3D object recognition is one of the most challenging tasks facing artificial systems. Thus, the ability to detect and localize the regions of interest is necessary to provide an enhanced searching and visualisation beyond a simple high-level categorisation. Recently, many approaches based on 3D objects learning have been proposed, they rely on learning objects characteristics from fully labelled 3D objects. However, such data training steps are difficult to be acquired at scale. In this paper we explore machine learning techniques to recognize objects, based on local parts, from a data base. The idea behind our approach is to compute and label the quantized local descriptor around 3D interest points using both intuitive geometric and spatial properties and an effective supervised classifier, named respectively Tri Spin Image and boosted-SVM classifier. First, we extract the salient points using Harris3D then the significant and robust feature vectors representing the keypoints are computed then quantized using bag of features. Second, we use these vectors to train a boosted-SVM classifier. The performance of the proposed method is evaluated and proves encouraging results.
CITATION STYLE
Abouqora, Y., Herouane, O., Moumoun, L., & Gadi, T. (2020). 3D Objects Learning and Recognition Using Boosted-SVM Algorithm. In Learning and Analytics in Intelligent Systems (Vol. 7, pp. 426–435). Springer Nature. https://doi.org/10.1007/978-3-030-36778-7_47
Mendeley helps you to discover research relevant for your work.