The perspective transform performed by a pinhole camera can be seen as a projectivity from the projective space P3 (the "world") to the projective plane P2 (the "picture"). A classic problem in Computer Vision is to estimate the position of the photo camera given a picture, so that one can infer 3D information from the image. This is basically getting the coefs of the matrix which represent the above projectivity. The literature offers tons af algorithm to do that, but they all assume the presence of non-natural artefacts in the viewed panorama: straight lines, known angles, orthogonalities, reference points. We claim that, in some particular cases, we can recover the position of the camera from the presence of random points in the scene, given that 1. these points all lay on the same plane 2. the probability distribution of these points as seen "before" the prespective transform is known. These conditions holds, for instance, if we are looking at a field of flowers.