In　this　paper,　we　offer　a　machine　learning　classifier　model,　later　considered　as　MLCM,　for　classifying　objects　such　as　road　signs　and　vehicles.　Showing　the　influence　of　vocabulary　size　on　accuracy　of　SVM　using　SURF.　Based　on　SURF　method　used　bag-of-words　model　as　feature　extractor.　Due　to　its　simplifying　representation,　it　accelerates　the　first　stage　of　our　MLCM.　We　tested　and　analyzed　accuracy　of　Support　Vector　Machines,　including　Linear,　Quadratic　and　Medium　Gaussian　SVM　as　flowed　step　model　and　automatically　use　best　result　for　further　estimation.　Furthermore,　we　provide　a　brief　introduction　of　applied　methods　and　experimental　results　analysis.　MLCM　introduces　combination　of　SURF　method　and　several　SVMs　as　well　as　optimized　SVM.　This　technique　shows　good　performance　with　minimum　failures.　Thereafter,　it　will　be　implemented　for　real-time　video　sequences.　The　achieved　goal　can　be　implemented　in　the　use　of　self-driving　of　industrial　machines　with　a　safe　speed.　©　2020　ACM.