1 of 15

BoSLcam�an IoT field-deployable camera enabling image-based water quality monitoring

S. E. Catsamas1,2, B. Shi1, M. Wang1, C. Thirkell1, D. T. McCarthy2

1BoSL Water Monitoring and Control, Monash University

2School of Civil and Environmental Engineering, Queensland University of Technology

​

5.2e 11 June 16:30 – 18:00

1

Funding: Monash University and QUT

Monash University &

Queensland University of Technology

2 of 15

Stormwater Monitoring & Maintenance

  • Waterway/ecosystem health
    • Sensors and autosamplers
  • Illicit-input detection
  • Blockages/ragging
  • Stormwater management treatment systems
    • Gross-pollutant traps
    • Wetland/biofilter plant health

​

​

  • Routine inspection/maintenance
    • Costly
    • Confined space entry
    • Coupled maintenance + inspection
  • Low-cost monitoring systems

​

5.2e 11 June 16:30 – 18:00

2

Monash University &

Queensland University of Technology

3 of 15

Imaged-based Sensors

  • Monitoring
    • Out-of-water:
      • Robust, easy installation, lower maintenance
    • Measurement based on image analysis:
      • Turbidity (Chai et al., 2017; J. Huang et al., 2021)
      • Water depth
      • Velocity (Watanabe et al., 2021)
      • Ecological diversity (Weinstein, 2018)
    • Compliment in-water sensors

​

  • Maintenance
    • Real-time view on infrastructure
      • Identify concerns as they occur
      • Decreased downtime and reduce maintenance

​

  • Commercial field cameras
    • Expensive ( > 200€)
    • Known theft target
  • Low-cost camera platform desired

5.2e 11 June 16:30 – 18:00

3

Chai, M. M. E., Ng, S. M., & Chua, H. S. (2017). AIP Conference Proceedings, 1808(1), 020014.

Huang, J., Qian, R., Gao, J., Bing, H., Huang, Q., Qi, L., Song, S., & Huang, J. (2021). Water Research, 202, 117406.

Watanabe, K., Fujita, I., Iguchi, M., & Hasegawa, M. (2021). Water, 13(15), Article 15.

Weinstein, B. G. (2018). Journal of Animal Ecology, 87(3), 533–545.

Monash University &

Queensland University of Technology

4 of 15

BoSLcam

5.2e 11 June 16:30 – 18:00

4

Monash University &

Queensland University of Technology

5 of 15

BoSLcam Hardware

  • OV7675, VGA: 640x480 colour images
  • Inbuilt LED flash: 3 m range
  • nRF9160 MCU+LTE modem
    • Cellular connected: CAT-M1, NB-IoT
    • GPS
    • RTC
  • SD card
  • UART interface
  • 79 mm x 15 mm
  • Low sleep current < 400 µA
  • Low power: 3.7 V Li-ion battery
    • 11200 mAh battery = ~ 7000 images (VGA w/o flash), � ~ 1 month @ 6 min upload
  • Open-source
  • Total cost: < 50€

5.2e 11 June 16:30 – 18:00

5

Monash University &

Queensland University of Technology

6 of 15

BoSLcam Software

  • Config via SD card: image size, capture frequency, flash usage
  • Auto exposure, auto white balance, & auto gain
  • JPEG compression: 80% file size reduction, saves battery and data costs
  • FTP image upload (~ 60 s for VGA JPEG)
  • Image metadata: date/time, total images, error codes, battery level
  • (Advanced) user programmable in C
    • Zephyr real time operating system
  • WIP: Arduino IDE integration

5.2e 11 June 16:30 – 18:00

6

FTP server

Monash University &

Queensland University of Technology

7 of 15

BoSLcam Applications

5.2e 11 June 16:30 – 18:00

7

Monash University &

Queensland University of Technology

8 of 15

Sensing: Turbidity

  • Stony creek stormwater drain
    • Suburban + industrial catchment

​

  • BoSLcam (w/lens) & in-water turbidity sensor

​

5.2e 11 June 16:30 – 18:00

8

2.6a. 14 June. 9:00-10:30. M. Wang.

