Reports
Contents
| Title: | Camera comparison at West Pond | ||||||||||||||||||
| Date: | 2023-01-19 - 2023-08-25 | ||||||||||||||||||
| Data File: | WP_RGB_2023.csv WP_RGBavg_2023.csv |
||||||||||||||||||
| Refers to: | WP,RPI0W-006 | ||||||||||||||||||
|
Starting in Jan. 19, 2023, we installed a Raspberry Pi camera (Picam) to complement the existing Autocam (Canon) camera that has been running forever there. Both cameras are installed on the same post, facing west. The Autocam and the Picam have a dynamic white balance that the camera adjusts according to the light conditions. The Picam also took fixed white balance (red=1.8, blue=1.8) photos, but we haven't analyzed those since it wasn't a big improvement over the auto white balance photos. The main issue with the Canon camera is that it does not restart automatically after a power failure. Additionally, in the wintertime the Canon camera enclosure windows gets a lot of condensation which blocks the view of the cameras. In 2023 we've had multiple power failures and are missing several blocks of Canon camera data (10 days in April, a week in May, a week in June, 3 weeks in July, no data in September). My objective is to compare the picam and Canon camera data to see if the picam data can be substituted for Canon data and what regression equations should be used. For this analysis, I defined midday as 11:15-13:15, inclusive. Summary: I uploaded picam data to the database starting 2023-01-19. Picam and autocam data are pretty close, although the picam has more noise before green-up in early May. I tried regressions using both individual data points (5 data points/day, 11:15-13:15 inclusive) and with the daily average (11:15-13:15 average), and they were pretty much identical. I don't think applying the calibration makes the picam data any more similar to the Canon data. In fact, applying the calibration actually makes the Picam daily average GCC more different than the Canon daily average GCC at the peak of the growing season in August (Fig 8). So, I don't recommend applying any calibration to the picam data. Figure 1. Time series of GCC. Five points are plotted per day, 11:15-13:15 inclusive. The Picam data is noisier before green-up and the peak is slightly lower than Canon GCC peak. Overall both cameras have GCC that actually match pretty well. Figure 2. Time series of midday average GCC. Overall both cameras have GCC that actually match pretty well. The Picam GCC is noisier before green-up and the peak is slightly lower than Canon GCC peak. Figure 3. Time series of midday Red data Figure 4. Time series of midday Green data Figure 5. Time series of midday Blue data
Figure 6. Linear regression of Red, Green, and Blue using 5 data points per day (11:15, 11:45, 12:15, 12:45, 13:15)
Figure 7. Linear regression of Red, Green, and Blue using a midday average Figure 8. Picam GCC where Red, Green, and Blue have been calibrated with the linear regression results using 5 data points per day Figure 9. Picam GCC where Red, Green, and Blue have been calibrated with the linear regression results using a midday average
|
|||||||||||||||||||
| |