This paper proposes a deep autoencoder-based approach to identify signal features from lowlight images and adaptively brighten images without over-amplifying/saturating the lighter parts in images with a high dynamic range. In surveillance, monitoring and tactical reconnaissance, gathering visual information from a dynamic environment and accurately processing such data are essential to making informed decisions and ensuring the success of a mission. Camera sensors are often cost-limited to capture clear images or videos taken in a poorly-lit environment. Many applications aim to enhance brightness, contrast and reduce noise content from the images in an on-board real-time manner. We show that a variant of the stacked-sparse denoising autoencoder can learn to adaptively enhance and denoise from synthetically darkened and noise-added training examples. The model can be applied to images taken from natural lowlight environment and/or are hardware-degraded. Results show significant credibility of the approach both visually and by quantitative comparison with various image enhancement techniques.

In order to identify and control the menace of destructive pests via microscopic image-based identification state-of-the art deep learning architecture is demonstrated on the parasitic worm, the soybean cyst nematode (SCN), Heterodera glycines. Soybean yield loss is negatively correlated with the density of SCN eggs that are present in the soil. While there has been progress in automating extraction of egg-filled cysts and eggs from soil samples counting SCN eggs obtained from soil samples using computer vision techniques has proven to be an extremely difficult challenge. Here we show that a deep learning architecture developed for rare object identification in clutter-filled images can identify and count the SCN eggs. The architecture is trained with expert-labeled data to effectively build a machine learning model for quantifying SCN eggs via microscopic image analysis. We show dramatic improvements in the quantification time of eggs while maintaining human-level accuracy and avoiding inter-rater and intra-rater variabilities. The nematode eggs are correctly identified even in complex, debris-filled images that are often difficult for experts to identify quickly. Our results illustrate the remarkable promise of applying deep learning approaches to phenotyping for pest assessment and management.

Non-intrusive load monitoring (NILM) of electrical demand for the purpose of identifying load components has thus far mostly been studied using univariate data, e.g., using only whole building electricity consumption time series to identify a certain type of end-use such as lighting load. However, using additional variables in the form of multivariate time series data may provide more information in terms of extracting distinguishable features in the context of energy disaggregation. In this work, a novel probabilistic graphical modeling approach, namely the spatiotemporal pattern network (STPN) is proposed for energy disaggregation using multivariate time-series data. The STPN framework is shown to be capable of handling diverse types of multivariate time-series to improve the energy disaggregation performance. The technique outperforms the state of the art factorial hidden Markov models (FHMM) and combinatorial optimization (CO) techniques in multiple real-life test cases. Furthermore, based on two homes' aggregate electric consumption data, a similarity metric is defined for the energy disaggregation of one home using a trained model based on the other home (i.e., out-of-sample case). The proposed similarity metric allows us to enhance scalability via learning supervised models for a few homes and deploying such models to many other similar but unmodeled homes with significantly high disaggregation accuracy.

The thermo-acoustic instabilities arising in combustion processes cause significant deterioration and safety issues in various human-engineered systems such as land and air based gas turbine engines. The phenomenon is described as selfsustaining and having large amplitude pressure oscillations with varying spatial scales of periodic coherent vortex shedding. Early detection and close monitoring of combustion instability are the keys to extending the remaining useful life (RUL) of any gas turbine engine. However, such impending instability to a stable combustion is extremely difficult to detect only from pressure data due to its sudden (bifurcationtype) nature. Toolchains that are able to detect early instability occurrence have transformative impacts on the safety and performance of modern engines. This paper proposes an endto- end deep convolutional selective autoencoder approach to capture the rich information in hi-speed flame video for instability prognostics. In this context, an autoencoder is trained to selectively mask stable flame and allow unstable flame image frames. Performance comparison is done with a wellknown image processing tool, conditional random field that is trained to be selective as well. In this context, an informationtheoretic threshold value is derived. The proposed framework is validated on a set of real data collected from a laboratory scale combustor over varied operating conditions where it is shown to effectively detect subtle instability features as a combustion process makes transition from stable to unstable region.

This paper presents a novel data-driven technique based on the spatiotemporal pattern network (STPN) for energy/power prediction for complex dynamical systems. Built on symbolic dynamic filtering, the STPN framework is used to capture not only the individual system characteristics but also the pair-wise causal dependencies among different sub-systems. For quantifying the causal dependency, a mutual information based metric is presented. An energy prediction approach is subsequently proposed based on the STPN framework. For validating the proposed scheme, two case studies are presented, one involving wind turbine power prediction (supply side energy) using the Western Wind Integration data set generated by the National Renewable Energy Laboratory (NREL) for identifying the spatiotemporal characteristics, and the other, residential electric energy disaggregation (demand side energy) using the Building America 2010 data set from NREL for exploring the temporal features. In the energy disaggregation context, convex programming techniques beyond the STPN framework are developed and applied to achieve improved disaggregation performance.