Question: First Infragram cam trials - Primeros ensayos con cámaras Infragram

ivalethia asked on September 20, 2014 04:23
3,191 | 0 answers | #11165


What I want to do

Quisiera poder analizar la salud y crecimiento de distintos cultivos, en sistemas protegidos y otros al aire libre. Para esto dispongo de tres cámaras: Sony DSC-P41, HTC Sensation (teléfono celular), Vstarcam T7837WIP.

My attempt and results

Con la primer cámara tuve problemas, y aun un poco, para alcanzar un enfoque considerable, lo que ven abajo en las imágenes es la mejor calidad alcanzada hasta ahora. Desde que retire el lente IR Cut, se perdió/disminuyó la calidad de enfoque (Imágenes: 1 y 2). Con la segunda, el enfoque se ha perdido por completo, tendré que cambiar el modulo completo y volver a intentar retirar el filtro IR sin fastidiar el autoenfoque (Imágenes: 3-13). Con la tercer cámara, el enfoque es bueno y con las características de un Cámara IP, mejor aun. El problema ahora radica en cómo controlar el cambio de filtros ya que responde al sensor de luz preinstalado en la cámara y hasta ahora no he encontrado una forma de incluir un comando o "botón" (algún script) para las aplicaciones existentes ya para su control en cualquier sistema operativo (Windows, Android, etc.) (Imagen: 14). El procesamiento de imágenes es mediante GIMP con el método que mencionan en esta pagina. El filtro usado en todas las cámaras es negativo de película fotográfica velada.

Questions and next steps

Cómo puedo mejorar el enfoque? Porque se perdió desde que retire el lente IR Cut? Cómo puedo hacer manual el cambio automático del filtro en la Cámara IP? Alguien tiene un HTC Sensation del que me pueda regalar/rematar el modulo de la cámara? Alguien mas ha modificado la cámara de algún teléfono celular? Pueden compartir sus experiencias? Quisiera mejorar todos estos puntos y ademas seleccionar una mejor escala de colores para hacer el degradado en GIMP, ya que, aunque visualmente se puede decir que hay una diferencia de las plantas con el resto, cuando uno revisa la escala cromatica vemos que en realidad lo que resalta no está cayendo en lo que supondríamos debería ir del 0.1 al 0.9, sino que incluso es inferior a 0. Tal vez no este entendiendo del todo como funciona la escala, ojala puedan ayudarme con esto.

Why I'm interested

Creo que el uso de cámaras infrarrojas y redes inalámbricas, podría convertirse en un buen apoyo para el monitoreo a distancia de cultivos a gran escala y al realizarlos uno mismo su precio seria bastante accesible.

Saludos.

SONY-DSC-P41-LANDSCAPE.1_1_.jpg

SONY-DSC-P41-LANDSCAPE.2_1_.jpg

HTC-SENSATION-INVERNADEROCHILE.2_1_.jpg

HTC-SENSATION-INVERNADEROJITOMATE.3_1_.jpg

HTC-SENSATION-INVERNADEROJITOMATE.1_1_.jpg

HTC-SENSATION-INVERNADEROCHILE.3_1_.jpg

HTC-SENSATION-INVERNADEROJITOMATE.2_1_.jpg

HTC-SENSATION-INVERNADEROCHILE.1_1_.jpg

HTC-SENSATION-LANDSCAPE.2_1_.jpg

HTC-SENSATION-LANDSCAPE.1_1_.jpg

HTC-SENSATION-MOSS.1_1_.jpg

HTC-SENSATION-NOGAL_1_.jpg

HTC-SENSATION-PLANTACHILE_1_.jpg

VSTARCAM-T7837WIP.1_2_.jpg



9 Comments

Gracias por esto! Muy interesante!


Replacing the camera's IR block filter with exposed, developed, color print film allows only near infrared light to be captured. These IR photos are not Infragram photos and cannot be transformed into NDVI or false color IR images for analyzing plant health. To make an Infragram camera, the IR block filter can be replaced with a red filter like Wratten 25A or Rosco #19. Producing meaningful NDVI images directly from Infragram photos requires that the camera has a custom white balance feature. Most web cams and cell phone cameras probably do not have this feature.

