Artificial Landmark based Pose-Graph SLAM for AGVs in Factory Environments

The paper addresses a new defect detection method that uses independent component analysis (ICA) for locating and also partially identifying defects in textile fabric images. The method is based on obtaining ICA basis vectors for a large number of textile images obtained from the same fabric. Feature vectors are computed for the test images and these are compared with the true feature vector predetermined from a defect-free image. The proposed method is shown to be invariant under rotation. Experimental results are also presented to demonstrate the use of this method for visual inspection of textile products obtained from a real factory environment.

Accurate and reliable localization is a base requirement for a mobile robot. Landmark localization is a very common technique to do this. The arrangement of landmarks is important for reliable and accurate estimation of localization data. In this paper the optimized arrangement of artificial landmarks is discussed and a solution for building an accurate and reliable localization system based on combining artificial and natural landmarks is presented. Landmark placement is done by definition and optimization of a confidence level describing the expected localization estimation error. Areas in the environment which are ill conditioned by natural landmarks are furnished with additional artificial landmarks. The proposed system has been tested in a simulation and with a mobile robot in a real office environment.

This paper introduces a concept for intuitive programming of an industrial HRI cell for non-experts. The main idea of this concept is to combine recently available technologies ranging from speech recognition, 3D visual surveillance (person tracking and object recognition) and handheld devices over to teach-in, visual programming and instruction based programming of compliant robots for heavy loads. With these components an easy method for programming or adapting workflows within an industrial packaging cell was realized and tested in a real factory environment.

Application of the Preprocessing System of Stretch Forming Over Multi-Point Tooling in Real Factory Environment

A YOLOv5-based object detection inventory algorithm was proposed to address the problem of low efficiency and high cost of inventorying the number of stacked compact double-sided laminates by manual means. The algorithm first collected images of laminates in a real factory environment and annotated them with the network; then constructed a double-sided copper clad laminate inspection dataset with a real application background and trained the dataset with the YOLOv5 object detection inventory algorithm to obtain a training model; finally, detected the number of laminates inventoried based on the validation set. The experimental data verified that the trained model could achieve 99% accuracy in detecting different numbers of sheets in stacked scenarios, which own more advantageous than manual inventory methods in terms of both detection accuracy and speed. The results show that the proposed YOLOv5-based object detection counting algorithm is reasonable, feasible and effective for counting stacked compact double-sided copper clad laminates.

This paper presents an implementation of a digital filtering inspection system applied on a paper pulp sheet production process. The automation of the inspection phase is particularly critical during this process and its solution is highly complex. The system is based on neural network learning, allowing a compromise between resolution and processing speed. The experimental results demonstrating the use of this algorithm for the visual detection of defects in images obtained from a real factory environment are presented. These results show that the developed learning method generates filters that fulfil the required inspection standard.KeywordsDigital FilterDefect TypeAutomatic GenerationMulti Layer PerceptronInspection SystemThese keywords were added by machine and not by the authors. This process is experimental and the keywords may be updated as the learning algorithm improves.

Global localization is one of the essential issues for mobile robot navigation. In this study, an indoor global localization method is proposed which uses a Kinect sensor and a monocular upward- looking camera. The proposed method generates an environment map which consists of a grid map, a ceiling feature map from the upward-looking camera, and a spatial feature map obtained from the Kinect sensor. The method selects robot pose candidates using the spatial feature map and updates sample poses by particle filter based on the grid map. Localization success is determined by calculating the matching error from the ceiling feature map. In various experiments, the proposed method achieved a position accuracy of 0.12m and a position update speed of 10.4s, which is robust enough for real- world applications.

In this paper, a new Haar wavelet-based approach to the detection and localization of defects in grey-level texture images is presented. This new approach explores space localization properties of the discrete Haar wavelet transform (HT) and generates statistically-based parameterized texture defect detection criteria. The criteria provide the user with a possibility to control the percentage of both the actually defect-free images detected as defective and/or the actually defective images detected as defect-free, in the class of texture images under investigation. The experiment analyses samples of ceramic tiles, glass samples, as well as fabric scraps, taken from real factory environment.

Intellectualization of manufacturing is a general trend due to the development of technology and science. This paper presents a preliminary research on intelligent mobile tool cart (IMTC) for working alongside workers in manufacturing process. Tool cart problem is considered as an exemplary intelligent mobile tool cart service system. A type of differential-driven mobile cart with a tool basket which is mounted on the top of the mobile cart compose the intelligent mobile tool cart. The IMTC is designed to provide workers with remote motion control and tools positioning services. Daily used instructions are exercised as commands to control the IMTC in the factory environment by using Leap Motion sensor for finger/hand position detection. The IMTC provides its real-time position information to the worker. To fulfill estimating the location of IMTC, QR localization method is proposed. We verified the features of the IMTC and feasibility of the proposed localization method through experimental trials.

