Hi

In some cases you need to recognize individuals, such as in capture-recapture analysis for abundance/density estimations, or you might want an index of individuals that are detected (instead of number of detection). 

If you need to recognize individuals is better if you have both flanks of the animal photographed, and even better if the animal has natural marks (such as rosettes or stripes like jaguars or tigers), but you can also use scars and general body complexion to help. 
You can still make a capture-recapture study with one camera trap, but you have to choose only pictures of one side of the animal to use, and discard all the other pictures of the other side (since you won't know if are the same animal), doing that you decrease the detection probability, because now it would be the product of the likelihood of the animal walking trough the camera and the likelihood of going in the direction that you need. 

Using one camera per site is a good setup for a site occupancy study, and you can even estimate abundance with the Royle-Nichols method, if your data fits it. 
I hope this helps

Bests

Juan

We have just started testing an array of three cameras recording video at an African wild dog marking site. The idea is to give seamless coverage of an area larger than can be monitored by one camera, with maximum chance of detecting animals approaching from any direction.  As a bonus it produces pictures of both sides. The cameras have only been out for a week, so it is too early to assess how well the arrangement works.

We have a set of three Bushnells covering an African wild dog scent-marking site. They are at the apices of a triangle, with each camera just at the edge of the field of view of one of the others. This gives us a higher probability of detecting animals at the site, a better view of the action on at least one of the cameras, and no dead spots below each camera.

I was hoping to use two Reconyx Ultrafire XR6s on game trails at each of ten scent stations, but the cameras cannot detect animals walking towards or away from them. 

Peter

Here's a set of tools that could be applicable to this idea

 https://github.com/pfr/VideoSpeedTracker

Hi Steph - just to follow up on your post: @MarcusRowcliffe , James Durrant and I have been working on a bit of software to implement the "computer vision" techniques that are mentioned in that paragraph. You can see a demonstration of it in action here. It requires camera-trappers to "calibrate" their camera traps during setup (or take-down), by taking pictures of a standard object (for example, we use a 1m pole held vertically) at different distances. The calibration takes ~10 mins per location. From this, you can reconstruct the paths that animals take infront of cameras, the total distance they travelled, and therefore their speed.

Hi Steph,

Only just discovered this site, so I'm a bit late to the game.

I'd love to hear what your main take-aways were from this meeting!

Best,

Louise

thanks for this very interesting post! i was also trying to develop an inexpensive camera trap but with not good results. i think the use of a PIR sensor can give more battery life than motion detection via software. Can you give more details about the components you used please?

Thanks
Paolo