Laser beam steering technology is essential for modern consumer and scientific optical devices including displays, microscopy, and Light Detection and Ranging (LIDAR) systems. Along with mechanical and completely non-mechanical beam steering approaches, Micro Electro Mechanical Systems (MEMS) are emerging beam steering devices that are especially suitable for LIDAR systems due to their fast scan rate and large scan angle. A class of MEMS-based devices, the Digital Micromirror Device (DMD), has been demonstrated for beam steering too by synchronizing its mirror movement to laser pulse. The tilt movement of micromirrors synchronizes with multiple pulses from multiple laser sources that sequentially redirect the pulses to multiple diffraction orders within μs. Based on the beam steering principle, multi-beam and multi-pulse beam steering in single-chip DMD LIDAR architecture provides a pathway to fast distance range finding having over 1M samples/s scan rate by leveraging a commercially available DMD, laser diodes and drivers. As a proof of concept, 3.34kHz and 15 points of range finding is demonstrated by using three pulsed laser diodes operating at 905nm. Additionally, multi-pulse beam steering for 5 points with an increased scanning rate of 6.63kHz demonstrates further enhancement of the scanning speed. The approach opens up a pathway to achieve a LIDAR system with a scanning rate over 1M samples/s while leveraging a state of the art DMD and a moderate number of laser sources.
Spectral matched filtering and its variants (e.g. Adaptive Coherence Estimator or ACE) rely on strong assumptions about target and background distributions. For instance, ACE assumes a Gaussian distribution of background and additive target model. In practice, natural spectral variation, due to effects such as material Bidirectional Reflectance Distribution Function, non-linear mixing with surrounding materials, or material impurities, degrade the performance of matched filter techniques and require an ever-increasing library of target templates measured under different conditions. In this work, we employ the contrastive loss function and paired neural networks to create data-driven target detectors that do not rely on strong assumptions about target and background distribution. Furthermore, by matching spectra to templates in a highly nonlinear fashion via neural networks, our target detectors exhibit improved performance and greater resiliency to natural spectral variation; this performance improvement comes with no increase in target template library size. We evaluate and compare our paired neural network detector to matched filter-based target detectors on a synthetic hyperspectral scene and the well-known Indian Pines AVIRIS hyperspectral image.
In this effort, random noise data augmentation is compared to phenomenologically-inspired data augmentation for a target detection task, evaluated on the Digital Imaging and Remote Sensing Image Generation (DIRSIG) model “MegaScene” simulated hyperspectral dataset. Random data augmentation is commonly used in the machine learning literature to improve model generalization. While random perturbations of an input may work well in certain fields such as image classification, they can be unhelpful in other applications such as hyperspectral target detection. For instance, random noise augmentation may not be beneficial when the applied noise distribution does not match underlying physical signal processes or sensor noise. In the context of a low-noise sensor, augmentation mimicking material mixing and other practical spectral modulations is likely to be more effective when used to train a target detector. It is therefore important to utilize a data augmentation strategy that emulates the natural variability in observed spectra. To validate this claim, a small fully connected neural network architecture is trained using an ideal hemispheric reflectance materials dataset as a trivial baseline. That dataset is then augmented using Gaussian random noise and the model is retrained and again applied to MegaScene. Finally, augmentation is instead performed using phenomenological insight and used to retrain and reevaluate the model. In this work, the phenomenological augmentation implements only simple and commonly encountered spectral permutations, namely linear mixing and shadowing. Comparison is made between the augmented models and the baseline model in terms of low constant false alarm rate (CFAR) performance.
Spatial light modulators (SLMs) that operate in a phase modulation mode enable beam steering with higher diffraction efficiency compared to amplitude modulation mode, thus potentially be used for an efficient beam steering with no moving part. Currently, Twisted Nematic phase SLMs are widely adopted for phase modulation. However, their refresh rate is typically in the range below kilohertz. Recently, a new method for binary and spatial phase modulation using Digital Micromirror Device (DMD) was proposed by a research group in Germany. In the method, complemental self-images of DMD, corresponding to on- and off-pixels, are formed by two auxiliary optics while adding a pi phase shift between two images. The optics function as recycling of light in a coherent manner. The method enables over kilohertz refresh rate and higher diffraction efficiency in binary phase modulation mode to conventional amplitude binary modulation.
As alternatives to the binary phase modulation, we propose and experimentally evaluated high-speed beam steering by DMD based on light recycling. In our experiment, with binary phase modulation mode, system output efficiency reaches 8%. It can be doubled to 16% with light recycling method. Efficiency is still low compared to the reported value of 27% without light recycling. To further increase beam efficiency, system loss was analysed.
A novel method of beam steering, utilizing a mass-produced Digital Micromirror Device (DMD), enables a reliable single chip Light Detection and Ranging (LIDAR) with a large field of view while having minimum moving components. In the single-chip LIDAR, a short-pulsed laser is fired in a synchronous manner to the micromirrors rotation during the transitional state. Since the pulse duration of the laser pulse is substantially short compared to the transitional time of the mirror rotation, virtually the mirror array is frozen in transition at several discrete points, which forms a programmable and blazed grating. The programmable blazed grating efficiently redirects the pulsed light to a single diffraction order among several while employing time of flight measurement. Previously, with a single 905nm nanosecond laser diode and Si avalanche photo diode, a measurement accuracy and rate of <1 cm and 3.34k points/sec, respectively, was demonstrated over a 1m distance range with 48° full field of view and 10 angular resolution. We have also increased the angular resolution by employing multiple laser diodes and a single DMD chip while maintaining a high measurement rate of 3.34k points/s. In addition, we present a pathway to achieve 0.65° resolution with 60° field of view and 23k points/s measurement rate.
An imaging lidar system is presented which combines the high speed of a Digital Micromirror Device (DMD) and the higher range of a 1D collimated scanning output. The system employing 1D line object illumination along with DMD placed at focal plane enables flexible optimization of system metrics, such as field of view, angular resolution, maximum range distance and frame rate.
A novel Digital Micromirror Device (DMD) based beam steering enables a single chip Light Detection and Ranging (LIDAR) system for discrete scanning points. We present increasing number of scanning point by using multiple laser diodes for Multi-beam and Single-chip DMD-based LIDAR.