Hybrid shape descriptor and meta similarity generation for non-rigid and partial 3D model retrieval

Abstract

Non-rigid and partial 3D model retrieval are two significant and challenging research directions in the field of 3D model retrieval. Little work has been done in proposing a hybrid shape descriptor that works for both retrieval scenarios, let alone the integration of the component features of the hybrid shape descriptor in an automatic way. In this paper, we propose a hybrid shape descriptor that integrates both geodesic distance-based global features and curvature-based local features. We also develop an automatic algorithm to generate meta similarity resulting from different component features of the hybrid shape descriptor based on Particle Swarm Optimization. Experimental results demonstrate the effectiveness and advantages of our framework, as well as the significant improvements in retrieval performances. The framework is general and can be applied to similar approaches that integrate more features for the development of a single algorithm for both non-rigid and partial 3D model retrieval

A comparison of 3D shape retrieval methods based on a large-scale benchmark supporting multimodal queries

“…Due to the non rigid properties of the models, it is more challenging to perform the retrieval. For a review of non rigid 3D retrieval techniques based on geodesic distance and spectrum analysis approaches, as well as different canonical form transforms for non rigid models based on multidimensional scaling, please refer to [12]. Another recent survey of non rigid shape retrieval is presented in [60], where a performance comparison of several descriptors derived from spectral geometry is given.…”

Classified 3d Model Retrieval Based on Cascaded Fusion of Local Descriptors

“…More specifically, following state of the art methods achieve high accuracy results evaluated against some of the most challenging datasets like SHREC'13 [1], '14 [2] and '15 [3]. The methods proposed include the hydric shape descriptors [4] and meta similarity generation for non-rigid 3D model retrieval [6] that integrates geodesic distance-oriented global features and curvature-oriented local descriptors and in a more general approach can be applied to similar methods that integrates more than one features in order to develop a specific algorithm for both partial and rigid 3D models. Another efficient descriptor methods are the Histograms of Area Projection Transform, the Radical symmetry Detection and the Shape Characterization with Multiscale Area projection transform [5] and theRBiHDM [2]method where it computes the project matrix k called reduced bi-harmonic distance matrix and the resulted equation is equal to:…”

Exploration in improving retrieval quality and robustness for deformable non-rigid 3D shapes

“…In the mean time, Li B. et al [17] created a hybrid ZFDR descriptor by summing directly over the distances of four features (Zernike moments, Fourier descriptor, Depth information and Ray-based features), which performs better than most view-based methods. Then, their another hybrid work [18] produced a further improvement by uniting curvature-based local feature, geodesic distance based global feature and MDS based ZFDR feature. For weight assignment, a Particle Swarm Optimization (PSO) algorithm was used.…”

“…Based on the investigation of previous works [7,12,18,20,21], feature extraction and fusion are two basic discussions for shape description. Usually, for single feature, it has limited ability in capturing shape content comprehensively, and, in most cases, it highlights the performance in some perspectives under some specific constrains, which has resulted in an urgent need for feature fusion to make full use of separate features for complement.…”

“…As stated in [20], it is easy to find that all of the compared methods could not perform equally well for different shape classes. (3) Feature fusion is usually expressed as a weighted sum of their individual dissimilarities, and, in most of existing methods [12,18,21], these weights are determined heuristically. However, this kind of weighted sum approach usually finds a global optimal value while ignores the local conficts, which still performs not such sufficient for feature combination.…”