Search

The number of patients who suffer from Binge Eating Disorder has been increased for the last 2 years. By understanding the causes and effects of Binge Eating Behavior, it is more accurate to find solutions since BED is a complexed clinic illness affecting both mental and physical. Majority of the patients are not seeking help. I want to design an application utilizing Augmented Reality and Artificial Intelligence to help them gaining back control of their actions and emotions; also, to rebuild their normal healthy eating habits.

Many studies have elucidated the importance of water stress on plant growth, yield, quality and susceptibility to disease. Via the vascular network of xylem, a water conductive tissue, plant water stress responds dynamically to variations in evaporative demand defined by micrometeorological conditions over minutes to hours, and soil water availability over hours to months. Stem water potential is believed to be an integrator of water stress across the soil-plantatmosphere-continuum (SPAC), and is difficult to measure. Despite its destructive nature, 50 years after its invention, the manually operated Schölander pressure chamber (SPC) is still the most widely accepted tool for stem water potential measurements. Limitations in available techniques have hindered the study of dynamic water stress in plants. In this dissertation, we introduce a micro-tensiometer (µTM) as a new technique, for probing the dynamic water stress of plants in an accurate and continuous manner. We examine the reliability of µTM against SPC, on apple (2 months), grapevine (12 months), and almond (4 months), to represent woody species in wet, semi-arid, and arid environments respectively. We observe: 1) nighttime disequilibrium in stem water potential that is challenging to acquire with the labor-intensive SPC; 2) rapid response of stem water potential to evaporative demand in wet environment; and 3) slow dynamics and persistent disequilibrium of the water-stressed almond in dry environment. With the advantage of continuous measurements and inspired by van den Honert, we use circuit models to interpret the observed dynamics. In a wet environment, a simple circuit with a single hydraulic resistance and a single hydraulic capacitance, is sufficient for elucidating the rapid response of plants to the high frequent variations in environmental demand. In a dry environment,an additional soil compartment defined by the soil retention properties, is used to address the long transient of soil dehydration. We now have models as new tools to resolve the complex dynamics of water stress. We also measure the dynamic water stress in maize, the first examination on herbaceous crops with this tool. The measured stress is less coupled to the rapid variations in evaporative demand, but more to the soil water potential around the roots. In fact, we extract an empirical water retention curve for the soil that coincides with the theoretical prediction. The µTM, therefore, opens up an opportunity to monitor the root-zone soil stress, a challenging property to access. Finally, we explore the response of plants to fine control of irrigation events, and discover that the transient of root response to irrigation events is shorter when less stressed (nighttime) and longer when more stressed (daytime). This phenomenon suggests more effective irrigation events when plants are less stressed with reduced water loss. The micro-tensiometer and the developed circuit models, together, provide opportunities to unveil the full dynamics of plant water stress, address the transient factor in plant physiological responses to both short and long-term dehydration processes, and guide more efficient management of agricultural water use.

Climate change has caused extreme weather conditions, and resulted in a large water stress in agriculture. Monitoring plant water stress is crucial for both the study of on plant drought responses and the improvement of the agricultural water use efficiency. However, current commercially available water stress sensors either lack of accuracy and resolution, or are too complicated to use. In this study, we developed a micro-tensiometer (µTM), which measures plant water stress in real time by monitoring the stem water potential (Ψ_stem) and the soil water potential (Ψ_soil) - the two most important plant water stress indicators - with high accuracy, high resolution, minimum sample destruction, and optimum local geometrical integration with the sample. The μTM translates the water energy state into electronic signal by implementing traditional tensiometry in a microelectromechanical system (MEMS) with the nanoporous silicon membrane (PoSi) technique. This design significantly increased the measurement range from >-0.1 MPa to >-10 MPa. With the MEMS approach, the sensing area was reduced by two orders of magnitude (from >10 cm^2 to 0.25 cm^2). In situ embedding strategies were developed for the µTM through testing on apple trees. In an in-plant experiment (GH2), the µTM (~ -2.5 MPa) showed up to 1.5 MPa difference from the traditional Scholander pressure chamber (~ -1.0 MPa). This result led to the hypothesis that a vapor gap existed between the µTM and the tissue, and could result in a 7.77 MPa error per degree Celsius of temperature difference (ΔT) between the sample and sensor at 25 ºC. Different strategies were tried to reduce the vapor gap in the fourth experiment (GH4). The µTM with the best contact showed a linear correlation (R^2=0.93) with the Scholander. Other discoveries from the GH4, and their related hypotheses were discussed as well.

An enormous amount of digital images are being generated and stored every day. Understanding text in these images is an important challenge with large impacts for academic, industrial and domestic applications. Recent studies address the difficulty of separating text targets from noise and background, all of which vary greatly in natural scenes. To tackle this problem, we develop a text detection system to analyze and utilize visual information in a data driven, automatic and intelligent way. The proposed method incorporates features learned from data, including patch-based coarse-to-fine detection (Text-Conv), connected component extraction using region growing, and graph-based word segmentation (Word-Graph). Text-Conv is a sliding window-based detector, with convolution masks learned using the Convolutional k-means algorithm (Coates et. al, 2011). Unlike convolutional neural networks (CNNs), a single vector/layer of convolution mask responses are used to classify patches. An initial coarse detection considers both local and neighboring patch responses, followed by refinement using varying aspect ratios and rotations for a smaller local detection window. Different levels of visual detail from ground truth are utilized in each step, first using constraints on bounding box intersections, and then a combination of bounding box and pixel intersections. Combining masks from different Convolutional k-means initializations, e.g., seeded using random vectors and then support vectors improves performance. The Word-Graph algorithm uses contextual information to improve word segmentation and prune false character detections based on visual features and spatial context. Our system obtains pixel, character, and word detection f-measures of 93.14%, 90.26%, and 86.77% respectively for the ICDAR 2015 Robust Reading Focused Scene Text dataset, out-performing state-of-the-art systems, and producing highly accurate text detection masks at the pixel level. To investigate the utility of our feature learning approach for other image types, we perform tests on 8- bit greyscale USPTO patent drawing diagram images. An ensemble of Ada-Boost classifiers with different convolutional features (MetaBoost) is used to classify patches as text or background. The Tesseract OCR system is used to recognize characters in detected labels and enhance performance. With appropriate pre-processing and post-processing, f-measures of 82% for part label location, and 73% for valid part label locations and strings are obtained, which are the best obtained to-date for the USPTO patent diagram data set used in our experiments. To sum up, an intelligent refinement of convolutional k-means-based feature learning and novel automatic classification methods are proposed for text detection, which obtain state-of-the-art results without the need for strong prior knowledge. Different ground truth representations along with features including edges, color, shape and spatial relationships are used coherently to improve accuracy. Different variations of feature learning are explored, e.g. support vector-seeded clustering and MetaBoost, with results suggesting that increased diversity in learned features benefit convolution-based text detectors.

An end-to-end license plate recognition (LPR) system is proposed. It is composed of pre-processing, detection, segmentation and character recognition to find and recognize plates from camera based still images. The system utilizes connected component (CC) properties to quickly extract the license plate region. A novel two-stage CC filtering is utilized to address both shape and spatial relationship information to produce high precision and recall values for detection. Floating peak and valleys (FPV) of projection profiles are used to cut the license plates into individual characters. A turning function based method is proposed to recognize each character quickly and accurately. It is further accelerated using curvature histogram based support vector machine (SVM). The INFTY dataset is used to train the recognition system. And MediaLab license plate dataset is used for testing. The proposed system achieved 89.45% F-measure for detection and 87.33% accuracy for overall recognition rate which is comparable to current state-of-the-art systems.