point and then handing over control to the autopilot to decrease the altitude and deploy a parachute. The autopilot calculates wind speed and direction and then flies upwind to deploy the parachute, after which the aircraft should float down to land close to the take-off point.
On this occasion, the telemetry reported a wind speed and direction that didn't correspond with my measurements on the ground. The aircraft flew overhead and we expected it to do another circuit, but it headed off towards some trees and bush to the west of our flying area. The chute deployed and we had a few moments of 'concern' as it floated towards the trees. In the event, the landing was in a small clearing and we were able to recover the aircraft without needing to climb trees.
Next week, we will be accompanied by a radio reporter, who wants to talk to our students on-site and obtain suitable sound bites for her report. We must make sure there are no further exciting moments and that we sound totally professional!
Our first flight netted 706 RGB+nIR images and the second 471 nIR images. This will keep our research student busy for many hours! These images will now be taken into Pix4UAV and stitched into 3-D orthomosaics.
The students also walked around the area with our hand-held spectroradiometer, taking measurements of various plants and trees. We hope that these measurements can be correlated with the UAV images.
What was the little scare? The Hawk is landed by bringing it back to the launch point and then handing over control to the autopilot to decrease the altitude and deploy a parachute. The autopilot calculates wind speed and direction and then flies upwind to deploy the parachute, after which the aircraft should float down to land close to the take-off point.
On this occasion, the telemetry reported a wind speed and direction that didn't correspond with my measurements on the ground. The aircraft flew overhead and we expected it to do another circuit, but it headed off towards some trees and bush to the west of our flying area. The chute deployed and we had a few moments of 'concern' as it floated towards the trees. In the event, the landing was in a small clearing and we were able to recover the aircraft without needing to climb trees.
Next week, we will be accompanied by a radio reporter, who wants to talk to our students on-site and obtain suitable sound bites for her report. We must make sure there are no further exciting moments and that we sound totally professional!
Labels: AUT Hawk, Data collection, Pix4UAV, UAV
We are heading out to fly the Hawk UAV to collect nIR and nIR+RGB images of our current study site. We flew last week but for some reason when we changed filters on the camera we shut down the camera and it did not restart to capture any images on our second flight! Frustration was felt by the entire team, but with frustration comes learning! We have now modified our protocol to download images after every flight and to reset the camera each time. The pre-flight check list now includes a test firing of the shutter. We are starting to build an extensive image library and we are looking forward to further work with the Pix4UAV mosaicing software.
Today we are taking two of our master thesis students into the field with us. They will be using the handheld spectroradiometer to collect reflectance data for several species of native trees.
Today we are taking two of our master thesis students into the field with us. They will be using the handheld spectroradiometer to collect reflectance data for several species of native trees.
Labels: AUT Hawk, Data collection, NIR camera, Pix4UAV
Following yesterday's successful sortie, I have come across another problem - data overload.
Last night I began to process the RAW data from the camera, converting the 450 or so images to TIFF format for processing in Pix4UAV. Each TIFF file is about 48MB, so, together with the original files, each flight generated approximately 18 GB of data. Not surprising that my computer warned that the hard drive was almost full. Time to call in a large external hard drive!
This raises another issue: besides storage capacity, the computer used for analysing and processing the images into orthomosaics will need to have lots of RAM and a high speed processor.
One thing is clear: in a single week of flying, the UAV Team will be able to submerge researchers in data.
Last night I began to process the RAW data from the camera, converting the 450 or so images to TIFF format for processing in Pix4UAV. Each TIFF file is about 48MB, so, together with the original files, each flight generated approximately 18 GB of data. Not surprising that my computer warned that the hard drive was almost full. Time to call in a large external hard drive!
This raises another issue: besides storage capacity, the computer used for analysing and processing the images into orthomosaics will need to have lots of RAM and a high speed processor.
One thing is clear: in a single week of flying, the UAV Team will be able to submerge researchers in data.
Labels: Data collection, Pix4UAV, UAV
Today we passed a major milestone. We went into the field once again and flew the same transects as last time. The firmware fault in the autopilot had been corrected by our suppliers, Skycam UAV, and we were keen to collect a set of data that could be analysed by our Masters student.
Because the strip of bush that we are surveying is only 1.4NM from a commercial airfield, we need the agreement of the aerodrome operator to fly. Professor John B wrote a draft Memorandum of Understanding and submitted it to the CFI of the aerodrome operator, an established aero club. Dr. John and John B went to visit the CFI and the club manager on Tuesday to discuss the form and content of the MoU and came away with an agreed form.
Today's operation was set up in accordance with the MoU; the proposed flight path, times of operation and an annotated Google map of the area was sent to the CFI 72 hours in advance.
After arrival at the operating site, the CFI was contacted to confirm the the operation would take place. The aircraft was very carefully checked against our standard pre-flight check list, as it had not been test flown after the firmware update. All checked out and the planned track was loaded into the autopilot. The Sony NEX5-N camera was installed in the payload bay and the autopilot was set to trigger the camera every 1.8 seconds. After a short period of monitoring 118.00MHz, the aerodrome circuit frequency, a general broadcast was made to indicate that the Hawk would be launched imminently.
