MINORA Project Home Page



Swiss priority program OPTIQUE II, funded by the ETH Council




Modern everyday life is characterized by an ever increasing interaction between man and machines leading to a growing number of potentially harmful situations for human beings. This holds true not only for factory environments but for virtually every facet of daily life: When using public or private transportation (trains, buses, trams, cars, etc.) or when entering through automatic sliding or revolving doors, the health or even the life of users might be endangered through accidents, malfunctions or human oversight. Most of these devices and many more, e.g., in the surveillance and security area, would profit from the availability of reliable presence detection systems which would be able to construct and evaluate a coarse three-dimensional representation of the world. For these purposes distance and lateral resolutions of about 10 cm would be sufficient for distances between a couple of meters to some 100 meters. These applications are already addressed by a wide variety of products on the market, based on ultrasound, radar, light barriers, coded light triangulation, stereo vision etc. Although each of these approaches has a certain success in a niche application, none comes really close to the ideal miniaturized range camera, which is reliable, inexpensive, robust, and safe to use.

The purpose of the new project is to develop two types of universal range cameras based on optical ranging techniques (time-of-flight or AM laser radar), with which a large part of today's safety, surveillance and automotive applications can be solved. The first type is a long-distance (up to 500 m) narrow field-of-view range camera, capable also of tracking moving targets Main applications are foreseen in the automotive area ("smart cruise control") and in the transportation safety sector ("train/tram radar"). The second type is a mid-range (up to 5m), wide field-of-view side-looking range sensor. While the first type of range sensor is typically a small box in a traditional camera configuration, the second will be rather used in thin profiles for safety applications. The 3D range image produced by this side-looking distance sensor is rather elongated, possibly consisting only of a strip or distance profile. The first type of range camera, however, is expected to produce range images with a typical resolution of about 64x16 pixels. These images are difficult to interpret due to several problems normally not encountered in intensity images The range images are spatially undersampled, they include incomplete data (target reflectivity either too low or so high that the pixels are saturated), the data can be ambiguous (multiple optical reflections), and the scenes and objects to be interpreted are constantly moving (analysis of range image sequences required). For this reason, preprocessing and interpretation of the temporal sequences of range images need novel algorithms which must be of low enough complexity to lend themselves well for the implementation in custom ASICs.

Objectives of the project: