Target Detection through Visual Recognition

A simple method for calculating the probability that an observer will detect a fixed object against background clutter in a limited time. Tailored to airborne search for ground targets, the method applies also to design of information display equipment and to the general limitations on system performance set by a human observer. Inputs to be supplied by the user of the model are estimates of target size and contrast, required search rate in area per second, and scene congestion (false target density). If optical aids or sensor displays are used, parameters also include image scale and contracts, system resolution, and signal-to-noise ratio. Input accuracy is expected to be no better than 20 to 30 percent. Six algebraic equations are used to estimate recognition probability as a function of the input parameters, for an average observer. The model shows that broad area search from high-speed aircraft is probably futile, while road recce up to a few hundred knots may be quite feasible.