An outlier correction procedure and its application to areal surface data measured by optical instruments

Abstract

An outlier correction procedure for areal surface topography data based on surrounding data for outlier detection and correction is reviewed. The outlier detection is based on the median relative height of the surrounding data within a defined detection window. The threshold value is calculated from a cumulative probability curve of the medians of all data points. The detection window size is selected based on the size of the largest outlier cluster observed on the topography data. The application of the procedure to dental implant surface topography data measured by a confocal laser scanning microscope shows excellent outlier removal compared to typical filtering methods. The procedure only affects outliers, and the information on normal data points remains unchanged. The application to topography data from 21 repeated measurements of a deterministic measuring standard shows great improvement in accuracy and repeatability of surface roughness parameters. The flexibility of the outlier correction procedure is observed through its application to several types of topography data from various types of optical surface measurement systems.

Application of Clustering Filter for Noise and Outlier Suppression in Optical Measurement of Structured Surfaces

“…However, it will fail when applied to data with clustered outliers since these outliers can result in biased medians. Ismail et al [8] developed an optimized median filter. The outlier detection is based on the median relative height of the surrounding data within a defined detection window.…”

“…In addition to measurement noise, optical measurement techniques can often produce unmeasured or faulty points and outliers. Outliers are presented as sudden sharp jumps compared to the surrounding data and occur at certain localities on the surface [8]. In ISO 16610-1 [9], an outlier is defined as a local portion in a data set that is not representative, or not typical, for the partitioned integral feature.…”

Rolling deformation of truck tires: Measurement and analysis using a tire sensing approach

“…Due to the high sampling frequency and limited reflectivity of the rubber surface, outliers, which have unreasonably high and low values compared to the surrounding data, were observed in the raw data. To eliminate those outliers, a filter based on the cumulative probability neighbor median (Ismail et al, 2010) was used to process data. The filtered signal shows clear peaks when the tire was being deformed by the road (Fig.…”

Geometrical transcription of diamond electroplated tool in ultrasonic vibration assisted grinding of steel

“…Outliers are errors which must be treated selectively. In the current study, an outlier correction procedure reported in [25] which the outliers detection is based on median of neighbouring relative height, is employed with the largest outlier cluster is defined as 5 Â 5 pixels and detection level L is 1.1. The height information of identified outliers is deleted and replaced with an artificial value based on the weighted average of available normal data points on 8 equiangular directions.…”