Density-based Clustering for 3D Stacked Pipe Object Recognition using Directly-given Point Cloud Data on Convolutional Neural Network

Pratama, Alfan Rizaldy; Dewantara, Bima Sena Bayu; Sari, Dewi Mutiara; Pramadihanto, Dadet

doi:10.24003/emitter.v10i1.704

Cited by 6 publications

(3 citation statements)

References 20 publications

Supporting

Mentioning

Contrasting

Unclassified

Order By: Relevance

“…[15] menyajikan pengenalan wajah baru dengan menggunakan titik sudut arah 3D (3D DCP). Masih pada dekade yang sama setelah arsitektur PointNet [16] dari Convolutional Neural Network (CNN) dikenalkan semakin banyak peneliti yang mencoba menerapkan arsitektur ini untuk berbagai pengolahan data 3D, seperti pengenalan objek 3D [17] dan pengenalan wajah 3D [5], [18]. PointNet muncul sebagai pendekatan yang lebih fleksibel dan kuat dalam pengenalan wajah 3D.…”

Section: A Pendahuluanunclassified

Pengenalan Wajah 3D dengan menggunakan PointNet

Arif Hidayah,

Dewantara,

Pramadihanto

2023

ijcs

View full text Add to dashboard Cite

Pengenalan wajah tiga dimensi (3D) telah menjadi topik penelitian yang menarik karena mampu mengatasi keterbatasan pengenalan wajah dua dimensi (2D) dalam menghadapi perubahan pose, pencahayaan, dan pemalsuan. Penelitian ini mengusulkan sebuah pipeline pengenalan wajah 3D yang invarian terhadap perubahan cahaya, dengan menggunakan teknik segmentasi euclidean clustering dan Convolutional Neural Network (CNN) PointNet. Data wajah diambil menggunakan kamera Time-of-Flight yang menghasilkan titik awan (point cloud). Proses segmentasi euclidean clustering berhasil memisahkan area wajah dengan akurat, membantu dalam pengenalan wajah 3D. Melalui pelatihan dengan 217 dataset dan 2048 titik per wajah, sistem mencapai akurasi pelatihan sebesar 99% dan akurasi validasi sebesar 84,4%, dengan loss pelatihan sebesar 1% dan loss validasi sebesar 15,6%. Evaluasi pada tiap kelas menunjukkan rata-rata akurasi 0.9887471867966992, presisi 0.8255813953488372, recall 0.8255813953488372, dan F1-score 0.8255813953488372. Hasil menunjukkan bahwa pipeline pengenalan wajah 3D ini memiliki potensi besar dalam aplikasi keamanan, pengawasan, dan pengenalan objek di lingkungan yang kompleks. Three-dimensional (3D) face recognition has emerged as an intriguing research topic, addressing the limitations of two-dimensional (2D) face recognition in handling pose variations, lighting changes, and spoofing. This study proposes an illumination-invariant pipeline for 3D face recognition, utilizing the euclidean clustering segmentation technique and Convolutional Neural Network (CNN) PointNet. Facial data is captured using a Time-of-Flight camera, generating point clouds. The euclidean clustering segmentation effectively isolates facial regions, aiding in 3D face recognition. After training with 217 datasets and 2048 points per face, the system achieved 99% training accuracy and 84.4% validation accuracy, with 1% training loss and 15.6% validation loss. Class-wise evaluation yielded an average accuracy of 0.9887471867966992, precision of 0.8255813953488372, recall of 0.8255813953488372, and F1-score of 0.8255813953488372. The results highlight the significant potential of this 3D face recognition pipeline in security, surveillance, and object recognition in complex environments.

show abstract

Section: A Pendahuluanunclassified

Pengenalan Wajah 3D dengan menggunakan PointNet

Arif Hidayah,

Dewantara,

Pramadihanto

2023

ijcs

View full text Add to dashboard Cite

show abstract

“…In recent years, the research of computer vision and autonomous driving has received more and more attention. Three-dimensional rigid PCR is a basic and critical problem in the field of 3D vision, such as 3D modeling [1,2], localization for robot navigation [3][4][5], object recognition [6][7][8], and surface alignment [9,10]. The purpose of PCR is to seek the best transformation parameters that precisely aligns a pair of point in different coordinate systems.…”

Section: Introductionmentioning

confidence: 99%

Establishment and Extension of a Fast Descriptor for Point Cloud Registration

et al. 2022

View full text Add to dashboard Cite

Point cloud registration (PCR) is a vital problem in remote sensing and computer vision, which has various important applications, such as 3D reconstruction, object recognition, and simultaneous localization and mapping (SLAM). Although scholars have investigated a variety of methods for PCR, the applications have been limited by low accuracy, high memory footprint, and slow speed, especially for dealing with a large number of point cloud data. To solve these problems, a novel local descriptor is proposed for efficient PCR. We formed a comprehensive description of local geometries with their statistical properties on a normal angle, dot product of query point normal and vector from the point to its neighborhood point, the distance between the query point and its neighborhood point, and curvature variation. Sub-features in descriptors were low-dimensional and computationally efficient. Moreover, we applied the optimized sample consensus (OSAC) algorithm to iteratively estimate the optimum transformation from point correspondences. OSAC is robust and practical for matching highly self-similar features. Experiments and comparisons with the commonly used descriptor were conducted on several synthetic datasets and our real scanned bridge data. The result of the simulation experiments showed that the rotation angle error was below 0.025° and the translation error was below 0.0035 m. The real dataset was terrestrial laser scanning (TLS) data of Sujiaba Bridge in Chongqing, China. The results showed the proposed descriptor successfully registered the practical TLS data with the smallest errors. The experiments demonstrate that the proposed method is fast with high alignment accuracy and achieves a better performance than previous commonly used methods.

show abstract

“…RoISC (Robotics and Intelligent System Center) technology has been penetrated into a wider area in various research fields, ranging from humanoid robots, to computer vision [1]. In computer vision research fields, various studies such as robust human body orientation estimation [2], densitybased clustering for stacked pipe object [3], and so on have been carried out recently. Not content with that, RoISC is currently preparing to delve deeper into computer vision research and plans to achieve novelty by combining stereo vision features with omnidirectional cameras features.…”

Section: Introductionmentioning

confidence: 99%

Omnidirectional Stereo Vision Study from Vertical and Horizontal Stereo Configuration

Pratama

Dewantara

Pramadihanto³

2022

EMITTER Int'l J. of Engin. Technol.

View full text Add to dashboard Cite

In stereo vision, an omnidirectional camera has high distortion compared to a standard camera, so the camera calibration is very decisive in its stereo matching. In this study, we will perform stereo matching for an omnidirectional camera with vertical and horizontal configuration so that the result of the image's depth has a 360-degree field of view by transforming the image using a calibration-based method. The result is that by using a vertical camera configuration, the image can be stereo matched directly, but by configuring a horizontal image, it is necessary to carry out a different stereo-matching process in each direction. Stereo matching with the semi-global matching method has better image results than block matching with more image objects detectable by the semi-global block matching method with a maximum disparity value of 32 pixels and with a window size of 21 pixels.

show abstract

Density-based Clustering for 3D Stacked Pipe Object Recognition using Directly-given Point Cloud Data on Convolutional Neural Network

Cited by 6 publications

References 20 publications

Pengenalan Wajah 3D dengan menggunakan PointNet

Pengenalan Wajah 3D dengan menggunakan PointNet

Establishment and Extension of a Fast Descriptor for Point Cloud Registration

Omnidirectional Stereo Vision Study from Vertical and Horizontal Stereo Configuration

Contact Info

Product

Resources

About