Evolutionary Object Detection by Means of Naieve Bayes Models Estimation

摘要

This paper describes an object detection approach based on the use of Evolutionary Algorithms based on Probability Models (EAPM). First a parametric object detection schema is defined, and formulated as an optimization problem. The new problem is faced using a new EAPM based on Naive Bayes Models estimation is used to find good features. The result is an evolutionary visual feature selector that is embedded into the Adaboost algorithm in order to build a robust detector. The final system is tested over different object detection problems obtaining very promising results.