Search results

A fine‐grained video content indexing, retrieval, and adaptation requires accurate metadata describing the video structure and semantics to the lowest granularity, i.e. to the object level. The authors address these requirements by proposing semantic video content annotation tool (SVCAT) for structural and high‐level semantic video annotation. SVCAT is a semi‐automatic MPEG‐7 standard compliant annotation tool, which produces metadata according to a new object‐based video content model introduced in this work. Videos are temporally segmented into shots and shots level concepts are detected automatically using ImageNet as background knowledge. These concepts are used as a guide to easily locate and select objects of interest which are then tracked automatically to generate an object level metadata. The integration of shot based concept detection with object localization and tracking drastically alleviates the task of an annotator. The paper aims to discuss these issues.

A systematic keyframes classification into ImageNet categories is used as the basis for automatic concept detection in temporal units. This is then followed by an object tracking algorithm to get exact spatial information about objects.

Experimental results showed that SVCAT is able to provide accurate object level video metadata.

The new contribution in this paper introduces an approach of using ImageNet to get shot level annotations automatically. This approach assists video annotators significantly by minimizing the effort required to locate salient objects in the video.

To make an augmented image realistic, the virtual objects should be correctly occluded by foreground objects. The purpose of this paper is to propose a new approach that resolves occlusion problems in augmented reality (AR). The main interest is that it can automatically obtain the proper spatial relationship between virtual and real objects in real time.

The approach is divided into two steps: off‐line disparity map constructing and on‐line occlusion handling. In the off‐line stage, the disparity map of the real scene is constructed using the global stereo matching method prior and then the disparities are refined by means of the fast mean shift method. Since the depth values of objects in different positions are different, the real object that occludes the virtual object can be specified according to the depth value. In the on‐line stage, the contour of the specified object is tracked using the real time object tracking method with the combination of feature tracking method and minimum s‐t cut method. The augmented image with correct occlusions is produced by redrawing all the tracked object pixels on the augmented image.

Compared with the existing methods, the proposed approach can automatically resolve occlusion problem in real time. The effectiveness of the method is demonstrated with several experimental results.

This paper makes three contributions. First, a novel framework is proposed to handle occlusion problem in AR. This framework is different to the previously proposed methods. The main procedure includes: obtain occluding real object, track the object, and redraw the pixels of the object on the composed image. It is much easier to implement and can achieve satisfactory results. Second, the disparity map is used to automatically obtain the contour of the occluding real object. To get the contour of the occluding real object precisely, the mean shift method is used to refine the disparity map. By comparing the depth value, the occluding real object can be extracted automatically. Third, the tracking method combining feature tracking method and minimum s‐t cut method ensures the real‐time requirement. The occlusion problem can be handled in real‐time.

The authors want to design an adaptive grasping control strategy without setting the expected contact force in advance to maintain grasping stable, so that the proposed control system can deal with unknown object grasping manipulation tasks.

The adaptive grasping control strategy is proposed based on bang-bang-like control principle and slippage detection module. The bang-bang-like control method is designed to find and set the expected contact force for the whole control system, and the slippage detection function is achieved by dynamic time warping algorithm.

The expected contact force can adaptively adjust in grasping tasks to avoid bad effects on the control system by the differences of prior test results or designers. Slippage detection can be recognized in time with variation of expected contact force manipulation environment in the control system. Based on if the slippage caused by an unexpected disturbance happens, the control system can automatically adjust the expected contact force back to the level of the previous stable state after a given time, and has the ability to identify an unnecessary increasing in the expected contact force.

Only contact force is used as feedback variable in control system, and the proposed strategy can save hardware components and electronic circuit components for sensing, reducing the cost and design difficulty of conducting real control system and making it easy to realize in engineering application field. The expected contact force can adaptively adjust due to unknown disturbance and slippage for various grasping manipulation tasks.

The purpose of this paper is to propose a method for hand measurement based on image and marker watershed algorithm, and combine the data to analyze the shape and characteristics of the hand.

A portable hand image capturing instrument was designed and manufactured, and the hand images and dimensions of 328 young men in Zhejiang area were obtained. The outer contour curve of the hand and the key points of finger root, fingertip, wrist and knuckle position were extracted. Then, the size of each hand part was calculated. The hand data obtained from the two-dimensional image was compared with the manual measurement data. Finally, the hands were classified according to the measurement data, and the relationship between hand control size and hand length, hand width and the relationship between hand length and height were explored.

The data comparison results show that the two measurement methods have high data consistency and are replaceable. In addition, analyzing the data obtained four major characteristic factors that affect the shape of the hand, divided the hands of young men in Zhejiang into five categories, and obtained the regression equations of basic hand size, hand length and hand width, and obtained the regression equation of hand length and height.

The method proposed in this study to obtain hand size based on the image and mark watershed algorithm has lower requirements on the external environment and testers, conforms to the development trend of applying artificial intelligence to anthropometric engineering and provides a useful reference value for data collection of gloves specification design. In addition, the results of data analysis can provide a valuable reference basis for consumer hand shape predictions, which can be used to guide the research and production of hand instruments, the design of specifications series and the purchase of hand products.

Active contour can describe target's silhouette accurately and has been widely used in image segmentation and target tracking. Its main drawback is huge computation that is still not well resolved. The purpose of this paper is to optimize the evolving path of active contour, to reduce the computation cost and to make the evolution effectively.

The contour‐evolution process is separated into two steps: global translation and local deformation. The contour global translation and local deformation are realized by average and normal gradient flow of the evolving contour curve, respectively.

When a contour is far away from the object to be segmented or tracked, the effective way of contour evolution is that it moves to the object without deformation first and then it deforms into the shape of the object when it moves on the object.

