KCF-Hash-Match Target Tracking Algorithm for Identifying Wheel Lateral Sway of High-speed Train

A good wheel-rail contact state is an important guarantee for the running safety of high-speed trains. The relative lateral movement of wheel and rail is an important indicator for judging the wheel-rail contact state, which directly determines the danger of wheel derailment. The wheel-rail online tracking monitoring technology based on machine vision can provide a new idea for the state identification of the wheel-rail contact point of high-speed trains. The test adopts the method of installing a camera at the bottom of the bogie to obtain images of wheel rail contact during high-speed train operation. Considering the advantages of fast tracking speed of the KCF algorithm and the high tracking accuracy of the template matching algorithm, and incorporating the principle of hashing algorithm, a Kernel Correlation Filtering-Hash-Match algorithm（KCF-Hash-Match）that can automatically identify wheel-rail contact geometry is proposed, which is applied to the monitoring of abnormal wheel lateral sway of high-speed trains under complex condition. The results show that the KCF-Hash-Match algorithm can solve the error accumulation problem existing in the traditional KCF algorithm when deal with tracking fast moving targets. When the allowable error threshold is 8 pixels, the accuracy rate can attain 99.8%. Compared with the field measured data, the KCF-Hash-Match algorithm has high tracking accuracy and can realize the safety monitoring of high-speed train wheel lateral sway behavior.