DRO: "DUDA Road Operator"

DRO: "DUDA Road Operator"

DUDA's detector is another method of classification, which uses four masks to convolve images, and to assign a score to each pixel.

A score function calculates

a value G(u) which is function of the uniformity along a curve, and
a value F(u) depending on the contrast with the neighbourhood.

The score functions are calculated for the four masks, the central pixel a₂ takes the maximum value.

**Figure 9:** The 4 masks of DUDA's roads detector

Horizontal Right Diagonal Vertical Left Diagonal

The score function is given by the following expression [M. A. Fischler, 1981]:

$\begin{displaymath}Score= \frac{G(\vert a_1- a_2\vert)\times G(\vert a_2- a_3\vert)}{\sum^3_{i= 1}F(a_i- b_i)+ F(a_i- c_i)}\end{displaymath}$

This detector is sensitive:

to road orientation,
to the effects of discretisation,
to rough changes of orientation and
to weak contrast with the environment.

It detects also the profiles of ``step'' type and causes a significant number of false detections. The result is however much less disturbed than in the case of a ``Top hat'' detector.

IRIT-UPS