As the failure of power line insulators leads to transmission systems, an insulator inspection system based on aerial platform is widely used. Insulator defect detection performed against complex backgrounds in images, presenting interesting but challenging problem. Traditional methods, handcrafted features or shallow learning techniques, can only localize and detect faults under specific conditions, such as when sufficient prior knowledge available, with low background interference, at...

Automatic metallic surface defect inspection has received increased attention in relation to the quality control of industrial products. Metallic detection is usually performed against complex scenarios, presenting an interesting but challenging problem. Traditional methods are based on image processing or shallow machine learning techniques, these can only detect defects under specific conditions, such as obvious contours with strong contrast and low noise, at certain scales, illumination...

Weld bead inspection is important for high-quality welding. This paper summarizes our work on weld profile measurement, monitoring, and defect detection using a structured light-based vision system. The configuration of the sensor described analyzed. In this configuration, system presented in can easily be calibrated. image processing extraction algorithms laser profiles feature points are presented. dimensional parameters measured, defects detected during multilayer welding processes....

In recent years, the application of deep learning based on convolutional neural networks has gained great success in face detection. However, one remaining open challenges is detection small-scaled faces. The depth network can cause projected feature map for small faces to be quickly shrunk, and most approaches with scale invariant hardly handle less than 15×15 pixel To solve this problem, we propose a different scales detector (DSFD) Faster R-CNN. new improve precision while performing as...

An automatic system is developed to realize high precision assembly of two components in the size mm level with an interference fit 3-dimensional (3-D) space 6-degree-of- freedoms (DOF), which consists a manipulator, adjusting platform, sensing and computer. The manipulator employed align component B A position. platform aligns orientations inserts into B. includes three microscopes force sensor. are mounted approximately orthogonal observe from different directions aligning stage. sensor...

A regular ceiling is common in many offices. The plentiful parallels and corner points on the can serve as features for visual positioning an indoor mobile robot. Based natural ceiling, a new method proposed. camera mounted top of robot pointed to ceiling. At beginning positioning, initial orientation position world frame estimated with specified block via perspective-n-point-based method. With motions robot, its global calculated from main secondary lines feature when has parallels. In...

A high-speed stereovision system with two smart cameras is presented to track a table tennis ball, which adopts distributed parallel processing architecture based on local area network. set of novel algorithms little computation and good robustness running in the also proposed recognize ball images. computer receives image coordinates from via network computes its 3-D positions working frame. Then, flying trajectory estimated predicted according measured rebound models. The main motion...

In this paper, an intelligent seam-tracking system is developed for fillet weld workpiece. Structured light vision employed to detect the deviation of welding torch. Two stripe lines are formed on surface workpiece by structured laser light. Their intercepts selected as image feature, which can result in approximately uncoupled relationship between feature and movement adjustment mechanism. A two-step feature-extraction method designed robustly two image. order get a favorable tracking...

In this paper, we propose a robust visual detection-learning-tracking framework for autonomous aerial refueling of unmanned vehicles. Two classifiers (D-classifier and T-classifier) are defined in the proposed framework. The D-classifier is linear support vector machine (SVM) classifier trained offline detecting drogue object low-dimensional normalized local binary pattern feature to describe D-classifier. T-classifier state-based structured SVM online tracking object. A combination strategy...

As a critical electrical connector component in the modern industrial environment, spring-wire sockets and their manufacture quality are closely relevant to equipment safety. These types of defects difficult properly distinguish due defect similarity diversity. In such cases, can only be determined using cumbersome human visual inspection. To satisfy requirements control, machine vision apparatus for inspection is presented this paper. With brief description system design, our emphasis put...

Visual tracking of free-swimming robotic fish remains a great challenge by considering imaging qualities in aquatic environments. In this paper, we propose visual identification and positioning method to obtain accurate position for 3-D control comprising depth directional control. Specifically, based on fuzzy sliding-mode is put forward make the swim preferred target remain at that depth. A with multiple stages proposed, series strategies developed integrate agile locomotion accuracy....

A vision-based target-following guider for mobile robot is presented in this paper. It consists of three parts: the visual tracking part; target redetection and servo part. In part, contour band proposed irregular sampling, which can improve performance correlation filter-based methods. potential targets are searched by an offline trained detector. Then online training module used to determine real achieve accurate positioning. The interaction matrix point features part motion control...

In this paper, a position measurement system, including drogue's landmark detection and computation for autonomous aerial refueling of unmanned vehicles, is proposed. A multitask parallel deep convolution neural network (MPDCNN) designed to detect the landmarks drogue target. MPDCNN, two networks are used, fusion mechanism proposed accomplish effective salient parts' detection. Considering target's geometric constraints, method based on monocular vision An strategy, which fuses results...

In this article, a robust insertion control method for precision peg-in-hole assembly is developed by visually supervising the deformation of passive compliance based on microscopic vision and force information. The advantage proposed task can be conducted with pose alignment errors since supervised. This significantly decreases requirement parts thus contributes to assembly. Specifically, relationship between contact status among deformations first analyzed. Afterward, vision-based estimate...

This paper proposes a skill learning approach for precision assembly robot, aiming to realize efficient transfer from teacher robot through several demonstrations. The framework is designed considering that has multiple controllers and procedures. A complex segmented an action sequence according the changes of teacher's selective attention settings on system variables. each select predefined class learn its key parameters demonstration data. forms finite-state machine. To execute action,...

Multiple peg-in-hole insertion control is one of the challenging tasks in precision assembly for its complex contact dynamics. In this article, an policy learning method proposed multiple assembly. The process separated into two phases: initial and residual learning. learning, a state-to-action mapping model based on Gaussian mixture (GMM) established. And regression (GMR) used to generalize reuse. reinforcement named normalized advantage function (NAF) employed refine via agent's...

Biomedical microelectrode implantation is a promising approach to achieve invasive brain-machine interface with high bandwidth and good biocompatibility. However, it difficult manipulate probe because of its micron-level width flexible material. Toward the challenging robotic implantation, this article addresses an essential operation step: automated hooking micro-needle. First, point-contour interest extraction network for microscopic image (MicroPCIE) proposed obtain key features visual...

In the wagyu industry worldwide, high-quality marbling beef is produced by promoting intramuscular fat deposition during cattle fattening stage through dietary vitamin A control. Thus, however, become susceptible to either deficiency or excess state, not only influencing performance and quality, but also causing health problems. Researchers have been exploring eye photography monitoring methods for blood levels based on relation between retina colour changes. But previous endeavours cannot...