Detecting the Source of Illegal Discharges by Applying Low-Cost Sensors

Monash University &

Queensland University of Technology

9 of 15

Sensing: Turbidity

  • in-water turbidity time series
    • Large turbidity event measured
    • T < 04:00 faulty sensor
      • Faulty sensor replaced
      • T > 04:00 new sensor
  • What did the BoSLcam see?

​

  • Image turbidity estimation
    • Mean grey-level in blue rectangle

5.2e 11 June 16:30 – 18:00

9

sensor replaced

Monash University &

Queensland University of Technology

10 of 15

Sensing: Turbidity

5.2e 11 June 16:30 – 18:00

10

  • Correlation
    • high linearity (T > 04:00)
    • R2 = 0.93

​

  • Linear fit for T > 04:00
    • Map grey-level to NTU

​

  • In-water alarm:
    • EPA crew dispatched 2h late
    • Source not found
  • May have been caught with BoSLcam alarm
  • BoSLcam: greater reliability
  • Demonstrates use-case of out-of-water sensors

y = 0.93 x + 11.1

Monash University &

Queensland University of Technology

11 of 15

Field Awareness

  • Vegetation
  • Vandalism
  • Flooding

​

  • Debris & ragging
  • Sensor installation
  • Dry weather flows

5.2e 11 June 16:30 – 18:00

11

wetland�+ RTC

Monash University &

Queensland University of Technology

12 of 15

Ecosystem Monitoring

  • Algae blooms
  • Species monitoring
    • AI automatic bird counting
    • Migrate to BoSLcam
      • Higher resolution desirable

5.2e 11 June 16:30 – 18:00

12

Monash University &

Queensland University of Technology

13 of 15

Summary

BoSLcam:

    • Field deployable
    • Low cost (< 50€), low power
      • 11200 mAh battery = ~ 7000 images (VGA w/o flash), ~ 1 month @ 6 min upload
    • Cellular image upload
    • Integrated flash
    • Open source

Applications:

  • Out-of-water sensing
    • Reliablity
  • Field awareness
    • Improve maintenance schedules
  • Ecosystem monitoring

​

5.2e 11 June 16:30 – 18:00

13

Monash University &

Queensland University of Technology

14 of 15

Future Work

  • Hardware:
    • Resolution upgrades
    • Lenses
    • Multi-spectral imaging (UV, IR)
  • Software:
    • Battery life
    • Video
    • Imaged-based alarms
    • AI analysis (onboard or on cloud)
  • Applications:
    • Sensing: water depth, velocity, illicit-discharge detection
    • Alarms: increase capture frequency for predicted rainfall
    • System integration: triggering RTC events & autosampling

​

5.2e 11 June 16:30 – 18:00

14

Monash University &

Queensland University of Technology

15 of 15

More information

5.2e 11 June 16:30 – 18:00

15

�BoSL presentations upcoming at ICUD:�

5.2b. 12 June. C. Pang:�A picture is worth a thousand measurements: a low-cost camera-based sensor for in-situ particle size estimation

5.2b. 12 June. B. Shi:�Next generation monitoring protocols to improve the load estimation accuracy

5.3b. 12 June. C. Thirkell:�Component Testing of Real-Time Control in a Stormwater Constructed Wetland for Pathogen Reduction

5.5b. 13 June. M Janmohammadi:�Transferable Machine Learning methods for Predicting nitrate in diverse catchments��2.6a. 14 June. M. Wang:�Detecting the Source of Illegal Discharges by Applying Low-Cost Sensors

5.2d. 14 June. M. Wang:�Identifying wetland weaknesses using high-spatial resolution and low-cost water quality sensing methods

BoSL website:

https://www.bosl.com.au/

​

Wiki:� https://www.bosl.com.au/wiki/BoSLcam

Github:� https://github.com/Monash-BoSL/BoSLcam

Hardware:

Talk to me at ICUD!

Monash University &

Queensland University of Technology