Subspace Clustering for Situation Assessment in Aquatic Drones: A Sensitivity Analysis for State-Model Improvement

Abstract

In this paper, we propose the use of subspace clustering to detect the states of dynamical systems from sequences of observations. In particular, we generate sparse and interpretable models that relate the states of aquatic drones involved in autonomous water monitoring to the properties (e.g., statistical distribution) of data collected by drone sensors. The subspace clustering algorithm used is called SubCMedians. A quantitative experimental analysis is performed to investigate the connections between i) learning parameters and performance, ii) noise in the data and performance. The clustering obtained with this analysis outperforms those generated by previous approaches.     

Chromatic adaptation in a 2D photograph demonstrated by a D‐up viewer

It is said that we cannot have color constancy in a photograph. The concept of recognized visual space of illumination (RVSI) asserts that chromatic adaptation occurs when one perceives the illumination that is filling a space and not the objects in the space. It predicts then that if one perceives a 3D scene in a photograph, then color constancy will occur in the photograph. In this work, a dimension‐up (D‐up) viewer was developed to perceive a 3D scene on a 2D photograph, and the effect of chromatic adaptation was measured by the color appearance of a gray patch placed at the center of the photograph. Subjects saw the patch as a vivid color when they saw a photograph that had been taken under colored illumination, which is a normal experience in a real space observation. When the color appearance was measured by the elementary color naming method, the amount of chromaticness of the patch in percentage and the apparent hue were very similar to those observed in the 2‐room technique, thus confirming the prediction by the RVSI theory.     

Visual customization: Diversity in color preferences in the automotive interior and implications for interior design

Evolving multimodal mobility needs influences established human–product relationships and requires a deeper insight into color preferences for car interiors. Hence, a study was conducted in which 204 members of a web contest created 1,265 designs. After a peer‐evaluation process, 53 most‐appreciated and 34 least‐preferred interior color compositions were identified and compared to identify patterns in color choices. Besides, visual lightweight design by layering of large interior components such as the seat, a modest use of color and patterns accompanied by repetition and the framing of the entire interior to create a feeling of spaciousness were found. Additionally, differences in the type of color between most and least favored color designs were found. Brown and beige occur more frequently among the top‐ than the worst‐rated designs. Larger surfaces are favored in lighter hues and smaller components in darker hues.     

Automatic restraint and visual detection of fatigue crack growth when applying an anti-fatigue smart paste to bolt holes

An anti-fatigue smart paste, which consists of fine alumina particles and silicone grease with low viscosity, was applied to a bolt hole (and its periphery in some cases) in a steel plate specimen, and the effects of the smart paste on automatic restraint and visual detection of fatigue crack growth were experimentally investigated through fatigue tests. Fractographic observations using a scanning electron microscope were also carried out. As a result, approximately 20～410% increase in failure life was produced by the wedge effect of the alumina particles in the smart paste. When the smart paste was applied, remarkable black colour developed in the white paste along the paths of crack growth, exceedingly facilitating the visual detection of the crack growth.     

Molecular Recognition in C-Type Lectins: The Cases of DC-SIGN,Langerin, MGL,and L-Sectin

Carbohydrates play a pivotal role in intercellular communication processes. In particular, glycan antigens are key for sustaining homeostasis, helping leukocytes to distinguish damaged tissues and invading pathogens from healthy tissues. From a structural perspective, this cross-talk is fairly complex, and multiple membrane proteins guide these recognition processes, including lectins and Toll-like receptors. Since the beginning of this century, lectins have become potential targets for therapeutics for controlling and/or avoiding the progression of pathologies derived from an incorrect immune outcome, including infectious processes, cancer, or autoimmune diseases. Therefore, a detailed knowledge of these receptors is mandatory for the development of specific treatments. In this review, we summarize the current knowledge about four key C-type lectins whose importance has been steadily growing in recent years, focusing in particular on how glycan recognition takes place at the molecular level, but also looking at recent progresses in the quest for therapeutics.     

基于混合特征LLE融合与SVM的质量异常模式识别

针对生产过程中存在的异常模式识别的问题,提出基于LLE融合与支持向量机的质量异常模式识别方法。首先从动态数据流中提取其原始特征、统计特征、几何特征并将其进行混合,形成动态数据流的混合特征,然后利用LLE算法对混合特征进行降维,将降维后的特征集作为MSVM分类器的输入进行训练,同时采用粒子群算法对MSVM分类器进行参数寻优。最后用训练好的模型对动态数据流进行异常模式的识别。并将所提方法与单一类型特征方法、混合特征方法的识别模型进行比较,仿真结果和应用实例表明,所提方法的识别精度较高,可用于生产过程的质量异常模式识别中。     

基于动态行为学习的空中目标识别方法

针对空中移动目标识别问题,提出了一种基于动态行为学习的识别方法。首先,从数据源中提取目标要素,经过标注形成目标数据特征集;其次,基于应用场景进行特征组合,形成特征模型要素;然后,基于深度学习算法进行模型训练,得到评估最优的模型参数;最后,利用训练后的模型进行目标识别处理。通过数值仿真验证,该方法能够有效地识别无业务规则的目标身份。     

基于改进Faster R-CNN及YOLO的大角度车牌识别算法

随着大数据时代的到来,智慧城市等一系列概念的提出,人工智能开始在城市各个角落得到广泛应用。其中,车牌识别作为城市道路交通重要的一环,也取得了重大突破。基于深度学习的车牌识别一经提出,各类算法应用于车牌检测及识别上,对比于传统的车牌识别,大大提高了识别的速率和准确率。然而,在非限制条件下,如大角度车牌识别准确率仍有提升的空间。提出了一种改进的FasterR-CNN与YOLO的深度学习车牌识别算法,将对大角度下的车牌识别准确率达到了99.7%。由于是轻量型的框架,为后续的移动设备部署工作提供了便利。     

Emotion recognition at the edge with AI specific low power architectures

Nowadays Deep Learning is applied in almost every research field and helps getting amazing results in a great number of challenging tasks. The main problem is that this kind of learning and consequently Neural Networks that can be defined deep, are resource intensive. They need specialized hardware to perform computation in a reasonable time. Many tasks are mandatory to be as much real-time as possible . It is needed to optimize many components such as code, algorithms, numeric accuracy and hardware, to make them “efficient and usable”. All these optimizations can help us to produce incredibly accurate and fast learning models. The paper reports a study in this direction for the challenging face detection and emotion recognition tasks.