Generally, in ray tracing, light paths are traced backward from the image plane into the scene; the digram suggests modeling the optics explicitly to capture certain effects (in this case, depth-of-field).  Note the diagram uses the term "film plane" instead of image plane or screen plane, as the diagram models a camera when film use was common, before high quality CCDs were available.

In this case, interaction with the lens implies the focal point: all paths originating from the same point on the image plane will converge at the focal point (assuming the paths first intersect the lens - note the diagram does not show an aperture, which would block paths that don't) - but paths are not coincident from the lens and diverge beyond the focal point - so they intersect the scene at different locations nearer and further than the focal point. The variation in scene intersection locations, given the same origin on the image plane (and therefore contributing to the same pixel in the computed image), is what gives the depth of field effect - i.e. something looking like circular blur outside of the plane of focus. Other effects are possible here, such as anti-aliasing by varying the origin of rays across a pixel's area in the image plane.