Deep Learning Methods for Calibrated Photometric Stereo and Beyond.

Photometric stereo recovers the surface normals of an object from multiple images with varying shading cues, i.e., modeling the relationship between surface orientation and intensity at each pixel. Photometric stereo prevails in superior per-pixel resolution and fine reconstruction details. However, it is a complicated problem because of the non-linear relationship caused by non-Lambertian surface reflectance. Recently, various deep learning methods have shown a powerful ability in the context of photometric stereo against non-Lambertian surfaces. This paper provides a comprehensive review of existing deep learning-based calibrated photometric stereo methods utilizing orthographic cameras and directional light sources. We first analyze these methods from different perspectives, including input processing, supervision, and network architecture. We summarize the performance of deep learning photometric stereo models on the most widely-used benchmark data set. This demonstrates the advanced performance of deep learning-based photometric stereo methods. Finally, we give suggestions and propose future research trends based on the limitations of existing models.

Surface Geometry Processing: An Efficient Normal-Based Detail Representation.

With the rapid development of high-resolution 3D vision applications, the traditional way of manipulating surface detail requires considerable memory and computing time. To address these problems, we introduce an efficient surface detail processing framework in 2D normal domain, which extracts new normal feature representations as the carrier of micro geometry structures that are illustrated both theoretically and empirically in this article. Compared with the existing state of the arts, we verify and demonstrate that the proposed normal-based representation has three important properties, including detail separability, detail transferability and detail idempotence. Finally, three new schemes are further designed for geometric surface detail processing applications, including geometric texture synthesis, geometry detail transfer, and 3D surface super-resolution. Theoretical analysis and experimental results on the latest benchmark dataset verify the effectiveness and versatility of our normal-based representation, which accepts 30 times of the input surface vertices but at the same time only takes 6.5% memory cost and 14.0% running time in comparison with existing competing algorithms.

Efficient Computer-Aided Design of Dental Inlay Restoration: A Deep Adversarial Framework.

Restoring the normal masticatory function of broken teeth is a challenging task primarily due to the defect location and size of a patient's teeth. In recent years, although some representative image-to-image transformation methods (e.g. Pix2Pix) can be potentially applicable to restore the missing crown surface, most of them fail to generate dental inlay surface with realistic crown details (e.g. occlusal groove) that are critical to the restoration of defective teeth with varying shapes. In this article, we design a computer-aided Deep Adversarial-driven dental Inlay reStoration (DAIS) framework to automatically reconstruct a realistic surface for a defective tooth. Specifically, DAIS consists of a Wasserstein generative adversarial network (WGAN) with a specially designed loss measurement, and a new local-global discriminator mechanism. The local discriminator focuses on missing regions to ensure the local consistency of a generated occlusal surface, while the global discriminator aims at defective teeth and adjacent teeth to assess if it is coherent as a whole. Experimental results demonstrate that DAIS is highly efficient to deal with a large area of missing teeth in arbitrary shapes and generate realistic occlusal surface completion. Moreover, the designed watertight inlay prostheses have enough anatomical morphology, thus providing higher clinical applicability compared with more state-of-the-art methods.

Mesh-Based Computation for Solving Photometric Stereo With Near Point Lighting.

We tackle the problem of dense reconstruction with a practical system, in which near point lighting (NPL) is employed. Different from the conventional formulation of photometric stereo (PS) that assumes parallel lighting, PS under the NPL condition is a nonlinear problem as the local surface normals are coupled with its distance to the camera as well as the light sources. After obtaining the locations of point lights by a calibration process, we develop a new framework to solve this nonlinear reconstruction problem via mesh deformation, in which each facet is corresponding to a pixel in the image captured by the camera. In our framework, mesh deformation is decoupled into an iteration of interlaced steps of local projection and global blending. Experimental results verify that our method can generate accurate estimation of surface shape under NPL in a few iterations. Besides, this approach is robust to errors on the positions of light sources and is easy to be implemented.

Highly chemoresistive humidity sensing using poly(ionic liquid)s.

A novel resistance type humidity sensor was fabricated using poly(ionic liquid)s PEVIm-Br, which not only exhibited high sensitivity at a relative humidity (RH) in the range of 11-98%, but also featured short response and recovery time, as well as small hysteresis and good repeatability, proving poly(ionic liquid)s to be promising sensing materials for high-performance humidity sensors.

Low-Temperature H2S Detection with Hierarchical Cr-Doped WO3 Microspheres.

