Outlier mining of a vision sensing databasefor SVM regression improvement

A 3D spatial measurement system has been enhanced by computational intelligence. The measurement system is based in opto-electronic scanning instrumentation for industrial task, robot navigation, medical scanning, and structural health monitoring applications. This paper presents new research performed in the data processing of a vision sensing database. Multivariate outlier analysis has been implemented by Mahalanobis distance in order to improve Support Vector Machine (SVM) regression algorithm results. Measurement error regression data has been used for spatial 3D measurements error correction. Demonstrating that optical systems for measurements that present non-linear behavior could be positive impacted by computational intelligence.