This paper aims to reduce and eliminate the abnormal peaks which, because of the reflection in the process of laser detection, make it easier to proceed with further analysis.
To solve the above problem, an abnormal data correction algorithm based on histogram, K-Means clustering and improved robust locally weighted scatter plot smoothing (LOWESS) is put forward. The proposed algorithm does section leveling for shear plant first and then applies histogram to define the abnormal fluctuation data between the neighboring points and utilizes a K-Means clustering to eliminate the abnormal data. After that, the improved robust LOWESS method, which is based on Euclidean distance, is used to remove the noise interference and finally obtain the waveform characteristics for next data processing.
The experiment result of liner tool mark laser test data correction demonstrates the accuracy and reliability of the proposed algorithm.
The study enables the following points: the detection signal automatic leveling; abnormal data identification and demarcation using K-Means clustering and histogram; and data smoothing using LOWESS.
