Vision Techniques Research Articles

Deep Learning for Forest Species Identification Based on Macroscopic Images

The fast and accurate identification of forest species is critical to support their sustainable management, to combat illegal logging, and ultimately to conserve them. Traditionally, the anatomical identification of forest species is a manual process that requires a human expert with a high level of knowledge to observe and differentiate certain anatomical structures present in a wood sample (Wiedenhoeft (2011)). In recent years, deep learning techniques have drastically improved the state of the art in many areas such as speech recognition, visual object recognition, and image and music information retrieval, among others (LeCun et al. (2015)). In the context of the automatic identification of plants, these techniques have recently been applied with great success (Carranza-Rojas et al. (2017)) and even mobile apps such as Pl@ntNet have been developed to identify a species from images captured on-the-fly (Joly et al. (2014)). In contrast to conventional machine learning techniques, deep learning techniques extract and learn by themselves the relevant features from large datasets. One of the main limitations for the application of deep learning techniques to forest species identification is the lack of comprehensive datasets for the training and testing of convolutional neural network (CNN) models. For this work, we used a dataset developed at the Federal University of Parana (UFPR) in Curitiba, Brazil, that comprises 2939 images in JPG format without compression and a resolution of 3.264 x 2.448 pixels. It includes 41 different forest species of the Brazilian flora that were cataloged by the Laboratory of Wood Anatomy at UFPR (Paula Filho et al. (2014)). Due to the lack of comprehensive datasets world wide, this has become a benchmark dataset in previous research (Paula Filho et al. (2014), Hafemann et al. (2014)). In this work, we propose and demonstrate the power of deep CNNs to identify forest species based on macroscopic images. We use a pre-trained model which is built from the resnet50 model and uses weights pre-trained on ImageNet. We apply fine-tuning by first truncating the top layer (softmax layer) of the pre-trained network and replacing it with a new softmax layer. Then we train again the model with the dataset of macroscopic images of species of the Brazilian flora used in (Hafemann et al. (2014), Paula Filho et al. (2014)). Using the proposed model we achieve a top-1 98% accuracy which is better than the 95.77% reported in (Hafemann et al. (2014) )using the same data set. In addition, our result is slightly better than the reported in (Paula Filho et al. (2014)) of 97.77% which was obtained by combining several conventional techniques of computer vision.

Medianatures

The article outlines the concept of medianatures. The term is a neologism and in debt to Donna Haraway’s rather eloquent and important coinage naturecultures that already functioned to mark the constant co-becomings of supposedly separated spheres of nature and culture. Medianatures is a further elaboration that elaborates the tie between the earth materialities that are mobilized for technological infrastructures, visual technologies, applications and devices, and the onto- epistemological stance that then feeds back into understanding those planetary scale earth materialities in the first place: the techniques of vision, observation, calculation, and circulation that are part of the governance of the earth and its various localities.

Impact of commercial sensors in human computer interaction: a review

Nowadays, the communication gap between humans and computers might be reduced due to multimodal sensors available in the market. Therefore, it is important to know the specifications of these sensors and how they are being used in order to create human computer interfaces, which tackle complex tasks. The purpose of this paper is to review recent research regarding the up-to-date application areas of the following sensors: (1) Emotiv sensor, which identifies emotions, facial expressions, thoughts, and head movements from users through electroencephalography signals, (2) Leap motion controller, which recognizes hand and arm movements via vision techniques, (3) Myo armband, which identifies hand and arm movements using electromyography signals and inertial sensors, and (4) Oculus rift, which provides immersion into virtual reality to users. The application areas discussed in this manuscript go from assistive technology to virtual tours. Finally, a brief discussion regarding advantages and shortcomings of each sensor is presented.

Towards NIRS-based hand movement recognition.

This work reports on preliminary results about on hand movement recognition with Near InfraRed Spectroscopy (NIRS) and surface ElectroMyoGraphy (sEMG). Either basing on physical contact (touchscreens, data-gloves, etc.), vision techniques (Microsoft Kinect, Sony PlayStation Move, etc.), or other modalities, hand movement recognition is a pervasive function in today environment and it is at the base of many gaming, social, and medical applications. Albeit, in recent years, the use of muscle information extracted by sEMG has spread out from the medical applications to contaminate the consumer world, this technique still falls short when dealing with movements of the hand. We tested NIRS as a technique to get another point of view on the muscle phenomena and proved that, within a specific movements selection, NIRS can be used to recognize movements and return information regarding muscles at different depths. Furthermore, we propose here three different multimodal movement recognition approaches and compare their performances.