The launch went without incident, the GCS announced that the Hawk had launched successfully and she headed off, climbing rapidly to our operating altitude of 500ft AMSL. After the Hawk reached altitude, Navigation Mode was commanded and she headed off on the pre-programmed track. The camera was turned on and the Hawk made three transects of the bush, returning to the launch point at the end of the track. Auto-land was commanded and, after a couple of descending circuits, the parachute deployed and Hawk landed gently. We have still not cured her of her predilection for fences, but she missed it, just! (This problem is exacerbated by changing wind speed during the descent after parachute deployment, but the landing is probably more gentle than a conventional approach and slide across the ground).
With some trepidation, we removed the memory card from the camera and downloaded the images to Prof B's computer. Magic! There were 239 images on the card.
An infrared bandpass filter (720nm up) was fitted to the camera and a second flight was organised. This too went off without incident and another 213 images, this time nIR were collected.
A call was made to the aerodrome to state that the Hawk UAV operation was complete, the A/C and ground station were packed up and we headed back to Auckland Central.
There is a lot of work to do, analysing the 452 images with Pix4UAV, and we hope to repeat the operation in about a week, this time carrying our vegetation stress camera over the same flightpath,
There is quite a lot of preparation for our next sortie - charging of batteries, issue of intent to fly, arranging for the team to be available at the same time (not so easy during semester), booking the vehicle, and cleaning the mud off the equipment; it is still winter here!
Infrared Image of the bush from about 350 ft
Because the strip of bush that we are surveying is only 1.4NM from a commercial airfield, we need the agreement of the aerodrome operator to fly. Professor John B wrote a draft Memorandum of Understanding and submitted it to the CFI of the aerodrome operator, an established aero club. Dr. John and John B went to visit the CFI and the club manager on Tuesday to discuss the form and content of the MoU and came away with an agreed form.
Today's operation was set up in accordance with the MoU; the proposed flight path, times of operation and an annotated Google map of the area was sent to the CFI 72 hours in advance.
After arrival at the operating site, the CFI was contacted to confirm the the operation would take place. The aircraft was very carefully checked against our standard pre-flight check list, as it had not been test flown after the firmware update. All checked out and the planned track was loaded into the autopilot. The Sony NEX5-N camera was installed in the payload bay and the autopilot was set to trigger the camera every 1.8 seconds. After a short period of monitoring 118.00MHz, the aerodrome circuit frequency, a general broadcast was made to indicate that the Hawk would be launched imminently.
The launch went without incident, the GCS announced that the Hawk had launched successfully and she headed off, climbing rapidly to our operating altitude of 500ft AMSL. After the Hawk reached altitude, Navigation Mode was commanded and she headed off on the pre-programmed track. The camera was turned on and the Hawk made three transects of the bush, returning to the launch point at the end of the track. Auto-land was commanded and, after a couple of descending circuits, the parachute deployed and Hawk landed gently. We have still not cured her of her predilection for fences, but she missed it, just! (This problem is exacerbated by changing wind speed during the descent after parachute deployment, but the landing is probably more gentle than a conventional approach and slide across the ground).
With some trepidation, we removed the memory card from the camera and downloaded the images to Prof B's computer. Magic! There were 239 images on the card.
An infrared bandpass filter (720nm up) was fitted to the camera and a second flight was organised. This too went off without incident and another 213 images, this time nIR were collected.
A call was made to the aerodrome to state that the Hawk UAV operation was complete, the A/C and ground station were packed up and we headed back to Auckland Central.
There is a lot of work to do, analysing the 452 images with Pix4UAV, and we hope to repeat the operation in about a week, this time carrying our vegetation stress camera over the same flightpath,
There is quite a lot of preparation for our next sortie - charging of batteries, issue of intent to fly, arranging for the team to be available at the same time (not so easy during semester), booking the vehicle, and cleaning the mud off the equipment; it is still winter here!
Labels: AUT Hawk, check list, Data collection, Pix4UAV, UAV
The AUT Hawk Team has just acquired the Pix4UAV software package. This elegant software is specifically designed for use with small UAVs and enables the stitching together of image sets to produce orthomosaics and 3-D images that can be manipulated to examine terrain surfaces.
For such a complex piece of software, it is surprisingly easy to run.
Over the last few days, a Masters research student has been processing images collected in NZD 522. Barbara and Prof John B have also worked with the software.
There have, of course, been a few teething problems. The format of the images collected by our Sony NEX5-N is not compatible with Pix4UAV and the photo log from the autopilot needs to be in the form of a space-delimited text file. However, we have produced a couple of mosaics with some success. Further tuning of our inputs is required before we can produce the sort of 3-D images shown in a previous post on the potential of our Hawk UAS.
For such a complex piece of software, it is surprisingly easy to run.
Over the last few days, a Masters research student has been processing images collected in NZD 522. Barbara and Prof John B have also worked with the software.
There have, of course, been a few teething problems. The format of the images collected by our Sony NEX5-N is not compatible with Pix4UAV and the photo log from the autopilot needs to be in the form of a space-delimited text file. However, we have produced a couple of mosaics with some success. Further tuning of our inputs is required before we can produce the sort of 3-D images shown in a previous post on the potential of our Hawk UAS.