The method presented in this paper can optimize the curve evolving path effectively without complicated calculation, such as rebuilding a new inner product, and its computation cost is largely reduced.

The purpose of this paper is to explore the role that computer vision can play within new industrial paradigms such as Industry 4.0 and in particular to support production line improvements to achieve flexible manufacturing. As Industry 4.0 requires “big data”, it is accepted that computer vision could be one of the tools for its capture and efficient analysis. RGB-D data gathered from real-time machine vision systems such as Kinect ® can be processed using computer vision techniques.

This research exploits RGB-D cameras such as Kinect® to investigate the feasibility of using computer vision techniques to track the progress of a manual assembly task on a production line. Several techniques to track the progress of a manual assembly task are presented. The use of CAD model files to track the manufacturing tasks is also outlined.

This research has found that RGB-D cameras can be suitable for object recognition within an industrial environment if a number of constraints are considered or different devices/techniques combined. Furthermore, through the use of a HMM inspired state-based workflow, the algorithm presented in this paper is computationally tractable.

Processing of data from robust and cheap real-time machine vision systems could bring increased understanding of production line features. In addition, new techniques that enable the progress tracking of manual assembly sequences may be defined through the further analysis of such visual data. The approaches explored within this paper make a contribution to the utilisation of visual information “big data” sets for more efficient and automated production.

Accurate and rapid tracking and counting of building materials are crucial in managing on-site construction processes and evaluating their progress. Such processes are typically conducted by visual inspection, making them time-consuming and error prone. This paper aims to propose a video-based deep-learning approach to the automated detection and counting of building materials.

A framework for accurately counting building materials at indoor construction sites with low light levels was developed using state-of-the-art deep learning methods. An existing object-detection model, the You Only Look Once version 4 (YOLO v4) algorithm, was adapted to achieve rapid convergence and accurate detection of materials and site operatives. Then, DenseNet was deployed to recognise these objects. Finally, a material-counting module based on morphology operations and the Hough transform was applied to automatically count stacks of building materials.

The proposed approach was tested by counting site operatives and stacks of elevated floor tiles in video footage from a real indoor construction site. The proposed YOLO v4 object-detection system provided higher average accuracy within a shorter time than the traditional YOLO v4 approach.

The proposed framework makes it feasible to separately monitor stockpiled, installed and waste materials in low-light construction environments. The improved YOLO v4 detection method is superior to the current YOLO v4 approach and advances the existing object detection algorithm. This framework can potentially reduce the time required to track construction progress and count materials, thereby increasing the efficiency of work-in-progress evaluation. It also exhibits great potential for developing a more reliable system for monitoring construction materials and activities.

Presents a robot control system dedicated for grinding and deburring robots. The control system is based on an active force feedback system using three axes force sensor attached to the robot’s end effector. This system offers new functionality in rapid programming of the robot by applying automatic programming and force supervision. The system is implemented and tested experimentally on a MultiCraft 560 robot with parallel kinematics. The experimental results show a significant reduction in the programming and set up time compared to conventional robot control systems.

The purpose of this paper is to describe a work that aims to solve contour detection problem using a planar deformable model and a swarm‐based optimization technique. Contour detection is an important task in image processing as it allows depicting boundaries of objects in an image. The proposed approach uses snakes as active contour model and adapts predator prey optimization (PPO) metaheuristic so that to define a new dynamic for evolving snakes in a way to reduce time complexity while providing good quality results.

In the proposed approach, contour detection has been cast as an optimization problem requiring function minimization. PPO has been used to develop a search strategy to handle the optimization process. PPO is a population‐based method inspired by the phenomenon of predators attack and preys evasion. It has been proposed as an improvement of particle swarm optimization (PSO) where additional particles are introduced to repel the other particles into the swarm. The introduced dynamic is intended to achieve better exploration of the search space. In the design, a representation scheme has been first defined. Each particle either a predator or a prey is represented as a curve (snake) defined by a set of control points. The idea is then to evolve a set of curves using the dynamic governed by PPO model equations. As a result, the curve that optimizes a defined energy function is identified as the contour of the target object.

Application of the proposed method to a variety of images using a multi agent platform has shown that good quality results have been obtained compared to a PSO‐based method.

Nature inspired computing is an emergent paradigm that witnesses a growing interest because it suggests a new philosophy to optimization. This work contributes in showing its suitability to solve problems even it is still at infancy. In another hand, despite the amount of work done in image processing, it is still required to define new methods for image segmentation. This work outlines a new way to deal with this problem through the use of PPO.

In response to these shortcomings, this paper proposes a dynamic obstacle detection and tracking method based on multi-feature fusion and a dynamic obstacle recognition method based on spatio-temporal feature vectors.

The existing dynamic obstacle detection and tracking methods based on geometric features have a high false detection rate. The recognition methods based on the geometric features and motion status of dynamic obstacles are greatly affected by distance and scanning angle, and cannot meet the requirements of real traffic scene applications.

First, based on the geometric features of dynamic obstacles, the obstacles are considered The echo pulse width feature is used to improve the accuracy of obstacle detection and tracking; second, the space-time feature vector is constructed based on the time dimension and space dimension information of the obstacle, and then the support vector machine method is used to realize the recognition of dynamic obstacles to improve the obstacle The accuracy of object recognition. Finally, the accuracy and effectiveness of the proposed method are verified by real vehicle tests.

The paper proposes a dynamic obstacle detection and tracking method based on multi-feature fusion and a dynamic obstacle recognition method based on spatio-temporal feature vectors. The accuracy and effectiveness of the proposed method are verified by real vehicle tests.