Orientation-ocularity maps: A technique for computer vision

Orientation-ocularity maps: A technique for computer vision

Motion-Adaptive Depth Superresolution

Multi-modal sensing is increasingly becoming important in a number of applications, providing new capabilities and processing challenges. In this paper, we explore the benefit of combining a low-resolution depth sensor with a high-resolution optical video sensor, in order to provide a high-resolution depth map of the scene. We propose a new formulation that is able to incorporate temporal information and exploit the motion of objects in the video to significantly improve the results over existing methods. In particular, our approach exploits the space-time redundancy in the depth and intensity using motion-adaptive low-rank regularization. We provide experiments to validate our approach and confirm that the quality of the estimated high-resolution depth is improved substantially. Our approach can be a first component in systems using vision techniques that rely on high-resolution depth information.

Construction of Visual Inspection Database for Catenary on High-speed Railways

With the rapid development of computer vision, techniques of machine vision and visual inspection have been applied into the inspection of catenary on high-speed railways. Visual inspection systems have been developed and super-high-resolution images are captured to check the status of catenary components. Automatic recognition of defects becomes very important since the number of images is too huge to be manually checked one by one. However, it is not easy for the development of recognition algorithms on catenary components. There are many types of defects to be checked on different kinds of catenary components, but the number of defect images is too small in real world. In this paper, a solution was proposed and implemented. An on-site data acquisition system was designed and developed, and different types of defects were manually made on different catenary components beforehand. Finally, a visual inspection database was successfully constructed, including plenty of different kinds of catenary components, different types of defects, in different inspection conditions. The visual inspection database will be of great use in the development and test of recognition algorithms for catenary.

A review on data-driven learning of a talking head model

Constructing a talking head model of a person allows generation of a novel talking face animation from an unseen audio sequence of the person. This has important applications such as building virtual avatars of people that can interact with real people in novel situations, model-based video compression, teleconferencing, human-computer interaction, computer graphics and video games. Traditionally, talking head models have been built by manual painstaking work. The advancement of computer vision and machine learning techniques, especially in the past decade, has made possible the automatic learning of a talking head model of a person from data. In this paper, we focus on this area of machine learning based data-driven facial animation and critically review the most common approaches, compare and contrast among them and identify promising research directions and prospects.

Obtaining Description for Simple Images using Surface Realization Techniques and Natural Language Processing

This paper aims at developing a simple mechanism to deduce corpora pertaining to an image through various computer vision and natural language processing techniques. The output of the vision detection is combined with the sentence formation approach to get the visual content in textual form. Vision detections are smoothed using a number of approaches to prune undesired combination of words that are semantically incorrect. Descriptions are generated based on syntactic trees and Markov Chains and compared for human likeness based on survey. The results of the survey indicate that the descriptions generated with the help of Markov Chains sound more human like. These generated descriptions can be indexed in lucene and image search can be made more efficient bridging the semantic gap.

3-D Imaging Systems for Agricultural Applications-A Review.

Efficiency increase of resources through automation of agriculture requires more information about the production process, as well as process and machinery status. Sensors are necessary for monitoring the status and condition of production by recognizing the surrounding structures such as objects, field structures, natural or artificial markers, and obstacles. Currently, three dimensional (3-D) sensors are economically affordable and technologically advanced to a great extent, so a breakthrough is already possible if enough research projects are commercialized. The aim of this review paper is to investigate the state-of-the-art of 3-D vision systems in agriculture, and the role and value that only 3-D data can have to provide information about environmental structures based on the recent progress in optical 3-D sensors. The structure of this research consists of an overview of the different optical 3-D vision techniques, based on the basic principles. Afterwards, their application in agriculture are reviewed. The main focus lays on vehicle navigation, and crop and animal husbandry. The depth dimension brought by 3-D sensors provides key information that greatly facilitates the implementation of automation and robotics in agriculture.

Vehicle Detection Based on Probability Hypothesis Density Filter.

In the past decade, the developments of vehicle detection have been significantly improved. By utilizing cameras, vehicles can be detected in the Regions of Interest (ROI) in complex environments. However, vision techniques often suffer from false positives and limited field of view. In this paper, a LiDAR based vehicle detection approach is proposed by using the Probability Hypothesis Density (PHD) filter. The proposed approach consists of two phases: the hypothesis generation phase to detect potential objects and the hypothesis verification phase to classify objects. The performance of the proposed approach is evaluated in complex scenarios, compared with the state-of-the-art.

