Traditionally, the hinge loss is used to construct support vector machine (SVM) classifiers. The related shortest distance between sets and corresponding classifier hence sensitive noise unstable for re-sampling. In contrast, pinball quantile result less sensitive. has been deeply studied widely applied in regression but it not classification. this paper, we propose a SVM with loss, called pin-SVM, investigate its properties, including insensitivity, robustness, misclassification error....

In the field of fabric defect detection, development algorithms has been hindered by issues such as poor quality and limited quantity open-source datasets. Traditional data augmentation methods offer improvements in model performance, while generative are plagued difficulties training models, susceptibility to artifacts, need for re-labeling. To address these challenges, this paper proposes a blind super-resolution algorithm augmentation. The is based on Real-ESRGAN optimized specifically...

Scene classification is a fundamental perception task for environmental understanding in today's robotics. In this paper, we have attempted to exploit the use of popular machine learning technique deep enhance scene understanding, particularly robotics applications. As images larger diversity than iconic object images, it more challenging methods automatically learn features from with less samples. Inspired by human based on knowledge, address problem encouraging neural networks incorporate...

Acquiring the accurate three-dimensional (3-D) position of a target person around robot provides valuable information that is applicable to wide range robotic tasks, especially for promoting intelligent manufacturing processes industries. This paper presents real-time 3-D human tracking system combines monocular camera with an ultrasonic sensor by extended Kalman filter (EKF). The proposed consists three submodules: module, and multisensor fusion algorithm. An improved visual algorithm...

This paper presents a new web mining scheme for parallel data acquisition.Based on the Document Object Model (DOM), page is represented as DOM tree.Then tree alignment model proposed to identify translationally equivalent texts and hyperlinks between two trees.By tracing identified hyperlinks, documents are recursively mined.Compared with previous schemes, benchmarks show that this improves coverage, reduces bandwidth, enhances quality of mined sentences.

Abstract This paper addresses automated mapping of the remaining wall thickness metallic pipelines in field by means an inspection robot equipped with nondestructive testing (NDT) sensing. Set context condition assessment critical infrastructure, integrity arbitrary sections conduit is derived a bespoke kinematic configuration that allows dense pipe discrimination circumferential and longitudinal direction via NDT sensing guaranteed lift‐off (offset sensor from wall) to wall, essential...

The perception of the environment around a land vehicle plays crucial role for its driving assistant system. Knowledge road terrain is useful handling characteristics while vehicles and improving passengers' safety comfort. In this article, an approach to classifying road-terrain presented. An accelerometer mounted on suspension measure vibration that represents terrain, profile can be calculated by knowing speed one-quarter-dynamic model vehicle. optimized classifier features, independency,...

Extending metric space representations of an environment with other high level information, such as semantic and topological enable a robotic device to efficiently operate in complex environments. This paper proposes methodology for robot classify indoor environments into categories. Classification task, using data collected from laser range finder, is achieved by machine learning approach based on the logistic regression algorithm. The classification followed probabilistic temporal update...

Acquiring the accurate 3-D position of a target person around robot provides fundamental and valuable information that is applicable to wide range robotic tasks, including home service, navigation entertainment. This paper presents real-time human tracking system which combines monocular camera with an ultrasonic sensor by extended Kalman filter (EKF). The proposed consists three sub-modules: model, model multi-sensor fusion. An improved visual algorithm presented provide partial location...

Place classification is a fundamental ability that robot should possess to carry out effective human-robot interactions. In recent years, there high exploitation of artificial intelligence algorithms in robotics applications. Inspired by the successes deep learning methods, we propose an end-to-end approach for place problem. With architectures, this methodology automatically discovers features and contributes general higher accuracies. The pipeline our composed three parts. First, construct...

Recent studies on the myoelectric control of powered prosthetics revealed several factors that affect its clinical performance. One important is variation in limb position associated with normal use which can have a substantial impact robustness Electromyogram (EMG) pattern recognition. To solve this problem, we propose paper new feature extraction algorithm based set spectral moments extracts relevant information about EMG power spectrum an accurate and efficient manner. The main goal to...

In the field of video-based fire detection, traditional color models have poor adaptability and susceptibleness to environmental interference. As can be divided into core, inner outer layers depending on its temperature color, a new detection model based dispersion components has been proposed by author. order further reduce false alarm, similarity consecutive video frames detected area is incorporated method, tracking algorithm flame kinematic characteristics proposed. Experimental results...

Human robot interaction is an emerging area of research, where human understandable robotic representations can play a major role. Knowledge semantic labels places be used to effectively communicate with people and develop efficient navigation solutions in complex environments. In this paper, we propose new approach that enables learn classify observations indoor environment using labeled grid map, which similar Occupancy Grid like representation. Classification the based on data collected...

Electric insulators as an indispensable device for electric power networks, maintaining its safe operation is of vital importance. Due to the large number and wide distribution, insulator state detection based on aerial images has important practical significance. Insulator are usually acquired by artificial or collection, at a specific angle, focal length complex background. For labor-detection, low efficiency, higher cost other interference, efficient accurate method proposed detect kinds...

Representation of spaces including both geometric and semantic information enables a robot to perform high-level tasks in complex environments. Therefore, recent years identifying semantically labeling the environments based on onboard sensors has become an important competency for mobile robots. Supervised learning algorithms have been extensively used this purpose with SVM-based solutions showing good generalization properties. The CRF-based approaches take advantage connectivity samples...

Road terrain identification is one of the important tasks for driving assistant systems or autonomous land vehicles. It plays a key role in improving strategy and enhancing fuel efficiency. In this paper, two-stage approach using multiple sensors presented. first stage, feature-based performed an accelerometer, camera, downward-looking forward-looking laser range finders (LRFs). This produces four classification label sequences. second majority vote implemented each sequences to match them...