The
light curtain constraint graph encodes the physical constraints of the light curtain device and can be used to compute feasible light curtain profiles. We extend the constraint graph to incorporate acceleration constraints of the device. We propose a mechanism to
sample light curtains from the constraint graph by defining a transition probability distribution over neighboring nodes. Then, a random light curtain is simply a random walk from the leftmost to the rightmost camera ray under these transition probabilities. Importantly, we perform
dynamic programming on the constraint graph to provide
analytic estimates of the probability of a random curtain detecting any user-specified object!
For each node in the graph, we compute the detection probability of a random light curtain starting from that node and ending on the rightmost ray. The solutions to these sub-problems satisfy a recursive relationship shown by the dynamic programming equation above. This recursive expression can be used to efficiently and analytically compute the detection probability in a single backward pass through the graph.