Assessing the potential value of bridge monitoring systems

On-going developments in smart technologies such as wireless sensor networks, micro-electro-mechanical systems (MEMS), computer vision, fibre optics and advanced data interpretation techniques may revolutionise structural health monitoring (SHM). Dedicated SHM of bridge assets has the potential to produce valuable data-sets and provide owners and managers with information to aid with key questions such as: current performance, margins of safety, actual loading, stress history and risk of fatigue, extent of deterioration and residual life. However, the parameters measured and value of the data obtained will differ when viewed from the perspectives of different stakeholders such as asset owners, designers, contractors and researchers. In this paper the purposes of monitoring are reviewed. A methodology is proposed to facilitate formal discussions between the key stakeholders before any deployment is specified and to ensure that scarce resources are not wasted in the pursuit of data as opposed to information. This approach can be used to determine if there is a prima facie case for the specification of SHM on a project and assess the potential value of any information that may be obtained. The developed methodology has been trialled with five historical monitoring case studies on bridges with which the authors are familiar.

Visual Tracking in Unknown Environments Using Fuzzy Logic and Dead Reckoning

This paper addresses the problem of making a non-holonomic wheeled mobile robot (WMR) move to a target object using computer vision and obstacle-avoidance techniques. If a priori information about the obstacles is available, pre-planning the desired path can be a good candidate method. However, in so many cases, obstacles are dynamic. Therefore, our first challenge is to make the WMR move to a desired target while autonomously avoiding any obstacle along its path. The second challenge deals with visual-tracking loss; that is, when the target is lost from the camera scope, the robot should use Dead Reckoning (DR) to get back on its path towards the target. The Visual Tracking (VT) algorithm then takes the relay to reach the final destination, compensating for any errors due to DR by calculating the distance to the target when it is within the scope of the camera. The proposed system also uses two fuzzy-logic controllers; the first controller avoids objects while the second manages the path to the target. Different complex scenarios have been implemented, showing the validity of our multi-controller model.

비전 기반 실시간 원거리 2차원 바코드 검출 시스템

In this paper, we propose a real-time two-dimensional bar code detection system even at long range using a vision technique. We first perform short-range detection, and then long-range detection if the short-range detection is not successful. First, edge map generation, image binarization, and connect component labeling (CCL) are performed in order to select a region of interest (ROI). After interpolating the selected ROI using bilinear interpolation, a location symbol pattern is detected as the same as for short-range detection. Finally, the symbol pattern is arranged by applying inverse perspective transformation to localize bar codes. Experimental results demonstrate that the proposed system successfully detects bar codes at two or three times longer distance than existing ones even at indoor environment.

Further development of adaptable automated visual inspection—part I: concept and scheme

The issue of adaptability in the design of automated visual inspection (AVI) systems has been a major road block in the application of AVI technology to manufacturing industry. The objective of an adaptable AVI system is to reduce system tuning time while maintaining the quality of inspection. In the authors’ previous work (Int J Adv Manuf Technol 59:655–667, 2012), we have proposed a novel adaptable AVI scheme, with which an AVI system can be trained online to accommodate new patterns. As such, the system can be deployed to undertake new inspection tasks or adapt to new operation conditions without excessive offline training efforts. The research presented in this two-part article is our further development of the adaptable AVI scheme, namely A2VIS. In this research, we focus on elaborating the concept, designing the structure, implementing the functions, and evaluating the performance of A2VIS in a systemic perspective. In part I, the proposed A2VIS is presented as a generic scheme at the conceptual level, with respect to the system components and structure, operating procedure, and performance measures. The key element in the design of the A2VIS is an adaptable inspection model that integrates subject localization, feature extraction, and classification functions as trainable submodels. In part I of this paper, we developed a generic tool for an adaptable AVI system without specifying the use of particular machine vision and machine learning techniques. Interested readers may choose techniques for implementing the proposed scheme according to their own preferences. In part II of this article, we focus on the system implementation and evaluation of A2VIS using such techniques as template matching, principal component analysis (PCA), and support vector machine (SVM).

