Two cameras, one geometric trick
Stereo vision mimics how human eyes judge distance: use two cameras spaced a fixed distance apart, the baseline, and compare the two images they capture at the same moment. Because the cameras are in slightly different positions, the same physical point in the world lands at a slightly different pixel position in each image. That pixel shift is called disparity, and it is the key measurement stereo vision is built around.
The relationship between disparity and distance is intuitive once you think through the geometry. Hold a finger close to your face and look at it with one eye closed, then the other, it appears to jump a large amount between the two views. Now hold your finger at arm's length and repeat, the jump is much smaller. Objects far away, like a distant mountain, barely shift at all. Nearby objects produce large disparity, distant objects produce small disparity, and that inverse relationship is what lets a stereo system convert a measured pixel shift into an actual depth value.