Vision can be a very complex problem and when you add in the requirements and restrictions from the multirotor it can be very hard to find a suitable algorithm.

• Requirements of system
• Capable of dealing with outdoor conditions during the day
• Global and local illumination changes
• Due to clouds, shadows (both copter and objects), changes in time of day and so on. This can really ruin color based approaches if they have assumptions about the background(which they tend to do)
• High amounts of IR
• The sun is a big ole ball of IR so IR based approaches don't tend to work during the day without some heavy assumptions or very powerful IR leds
• Cluttered environment
• In order to remove any assumptions about the operating environment. Which can make model based searches fail if similar objects exist in the scene
• Arbitrary backgrounds
• Once again to remove operating assumptions this means that color based approaches can fail if they have assumptions about the background (which they tend to do)
• Capable of dealing with the operating conditions of a multirotor UAV
• Partial occlusion
• Due to the pitching and rolling movement of the copter in order to translate it can cause the image to partially move out of the image.
• Heavy skew
• As above but it causes skew
• Blur
• Depending on a large number of factors such as exposure, sensor quality, vibrations and speed of movement there can be some blurring in the image
• High variability in scale
• The vision system has to be capable of dealing with the copter at as many heights possible from 20cm to 20m away
• Arbitrary changes in orientation
• The camera and copter can be orientated arbitrarly compared to the target so the system has to be capable of dealing with that.
• Needs to operate as quickly as possible as to run in realtime (5hz minimum)
• If its too slow the control will not be capable of dealing with perturbations and compensate for error
• Limited processing power
• Due to the time constraints and onboard computer expensive visual operations are not suitable
• Needs a high detection rate and low false detection rate
• Cant have the copter trying to land somewhere that’s not the target
• Needs a way of determining the confidence of the detection
• This is useful for use in data fusion such as a Kalman filter and determining if there are false detections.

  Approaches tried and some results / notes
  

  • Approaches tried
  • Color blob detection of the centers and using the color to solve the correspondence problem and PNP for pose estimation
  • Color blob detection with cropping(based on looking for the white outline) and PNP for pose estimation

  These approaches worked at about 3 hz or 15 hz with the cropping approach and most of the image is cropped out ,but PNP is very sensitive to pixel coordinates of the detected colours. Also colour based approaches are sensitive to changes in illumination and this approach doesnt allow any occlusion of the target.

  Example of working blob detection (circles are detected centers)
  Example of breakdown of blob detection

  • SIFT / SURF / BRIEF / ORB based template search using RANSAC and RPP

    Example working output of image feature based approach

  Example breaking down output of image feature based approach

  

  • CMT
    Consensus-based Matching and Tracking of Keypoints (CMT) is a new algorithm from a conference this year it works by tracking image features or keypoints between frames and comparing keypoints based on their descriptors. For CMT I altered the landing target to have a checkerboard in the center which gives a higher number of key points to detect at lower heights.

    Examples of CMT working at 2 meters and 15 meters above the ground from the copters downward facing camera. The blue squares are the detected markers and the dots are the matched keypoints.

Output from CMT outdoors with hexacopter 2m above ground
Output from CMT outdoors with hexacopter 15 meters above the ground

Using the free vrep simulation environment and my access to matlab and simulink via the university the simulation model of the copter has been developed and I can work on designing vision based control algorithms without the risk of crashing copters.

The autonomous platform was assembled based on the design documents and all the electronics were validated in the lab. Once the system was confirmed to be working the platform was taken out to the ACFR test area on the Marulan farm.

A number of tests were performed and the pixhawk running the arducopter firmware performed as expected.

Preliminary testing of the visual system was started ,but the software integration requires more work and the addition of more failsafes to deal with fail cases

After considering the options avalible the three best possible solutions which are very close are the udoo, odroidx2 and the BBB.

The odroidx2 is the most powerful processing unit by far it is almost twice as powerful as the udoo and almost 7 times as powerful as the BBB (assuming that all the cores are fully utilised) it also has twice and four times the offering of ram as the others.

The downsides of the odroidx2 is that it does not have tha ardunio envrionment avalible on the udoo but since the FC(pixhawk) is receiving the sensor data it does not need many gpio pins itself. Also the selected sonar sensors have the option to use pulsewidth or serial transfer. It also has 4 uart, 3spi and 8i2c ports avlialbe so there is plenty of interfaces. The biggest downside is lack of good documentation and higher complexity to get things running due to having to deal with the linux to gpio interface.

But odroid x2 + console cable + wifi = much more powerful processing for vision / control modelling

With the pixhawk doing stabilisation/reading sensors and parroting the data over the uart

The selected autopilot will be either an APM or a Pixhawk

If the official APM is used then there is no real benefit to choosing it over the Pixhawk but if the arduflier or some other knock off version is considered instead than it may be worth it financially choosing it over the pixhawk

The final selected autopilot is the pixhawk due to the higher processing power which allows an extended kalman filter to be implemented which can lead to better state estimation. Another benefit of the pixhawk is the extra ports avalible for external sensors to be integrated with the software framework

I have a lengthy design document for the selection and design of the copter system being used in my development. Some select sections have been posted here for documentation purposes and the document will be included with my project. 

