The Classification of Skateboarding Trick Manoeuvres Through the Integration of Image Processing Techniques and Machine Learning

More often than not, the evaluation of skateboarding tricks executions is assessed intuitively according to the judges’ observation and hence are susceptible to biasness if not inaccurate judgement. Hence, it is crucial to underline the benchmark for analyzing the rate of successful execution of skateboarding trick for high level tournaments. The common tricks in skateboarding such as Kickflip, Ollie, Nollie, Pop Shove-it and Frontside 180 are investigated in this study via the synthetization of image processing and machine learning classifiers. The subject used for accomplishing the tricks is a male amateur skateboarder at the age of 23 years old with ±5.0 years’ experience using ORY skateboard. Each trick is collected upon five successful landings and the camera is placed 1.26 m from the subject on a flat cemented ground. The features extracted from each trick were engineered using Inception-V3 image embedder. Several classification models were evaluated, namely, Support Vector Machine (SVM), k-Nearest Neighbour (kNN), Logistic Regression (LR), Random Forest (RF) and Naïve Bayes (NB) on their ability in classifying the tricks based on the engineered features. It was observed from the preliminary investigation that the SVM model attained the highest classification accuracy with a value of 99.5% followed by LR, k-NN, RF, and NB with 98.6%, 95.8%, 82.4% and 78.7%, respectively. It could be inferred that the method proposed decisively provide the classification of skateboarding tricks efficiently and would certainly provide a more objective based judgment in awarding the score of the tricks.