A Modified Binary Descriptor for Object Detection

A Modified Binary Descriptor for Object Detection

Ritu Rani (HMR Institute of Technology and Management, India), Ravinder Kumar (Skill Faculty of Engineering, Shri Vishwakarma Skill University, India) and Amit Prakash Singh (Guru Gobind Singh Indraprastha University, Delhi, India)
Copyright: © 2021 |Pages: 17
DOI: 10.4018/JITR.2021010102
OnDemand PDF Download:
Available
$37.50
No Current Special Offers
TOTAL SAVINGS: $37.50

Abstract

The reliability of computer vision applications highly depends on the extraction of compact, fast, and accurate and robust feature description. This paper presents a better and modified binary descriptor based on ORB (oriented and rotated brief) with the SVM-RBF-RFE (support vector machine-radial basis function-recursive feature elimination) to achieve a better extraction and representation of local binary descriptors. This work presents the extensive comparison of the proposed modified descriptor with the state-of-the-art binary descriptors on various datasets. The results show that the proposed descriptor is highly distinctive and efficient as compared to the other state-of-the-art binary descriptors. The experiments were performed on the four benchmark datasets PASCAL, CALTECH, COIL, and OXFORD to demonstrate the robustness and effectiveness of the proposed descriptor. The robustness and effectiveness of the proposed descriptor is tested under the various transformations like scaling, rotation, noise, intensity variation.
Article Preview
Top

Introduction

Object detection is the process of an extraction and identification of an object, and its location in a given image/scene. These objects could be some bird, animal, bicycle, mug, chair, building etc in an image or video. Object recognition involves identification of objects in image and assigning of that object to the particular class labels through machine learning. Since, humans can easily distinguish the object classes like bird, animal, chair etc based on their characteristics; similarly the same property of recognising the object classes by machines can prove to be very useful in various domains. Thus object detection is the subset of object recognition. Object detection can be understood in two ways; soft detection, which only tells the presence of an object and hard detection which not only detects the presence but also the location of the object in the image.

Object detection is performed by searching almost each part of an image to localize parts, whose photometric or geometric properties match those of the target object in the training database. An object template is scanned across an image at various locations, scales and rotations .The similarity between the template and the image should be high to perform detection. Object detection has various applications like face detection, pedestrian detection, image retrieval, security, surveillance etc. Variety of models can be used for object detection; in this paper feature-based object detection has been done. Feature based object detection involves the extraction of the best features for matching so as to perform accurate detections. Thus, extraction of local features plays an important role in almost all the computer vision applications (Shapiro, 2001; Klette, 2014; Golnabi & Asadpour, 2007; Morris, 2003). Design of such robust and efficient features are under intensive investigation by the research community. These features are highly important for the various applications in biometrics (Kumar, Chandra & Hanmandlu, 2016, 2014a, 2014b) military applications, automated industry applications, object detection, object recognition etc. The design of feature descriptors needs to be invariant under various transformations.

Descriptors are also classified as the gradient and binary descriptors. The recent development focuses on design of binary features descriptors, due to their simplicity, less complexity and ease of implementation (Muja & Lowe, 2012). The detailed objectives of the proposed work are as under:

  • To design an invariant binary descriptor for object detection under various circumstances like scaling, rotation, intensity variation, noise etc. The empirical evaluation of the modified binary descriptor on various standard benchmark datasets (CALTECH, COIL, PASCAL, OXFORD) in order to show its robustness, efficiency and matching accuracy.

  • To evaluate the performance of the proposed binary descriptor against various other binary descriptors using the recall and precision performance metrics so as to highlight its accuracy and suitability.

  • To introduce the readers to the proposed ORB based binary descriptor with the SVM-RBF-RFE algorithm (for feature selection) and show the suitability of the descriptor in terms of compact size, low memory requirement, fast execution and higher matching accuracy.

It is evident from the results that the proposed binary descriptor outperforms the other binary descriptors of its class in terms of robustness, efficiency and matching accuracy. Its performance is better as compared to the ORB descriptor under intensity variation, rotation, and scaling and noise variation.

The paper is organised as follows: Next Section presents the review of the literature in this domain. The section after that presents the proposed work in the form of the flow chart and explains all the modules in detail along with their mathematical formulation. The next section presents the experimental setup and discussion of the results. The next Section demonstrates the performance of the proposed descriptors and finally the paper is concluded in the last section of the paper.

Top

Literature Review

Owing to the advancement in the domain of computer vision applications, the design of reliable binary descriptor faces various challenges like compactness and invariance. Lot of work has already been published in this domain to address the issues in design of the binary descriptor.

Complete Article List

Search this Journal:
Reset
Volume 15: 6 Issues (2022): 1 Released, 5 Forthcoming
Volume 14: 4 Issues (2021)
Volume 13: 4 Issues (2020)
Volume 12: 4 Issues (2019)
Volume 11: 4 Issues (2018)
Volume 10: 4 Issues (2017)
Volume 9: 4 Issues (2016)
Volume 8: 4 Issues (2015)
Volume 7: 4 Issues (2014)
Volume 6: 4 Issues (2013)
Volume 5: 4 Issues (2012)
Volume 4: 4 Issues (2011)
Volume 3: 4 Issues (2010)
Volume 2: 4 Issues (2009)
Volume 1: 4 Issues (2008)
View Complete Journal Contents Listing