Presentation slides of 30 minute lecture about machine vision camera as a sensor.
Applicable in industry, traffic, agro and more.
The value of a sensor-network to help convert data into information, projected on automation in AgriCulture / Agribotics.
1. Intelligent Sensor Networks Conference
Ron van Dooren, Technical Sales & Marketing
“What do eye see?”
Machine vision solutions in Smart Industry
2. How is an Intelligent Sensor Network defined?
Intelligent:
Psychological term to describe the ability to notice differences and similarities, to orient in spaces,
to reason, to plan, to solve problems, understand ideas and languages, store information in
memory and retrieve it later, to learn from experiences.
Sensor:
Man-made device, often mechanical, electronical, by software or virtual, to fulfill the function
of biological senses.
A sensor measures a physical property and converts it into a standardized signal.
e.g. making a water-temperature from 0-100 °C result in a current from 4-20 mA.
Network:
Two or more sensors linked together
3. Narrowing down to one sensor type
Measured property:
Waves.
Output:
Bits. (Gigabits)
Camera:
Derived from Latin, meaning “room”, is a device where light can enter in a controlled way and leave an imprint on the wall.
In digital industrial camera’s, the light actuates pixels on a chip, resulting in a stream of bits to be interpreted by an external
software device.
4. Why sensors
A machine can experience (or sense) its environment by use of sensors.
In process-control, sensors provide real-time process-information to the controller to determine if any adjustments are required.
5. Why good process-control:
The driver’s eyes are sensors, needed to operate the actuator (steering wheel) to stay in the lane
1. Have a safe trip, no worries about the safety on-the-way
2. Get the right outcome at the end of the trip
3. Repeatability in next trips
Three reasons to want good process-control
6. From sampling to 100% process-control …..
Sometimes it is enough to only look at the beginning
More often we want better statistic data to gain our trust.
High resolution camera’s will record the scene in an instant [µsec] and a powerful PC ensures the software algorithm
analyzes the image in a glimpse [msec]
….. it’s possible nowadays
and the end…..
7. Two hardware approaches to achieve this
Industrial camera, cable connection, software on a pc, output through a digital I/O board
Smart-camera, software in onboard fpga, output
8. Stand Alone or Network
Sensors working together, examples in the field.
Fugro / Raildata scan HSL- railtrack
• 2 laser triangulation scanunits, accurate speed & GPS, inclinometer
• Forming a 3D image of a traintrack, recorded at 80 km/h
• Additional 2D imaging
• Processing of the 3D point-cloud and compare to previous to see the areas where track needs maintenance.
9. Stand Alone or Network
Robots acting on 3D information
• Random bin picker, cheese slicer
• Forming a 3D image of a bin with parts
• Processing of the 3D pointcloud and identify best possible pick.
• Telling the robot x,y,z co-ordinates, the a,b,c rotation and required tool-type to pick the part
10. Our vision on sensor networks: agribotics
Connected sensors can deal with more parameters at the same time and provide the software with multiple inputs.
Examples in the near future
Cucumber harvesting
• multiple laser triangulation scanunits,
• 2D colour imaging overlay
• Ambient sensors for moisture, temperature, CO₂, etcetera
• Forming a 3D image of a whole greenhouse
• Forming a map of ambient conditions in the greenhouse
• Processing of the 3D point-cloud where to pick the cucumber.
• Detect anomalities in leaf-colour
• Predict outbreak of fungus by analyzing multiple parameters
• Prevent outbreaks by disturbing one of the key parameters
(temperature, moisture)
• Actually, it is a data-harvester which can also pick a cucumber
11. Our vision on sensor networks: agribotics
Connected sensors can deal with more parameters at the same time and provide the software with multiple inputs.
Examples in the near future
Weed removal
• 2D colour imaging
• Multi spectral imaging overlay
• Looking at the same image at different wavelengths instantaneously
• Compare the images and find details normally invisible
• High resolution GPS-RTK
• Ambient sensors for moisture, temperature, CO₂, etcetera
• Self-driving
• Distinguish green weed from green crop
• Forming a map of ambient conditions on the field
12. Our vision on sensor networks: logistics
Traffic monitoring
• Automated Number-Plate Recognition (ANPR) relates to databases and suspect behavior
• Pattern recognition on motorways, all traffic cams read by software, only the suspicious get shown to operator
• Automated warning systems : truck-height, low tire-pressure, hot brakes
• Truck platooning, Container numbers tracking in terminal, site-safety
• Smart entry
Warehouse logistics
• Warehouse AGV’s safe-guarding the load dimensions, validating the availability of ‘open’ positions
• Robotic bin picking of parcels and packages
• Robot palletizing of random boxes (weight, dimension, reverse de-palletizing order)
13. Wrap up
Available technology allows for 100% control
Applicable to any process, indoor or outdoor
Collecting big data of combined properties is valuable
Agribotics is an emerging market with high demand of sensors
Multi-spectral imaging will grow strongly
14. Ron van Dooren
Technical Sales & Marketing
Beltech.nl
Thanks for your attention!
Crux Agribotics, Vimec, Beltech & Smart Vision Center are members of