(link to lengthier design document)

Project requirements

• The project shall be completed within the budget.
• The system should be able to resolve a 5cm by 5 cm logo.
• The system shall be able to resolve 20 by 20cm objects.
• The system shall fail in a safe manner that will not damage either the system itself nor pose any risk to the user, other people or objects in the vicinity of operation.
• The system shall be capable of imaging an area of 100m by 100m.
• The system should be able to image an area of 400m by 400m.
• The system should be capable of flying at a variable height with a minimum of 5m.
• The system should have minimum cost of upkeep.
• The system should require minimum training to operate.
• The system should be capable of operating outdoors.
• The system shall be capable of operating in light adverse conditions (light breeze).
• The system shall carry a 100g payload.
• The system should have enough battery to image the area in as few trips as possible.
• The system should have enough payload to support future sensor additions.
• The system shall carry a camera, onboard stabilisation, some means of connecting to a computer and be able to store the images.
• The budget shall include the mechanical parts of the craft, control electronics and batteries
• The budget should include radio receiver and camera.
• The system shall include the electronics required to allow autonomous recharging

Requirements due to legality

• The system shall have power monitoring.
• The system shall have a means of transmitting telemetry data in order to inform the operator of the systems operational status.
• The system shall have a means of failsafe.
• Radio transmission power and frequency shall comply with Australian regulations for all transmitters used.
• The system shall failsafe systems like return to base / flight termination.
• The system shall have a means to terminate the current operation and allow user control.
• The system shall be considered a small uav under casa regulations (under 100kg gross weight).
• The system shall not be commercial applications.
• The system shall not operate near structures, people or an aerodrome.
• The system shall not operate at a height higher than 400m.

Notes on copter design

The selection of components for the H550 and Q450 designs required many tradeoffs between price, payload, flight time and availability of components from as few distributors as possible in order to minimise money spent on shipping. Some justifications and reasons for the selections of components were

Propeller selection:

Larger Propellers

Higher thrust/payload capacity

lower flight time

Less responsive

Heavier weight

May burn out motors if they are too large

Limited by frame size

Smaller propellers

Lower thrust/payload capacity

Higher flight time

More responsive

Lighter weight

Useable on high kv motors

Battery selection:

Larger batteries

Higher cost

Higher flight time (with diminishing returns)

Less responsive

Heavier weight

Higher more battery cells

Higher thrust /payload capacity

Less flight time

Smaller batteries

Lower cost

Shorter flight time

More responsive

Lighter weight

Fewer battery cells

lower thrust /payload capacity

more flight time

Motor selection:

Higher KV

Higher thrust/payload capacity

lower flight time

More responsive

Limits propeller size

Higher power

Higher thrust /payload capacity

Less flight time

Able to turn larger propellers

Lower KV

lower thrust/payload capacity

Higher flight time

less responsive

can have larger propellers

Lower power

lower thrust /payload capacity

more flight time

limited in propeller size

There is a delicate balancing act between those three main copter components as seen above so there are a few rules of thumb that have been developed to try and guide designers

Try not to exceed a ratio of more than 0.5*weight to thrust to remain responsive for flight or more than 0.8*weight to thrust to be able to maintain stable flight. Lower weight to thrust ratios result in a more responsive system allowing acrobatics and advanced maneuvers. Hence aerial photography systems usually have a weight to thrust ratio of 0.5-0.8 since the system does not need to be responsive for flight and payloads are normally heavy due to gimbals cameras etc

For long flight times try to have a battery weight to copter weight ratio of 1gram copter and payload for every 2grams of battery you are carrying or even more battery. This is because the motors on copter systems require a large amount of energy to run and deplete batteries very quickly.

For long flight times pancake style motors with low KV, high power and large propellers ,but they are limited in the payload they are capable of carrying and these systems are more expensive as well

Other craft design considered

-Hexacopter, Octocopter and n-copter

Decreased flight time and increased cost compared to quad

Higher lifting capacity for additional future upgrades such as sensors or camera gimbals

Depending on the design there may be more stability in wind

This project does not require large payload capacities

Sometimes able to land safely after losing a motor or propeller mid flight

-Fixed wing

Can have much higher flight times ,but it comes with the trade off of maneuverability

Unable to hover to take still photos

No vtol

-Blimp

Can have large carrying capacity, but needs larger envelope. Can have huge flight time and hover

No off the shelf systems so everything must be custom

Considering the design limitations of the above systems I chose a build a hexacopter system to allow for the overhead of the experimental payload (which could be reduced in weight once sufficiently prototyped).