Cellular neural networks for motion estimation and obstacle detection

[摘要] Obstacle detection is an important part of VideoProcessing because it is indispensable for a collision preventionof autonomously navigating moving objects. For example,vehicles driving without human guidance need a robustprediction of potential obstacles, like other vehicles or pedestrians.Most of the common approaches of obstacle detectionso far use analytical and statistical methods like motion estimationor generation of maps.

In the first part of this contribution a statistical algorithmfor obstacle detection in monocular video sequences is presented.The proposed procedure is based on a motion estimationand a planar world model which is appropriate to trafficscenes. The different processing steps of the statistical procedureare a feature extraction, a subsequent displacementvector estimation and a robust estimation of the motion parameters.Since the proposed procedure is composed of severalprocessing steps, the error propagation of the successivesteps often leads to inaccurate results.

In the second part of this contribution it is demonstrated,that the above mentioned problems can be efficiently overcomeby using Cellular Neural Networks (CNN). It will beshown, that a direct obstacle detection algorithm can be easilyperformed, based only on CNN processing of the inputimages. Beside the enormous computing power of programmableCNN based devices, the proposed method is alsovery robust in comparison to the statistical method, becauseis shows much less sensibility to noisy inputs. Using the proposedapproach of obstacle detection in planar worlds, a realtime processing of large input images has been made possible.

[学科分类] 电子、光学、磁材料