Get an idea on the spectral sensitivity and quantum efficiency of your camera. The data are provided by the camera manufacturer and/or chip manufacturer. It's easier for monochrome cameras as there is no colour filter (bayer pattern) involved. They are also more light sensitive. CCDs are more sensitive than CMOS and CCD is also more linear. (Your Canon camera has a CMOS sensor). For colour cameras you'll get separate curves of quantum efficiency for each colour.
An important aspect is the colour spectrum you want to observe. Cameras and normal lenses are designed to detect visible light only. That is only an extremely small part of the electromagnetic spectrum.. Any camera, including the monochrome ones, do have an IR filter to disallow any light with wavelength longer than red to pass. These IR filters can be removed from some chips (there are offers for certain cameras). Yet even then, the material of the chips has a cut-off around 1.5µm.
If you are using some optics, get the spectral transmission curve for that as well. Different coatings block different wavelengths. Data are availble from manufacturers - usually they are designed for visible wavelength exclusively.
In order to get a rough estimate of your photon count you thus have to take the pixel value, de-convolve that with the quantum efficiency of the pixel (per colour if you use a colour camera) and de-convolve that with the transmission curve of your optics and then the spectral density of the light source you observe. You might want to make several measurements at different wavelength (small band pass filters) to get a better idea of the spectral content.
If you have a monochrome colour source with known intensity (i.e. laser), you can use that in order to calibrate your guestimate.
In any case: use raw images. Try to get a handle on the actual values per pixel without any colour corrections via white balance etc done.