Regularization techniques and parameter estimation for object detection on hyperspectral data

The main challenge for the retrieval of information using hyperspectral sensors is that due to the high dimensionality provided by them there is not comparably enough training samples data to produce well-estimated parameters to solve our supervised detection problem. This lack of enough training samples information for an estimation yields to an ill-posed and ill-condition problem. As a consequence, this leads to an increment in false alarms and an increase in the probability of missing throughout the classification process.; An approach based on a regularization technique applied to the parameters obtained from the data collected from the hyperspectral sensor is used to simultaneously minimize the probabilities of error and missing. This procedure is implemented using algorithms that apply regularization techniques by biasing the Maximum Likelihood estimate of the covariance matrix, which enable the reduction of the probability of error and the decrease of the probability of missing.; This work presents a statistical approach in which the optimum regularization parameter is automatically selected minimizing the probability of error and simultaneously decreasing the probability of missing.