3D Reconstruction based on NIR single-pixel for drone navigation under rainy condition

3D reconstruction has become a challenging task for navigation systems in recent years. Therefore, different technologies such as RGB cameras, LIDARs, or RADAR can capture information from the environment and perform the 3D reconstruction. However, in an environment of high dispersion and low illumination, it is necessary to have a robust solution that operates in the infrared spectrum. One solution is to integrate Single-Pixel Near Infra-Red Imaging (NIR-SPI) technology, allowing mage reconstruction of a few samples and operating high dispersion conditions. Using an SPI-NIR vision system with active illumination for 3D image reconstruction in rainy environments is a solution. The reconstruction of 3D objects is performed from their construction of a low-resolution SPI-NIR 2D image, using a robust unified reflectance model that combines the Lambertian, Oren-Nayar, and Blinn-Phong models to improve the 3D image of objects with surface roughness or with low reflectance. For 3D reconstruction, single-view Shape-From-Shading (SFS) based on fast Eikonal solvers was used. Which makes it possible to improve the shape of the 3D object, reduce computation time for future applications and generate real-time 3D images in harsh environments.

Single-Pixel Imaging a solution 3D reconstruction in harsh environments

Over the last decades, different imaging solutions have been proposed to achieve three-dimensional (3D) space reconstruction and obstacle detection. Either based on stereo-vision principles using active pixel (RGB) sensors operating in the visible (VIS) part of the spectra, or applying the Time-of-flight (TOF) principle based on active Near Infra-Red (NIR) illumination, to mention just a few. Suppose a silicon-based detector yields extremely low quantum efficiencies for NIR active illumination. Solutions are considered together with the huge photon noise levels produced by the background illumination accompanied by Rayleigh scattering effects in outdoor applications and the operating limitations of these systems in harsh weather conditions, especially if relatively low-power active illumination is used result evident. Suppose longer wavelengths for active illumination are applied to overcome these issues. In that case, Indium Gallium Arsenide (InGaAs) based photodetectors become the technology of choice, and for low-cost solutions, using a single InGaAs photodetector or an InGaAs line-sensor become a promising choice. In this case, the principles of Single-Pixel Imaging (SPI) and Compressive Sensing (CS) acquire paramount importance. The introduction of a Near Infra-Red Single Pixel Imaging (NIR-SPI) sensor aimed at detecting static and dynamic objects under outdoor conditions increased applications field widely from Unmanned Aerial Vehicle (UAV) to sensor redundant for vehicles.

SPI simulation results of different algorithms under different sampling ratios of : a) Sampling ratio=0.2, b) Sampling ratio=0.8, and c) Sampling ratio=3. We can see CS and TV methods need few samples for SPI reconstruction, while DGI, GD, and Poisson need a higher sampling ratio >1.

Araguaney CS1

The first solar vehicle of Venezuela, built-in 2007, participated in the World Solar Challenge in Australia (WSC).

Technical characteristics: Dimensions 5x1.8 cell arrangement 6m2 with a power of 1920W, ABB AC asynchronous motor of 4.5KW, maximum speed 70km/h, Bus DC 72V LIPO battery with an autonomy of 4hs and telemetry system.

---

https://www.youtube.com/watch?v=xDk8dpy0UGk&t=1s