Chris


Hi Chris (@cfastie). This pictures (the third in every serie of 3) are not produced directly from the camera, as you say. First, I take a picture without any filter (the first picture in the serie of 3); then, I take another picture with the exposed film as filter (the second picture in the serie of 3); finally, with these 2 pictures align, in GIMP, I do exactly as @Warren says it here: http://publiclab.org/notes/warren/10-27-2011/video-tutorial-creating-false-color-ndvi-aerial-wetlands-imagery. Resulting in the third picture of every serie. So, it is really necessary do it with a custom white balance? Maybe I don't fully understand yet how the NDVI works, but it cannot be calculated with this filter? Why? Greetings!

Is this a question? Click here to post it to the Questions page.


Okay, I think I understand now. What you did is:

  1. remove IR block filter from a camera
  2. take a photo
  3. take the same photo with exposed and developed color print film in front of the lens
  4. align the two photos and make NDVI

If that is what you did, there is one problem. For NDVI you want two photos (or channels in a photo): one with only visible light and one with only near infrared light. Your photo with the film filter is a good near infrared photo (the film blocks most visible light), but the photo with no filter has visible and NIR light mixed in all three channels. So you would need two cameras: one unmodified so it takes normal visible light photos, and one with the IR block filter replaced with film (or Wratten 87). Or you would need another filter, just like the one you removed from inside the camera, to put in front of the lens so normal photos could be taken again. Then to make traditional NDVI, the red channel in the normal photo is used for visible light and any of the three channels in the IR photo is used for NIR.

The white balance setting on the NIR camera does not matter much. If you do a custom white balance while filling the frame with a gray surface, all three channels will be the same and the photo will be grayscale. Without the white balance the photo will look pink, but that is just because the red channel is a little brighter than the others. Any of the three channels can be used for NIR. White balance is more important when the filter used is intended to make a photo with visible light in one channel and NIR light in another. For example a Wratten 25A filter allows capture of NIR in the blue channel and mostly red light in the red channel.

An easier way to make NDVI from two photos is to use Fiji (ImageJ) with the Photo Monitoring plugin written by Ned Horning. The two photos are automatically aligned and a new image is created with NDVI computed for each pixel.

Chris


After add the IRCut filter in the IPCam, with the GIMP (@Warren) method, I get this:

VSTARCAM_T7837WIP_TRIGO.GERANIO.ROSASECA_NDVI_1_.jpg

It looks like go better according the color scale (except the pot), the living parts respond to the expected. Now I will try with ImageJ and the plugin, maybe that way the shadows could be eliminated (the one at the right of the pot). Thanks for the advise, not only I correct the filter issue, even I clarify the doubts that I had but didn't do explicit. Greetings. Ivan.


Hi Chris (@cfastie). Can you help one more time? I can't make it work with the plugin in Fiji (also in ImageJ), could you put it in simple words how to install and run it? Thanks. Ivan Morales.

Is this a question? Click here to post it to the Questions page.


Let us know which parts of the user guide (PhotoMonitoringPluginGuide.odt, at Github) you need help with.

Chris


I belive the problem is in the first plugin ("Create dual image list" ), 'cause I don't understand from where this file (list) comes, do I create it and then work with it in Fiji? Or the plugin fully creates it? Could you put an example of how you do it in your computer? Ivan.

Is this a question? Click here to post it to the Questions page.


Ivan, Yes, the plugin to make the file pairs list requires some study. It also requires that you have:

  1. synchronous pairs of VIS and NIR photos
  2. all the VIS photos in one directory and the NIR photos in another directory.
  3. not too many pairs (I think I had trouble with more than 200)
  4. accurate time stamps in the EXIF headers (otherwise it tries to use the file date)
  5. pairs with time stamps that are close to each other (this should not matter but seems to make it work more often)

If you run the plugin and enter the information at the prompts (some obscure at the upper right of the windows) it might work and create a new text file with the filenames and paths of all the pairs. There are three ways to help the plugin find the matches for each pair. If one does not work, another might. If your photo pairs cannot be identified by their time stamps (as is probably the case with yours) you might have to make the text file of the pairs by hand. This is not hard for just a few pairs, but the format of the text file has to be perfect.

It would be good for Ned to add a feature to make the text file from two directories of files, so all you have to do is change the filenames so the two directories sort with the files in the proper order. Then it could work for just a few pairs that might not have matched time stamps.

Chris


Log in to comment

0 Answers

Sign up or Login to post an answer to this question.