The aim of this paper is to explore the possibility of using geo-referenced aerial images to augment an Unmanned Aerial Vehicle (UAV) navigation system in case of GPS failure. The proposed method is based on image matching between the current view from a monocular video camera and a previously known database of geo-referenced images. Only natural landmarks provided by a feature tracking algorithm will be considered, without the help of visual beacons or artificial landmarks. The novel idea is to perform global localization, position tracking, localization failure recovery and safe landing. Key points of current video frame and reference database are extracted by implementing Speeded-Up Robust Features (SURF) algorithm on a NVIDIA Geforce 240M graphics processing unit (GPU). The proposed system has been tested on real flight data and testing results confirmed that the proposed system is real-time and sufficiently robust to scale, rotation, illumination, noise and affine projection, thus it can be an ideal scheme for autonomous navigation system for UAVs under GPS blockage conditions.

Autonomous underwater vehicles (AUVs) can perform flexible operations in complex underwater environments due to their autonomy. Localization is one of the key components of autonomous navigation. Since the inertial navigation system of an AUV suffers from drift, observing fixed objects in an inertial reference system can enhance the localization performance. In this paper, we propose a method of localizing AUVs by exploiting visual measurements of underwater structures and artificial landmarks. In a framework of particle filtering, a camera measurement model that emulates the camera's observation of underwater structures is designed. The particle weight is then updated based on the extracted visual information of the underwater structures. Detected artificial landmarks are also used in the particle weight update. The proposed method is validated by experiments performed in a structured basin environment.

Deformable models have been intensively studied in image analysis through the last decade, and are used for detection and recognition of flexible or rigid templates under diverse viewing conditions. Genetic algorithm (GA) based deformable models are used for generic visual landmark detection and interpretation. The developed system allows topologic localization and navigation using natural and artificial landmarks, exploiting deformable models' ability for handling landmark perspective variations. The resulting perception module has been integrated successfully in a complex navigation system. Various experimental results in real environments are presented on this paper, showing the effectiveness and capacity of the landmark detection and reading system.

In manufacturing and logistics, various applications exploiting IoT devices are started to be used. Although there is a demand for wireless connection between the IoT devices to networks, obstacles such as radio frequency interference, multipath-rich propagation, and movement of objects make communication unstable. The instability can cause a system failure of IoT applications. Since the QoS outage does not occur frequently in actual systems, the training of QoS outage events may be insufficient to accurately predict future QoS values. Therefore, we propose a method to estimate QoS outage probability from the predicted future QoS, distribution of prediction error, and required QoS. The proposed method can achieve high estimation accuracy by obtaining and using different error distributions according to the rarity of the observed sequence. Simulation results based on a theoretical throughput analysis of wireless LAN show that the proposed method can appropriately predict the QoS outage probability.

In this paper, we present an omnidirectional artificial landmark model and a robust artificial landmark recognition algorithm for indoor mobile robot positioning. The landmark model encodes identities with nested circles in black and white, which provides stable edge response and enables strong tolerance to various lighting conditions and perspective distortions. The corresponding positioning system uses a single upward-facing webcam as the vision sensor to capture landmarks. To address the effect of the lighting and sensing noise, the topological contour analysis is applied to detect landmarks, and the dynamic illumination adjustment is used to assist landmark recognition. Based on the landmark recognition, the absolute position of the camera in the environment is estimated using a trilateration algorithm. The landmark model and positioning system are tested with a mobile robot in a real indoor environment. The results show that the purposed technique provides autonomous indoor positioning for mobile robots with high robustness and consistency.

This paper presents a preliminary evaluation of a framework that allows an overhead RGBD camera to segment and track workers skeleton in an unstructured factory environment. The default Kinect skeleton tracking algorithm was developed using front-view artificial depth images generated from a 3D model of a person in an room. The proposed framework is inspired in this concept, and works by capturing motion data of worker movements performing a real factory task. That motion data is matched to the 3D model of the worker. In a novel approach, the largest elements in the (e.g. desks, racks) are modeled with simple shapes, and the artificial depth images are generated in a workspace in contrast with an empty workspace. We show in preliminary experiments that the addition of the simplified models during training can increase, ceteris paribus, the segmentation accuracy by over 3 times and the recall by about one and a half times when the is highly cluttered. Evaluation is made using real depth images obtained in a factory environment, and as ground-truth manually segmented images are used.