Fitting primitive shapes in point clouds: a practical approach to improve autonomous underwater grasp specification of unknown objects

This article presents research on the subject of autonomous underwater robot manipulation. Ongoing research in underwater robotics intends to increase the autonomy of intervention operations that require physical interaction in order to achieve social benefits in fields such as archaeology or biology that cannot afford the expenses of costly underwater operations using remote operated vehicles. Autonomous grasping is still a very challenging skill, especially in underwater environments, with highly unstructured scenarios, limited availability of sensors and adverse conditions that affect the robot perception and control systems. To tackle these issues, we propose the use of vision and segmentation techniques that aim to improve the specification of grasping operations on underwater primitive shaped objects. Several sources of stereo information are used to gather 3D information in order to obtain a model of the object. Using a RANSAC segmentation algorithm, the model parameters are estimated and a set of feasible grasps are computed. This approach is validated in both simulated and real underwater scenarios.

László Moholy-Nagy and Chicago’s War Industry: Photographic Pedagogy at the New Bauhaus

László Moholy-Nagy came to Chicago to head the New Bauhaus in 1937 after the Nazis closed the school in Dessau. By the time the USA became involved in the war, investors had forced the school to change its name from the New Bauhaus to the American School of Design – a change that resulted directly from US ambivalence toward foreign, and especially German, influence. Of course, the school’s name was not the only change brought by the war. A study of the wartime context of Chicago industry and politics, as well as the school’s unique photographic pedagogy based on the photogram and its application to visual war strategies, reveals a fascinating attempt to convert an interdisciplinary Bauhaus curriculum based on the melding of art and industry into an artistic contribution to Chicago’s war industry. This article investigates how the interdisciplinary pedagogy of the Bauhaus was influenced by its new context in the heart of US industrial markets, and how the war affected both the city and the course of the school, its faculty and students. The article discusses how photographic pedagogy was adapted from its modernist origins to address the most pressing question facing the art school – namely, how artists could productively contribute to wartime efforts. Pedagogy based on mastering light and shadow was directly applied to new theories of combat vision and camouflage techniques developed by the faculty alongside numerous military guest lecturers, while students and faculty used art as a source of catharsis, producing haunting wartime images of destruction.

IMM Filter Based Human Tracking Using a Distributed Wireless Sensor Network

This paper proposes a human tracking approach in a distributed wireless sensor network. Most of the efforts on human tracking focus on vision techniques. However, most vision-based approaches to moving object detection involve intensive real-time computations. In this paper, we present an algorithm for human tracking using low-cost range wireless sensor nodes which can contribute lower computational burden based on a distributed computing system, while the centralized computing system often makes some information from sensors delay. Because the human target often moves with high maneuvering, the proposed algorithm applies the interacting multiple model (IMM) filter techniques and a novel sensor node selection scheme developed considering both the tracking accuracy and the energy cost which is based on the tacking results of IMM filter at each time step. This paper also proposed a novel sensor management scheme which can manage the sensor node effectively during the sensor node selection and the tracking process. Simulations results show that the proposed approach can achieve superior tracking accuracy compared to the most recent human motion tracking scheme.

Practical Approach for Product Dimension Measurement Using Stereoscopic Vision

Vision techniques have become, since many years, a relevant means for mastering product quality parameters, as dimensions, especially in on line control and in a hard working environment. This development has been reached thanks to technologic progress in equipements, sensors and data processing power. Indeed, in dimensionnal control, industrial vision allows the detection of specific points of the object, to compute the sought dimension. This latter is, in a cartesian coordinates system, the two points Euclidean distance. However, depending of the application type and in spite of their performance, used cameras present, some constraints as parallaxe error. It is due to the natural shift, at the installation, of the cameras optical axis regarding target object; or to the dynamic state of the object. This error may affect significantly the measure precision. Stereoscopic vision is a solution that allows minimizing the above error and presenting the compromise between precision, installation space and costs. Two cameras are used to observe, each one, the whole object to be measured. In the proposed practical approach, the application is highly depending of a calibrating step, that consists in creating a data base composed of a maximum image points related to the target plan and their geometric positions in a considered coordinates system. When executed, the algorithm detects, thanks to the two cameras acquired information, pixels corresponding to object edges, find their correspondance in the created data base, computes their coordinates (xi,yi) in the defined coordinates system to, finally, computes the sought dimension. 

Application of Camera Image Processing to Control of Humanoid Robot Motion

The wide potential applications of humanoid robots require that the robots can move in general environment, overcome various obstacles, detect predefined objects and control of its motion according to all these parameters. The goal of this paper is address the problem of implementation of computer vision to motion control of humanoid robot. We focus on using of computer vision and image processing techniques, based on which the robot can detect and recognize a predefined color object in a captured image. An algorithm to detection and localization of objects is described. The results obtained from image processing are used in an algorithm for controlling of the robot movement.