Hierarchical Cr-doped WO3 microspheres have been successfully synthesized for efficient sensing of H2S gas at low temperatures. The hierarchical structures provide an effective gas diffusion path via well-aligned micro-, meso-, and macroporous architectures, resulting in significant enhancement in sensing response to H2S. The temperature and gas concentration dependence on the sensing properties elucidate that Cr dopants remarkably improve the response and lower the sensor' operating temperature down to 80 °C. Under 0.1 vol % H2S, the response of Cr-doped WO3 sensor is 6 times larger than pristine WO3 sensor at 80 °C. We suggest the increasing number of oxygen vacancies created by Cr dopants to be the underlying reason for enhancement of charge carrier density and accelerated reactions with H2S.

Temperature-Dependent Abnormal and Tunable p-n Response of Tungsten Oxide--Tin Oxide Based Gas Sensors.

We observed the sensing response of temperature-dependent abnormal p-n transitions in WO3-SnO2 hybrid hollow sphere based gas sensors for the first time. The sensors presented a normal n-type response to ethanol at elevated temperatures but abnormal p-type responses in a wide range of operation temperatures (room temperature to about 95 °C). By measuring various reducing gases and applying complex impedance plotting techniques, we demonstrated the abnormal p-type sensing behavior to be a pseudo-response resulting from the reaction between target gas and adsorbed water on the material surface. The temperature-controlled n-p switch is ascribed to the competition of intrinsic and extrinsic sensing behaviors, which resulted from the reaction of target gas with adsorbed oxygen ions and protons from adsorbed water, respectively. The former can modulate the intrinsic conductivity of the sensor by changing the electron concentration of the sensing materials, while the latter can regulate the conduction of the water layer, which contributes to the total conductivity as an external part. The hollow and hybrid nanostructures facilitated the observation of extrinsic sensing behaviors due to its large-area active sites and abundant oxygen vacancies, which could enhance the adsorption of water. This work might give new insight into gas sensing mechanisms and opens up a promising way to develop practical temperature and humidity controllable gas sensors with little power consumption based on the extrinsic properties.

Surrounding sensitive electronic properties of Bi2Te3 nanoplates-potential sensing applications of topological insulators.

Significant efforts have been paid to exploring the fundamental properties of topological insulators (TIs) in recent years. However, the investigation of TIs as functional materials for practical device applications is still quite limited. In this work, electronic sensors based on Bi2Te3 nanoplates were fabricated and the sensing performance was investigated. On exposure to different surrounding environments, significant changes in the conducting properties were observed by direct electrical measurements. These results suggest that nanostructured TIs hold great potential for sensing applications.

Strongly coupled hybrid nanostructures for selective hydrogen detection--understanding the role of noble metals in reducing cross-sensitivity.

Noble metal-semiconductor hybrid nanostructures can offer outperformance to gas sensors in terms of sensitivity and selectivity. In this work, a catalytically activated (CA) hydrogen sensor is realized based on strongly coupled Pt/Pd-WO3 hybrid nanostructures constructed by a galvanic replacement participated solvothermal procedure. The room-temperature operation and high selectivity distinguish this sensor from the traditional ones. It is capable of detecting dozens of parts per million (ppm) hydrogen in the presence of thousands of ppm methane gas. An insight into the role of noble metals in reducing cross-sensitivity is provided by comparing the sensing properties of this sensor with a traditional thermally activated (TA) one made from the same pristine WO3. Based on both experimental and density functional theory (DFT) calculation results, the cross-sensitivity of the TA sensor is found to have a strong dependence on the highest occupied molecular orbital (HOMO) level of the hydrocarbon molecules. The high selectivity of the CA sensor comes from the reduced impact of gas frontier orbitals on the charge transfer process by the nano-scaled metal-semiconductor (MS) interface. The methodology demonstrated in this work indicates that rational design of MS hybrid nanostructures can be a promising strategy for highly selective gas sensing applications.

Extracting valley-ridge lines from point-cloud-based 3D fingerprint models.

3D fingerprinting is an emerging technology with the distinct advantage of touchless operation. More important, 3D fingerprint models contain more biometric information than traditional 2D fingerprint images. However, current approaches to fingerprint feature detection usually must transform the 3D models to a 2D space through unwrapping or other methods, which might introduce distortions. A new approach directly extracts valley-ridge features from point-cloud-based 3D fingerprint models. It first applies the moving least-squares method to fit a local paraboloid surface and represent the local point cloud area. It then computes the local surface's curvatures and curvature tensors to facilitate detection of the potential valley and ridge points. The approach projects those points to the most likely valley-ridge lines, using statistical means such as covariance analysis and cross correlation. To finally extract the valley-ridge lines, it grows the polylines that approximate the projected feature points and removes the perturbations between the sampled points. Experiments with different 3D fingerprint models demonstrate this approach's feasibility and performance.