Image Computer Vision Research Articles

Quality evaluation of Mono & bi-Colored Apples with computer vision and multispectral imaging

In the field of agriculture science, the presence of disease in fruits affects the quantity and quality of production. To sort the fruits based on quality is a challenging task. Human grades the fruit but this process is inconsistent, stagnant, expensive and get influenced by the surrounding. Thus an effective system for grading fruit is desired. In this paper, an automated fruit grading system has been developed for mono and bi-colored apples. An automated fruit grading system involves three steps, namely, segmentation, feature extraction, and classification. In this work, segmentation of defected area has been carried out using fuzzy c-means and for feature extraction, the various combination of Statistical/ Textural, Geometrical, Gabor Wavelet and, Discrete Cosine Transform feature have been considered. For classification three different classifiers i.e. k-NN (k- Nearest Neighbor)), SRC (Sparse Representation Classifier), SVM (Support Vector Machine) have been applied. The proposed system has been validated for four different databases of apples one having 1120 samples of which 984 were defective, the second having 333 samples of which 247 were defective, the third having 100 samples with 26 defective and the fourth with 56 defectives. The maximum accuracy of 95.21%, 93.41%, 92.64% and 87.91% for four datasets respectively, achieved by the system is encouraging and is comparable with the state of art techniques. The system performance has been validated using the k-fold cross-validation technique by considering k = 10. Results showed that a combination of features provides improved performance and the SVM classifier has the highest performance among k-NN, SRC. As compared to the state of art, our proposed solution yields better accuracy. Hence, the proposed algorithm showed great potential for the classification of apple and the possibility of its uses for further different fruit.

Utilization of the methods of artificial intelligence for organization of unmanned traffic

This article meticulously examines the questions of application of certain technologies of multi-agent systems theory in the area of unmanned traffic management for combatting the so-called “generative adversarial attacks” on the computer vision systems that are used in such vehicles. The article provides examples of generative-adversarial attacks on various types of neural networks, as well as describes the problems that arise when using computer vision. Possible solutions to these problems are proposed. Research methodology includes the theory of multi-agent systems applicable to automobile transport, which suggests using the so-called V2X-interaction, i.e. constant exchange of information between the vehicle and various actors involved in road traffic – a central control system, other vehicles, roadside infrastructure and pedestrians. The authors’ special contribution to this research lies in application of the theory of multi-agent systems for traffic arrangement with consideration of its actors as the agents with diverse roles. The novelty consists in employment of one of the methods of artificial intelligence in solution of the problems, obtained due to the use of other methods of artificial intelligence (recognition of images in computer vision). The relevance of the study is based on the detailed coverage of the questions of organization of unmanned traffic on training grounds and public roads.

Automatic Vehicle License Plate Recognition Using Optimal K-Means With Convolutional Neural Network for Intelligent Transportation Systems

Due to recent developments in highway research and increased utilization of vehicles, there has been significant interest paid on latest, effective, and precise Intelligent Transportation System (ITS). The process of identifying particular objects in an image plays a crucial part in the fields of computer vision or digital image processing. Vehicle License Plate Recognition (VLPR) process is a challenging process because of variations in viewpoint, shape, color, multiple formats and non-uniform illumination conditions at the time of image acquisition. This paper presents an effective deep learning-based VLPR model using optimal K-means (OKM) clustering-based segmentation and Convolutional Neural Network (CNN) based recognition called OKM-CNN model. The proposed OKM-CNN model operates on three main stages namely License Plate (LP) detection, segmentation using OKM clustering technique and license plate number recognition using CNN model. During first stage, LP localization and detection process take place using Improved Bernsen Algorithm (IBA) and Connected Component Analysis (CCA) models. Then, OKM clustering with Krill Herd (KH) algorithm get executed to segment the LP image. Finally, the characters in LP get recognized with the help of CNN model. An extensive experimental investigation was conducted using three datasets namely Stanford Cars, FZU Cars and HumAIn 2019 Challenge dataset. The attained simulation outcome ensured effective performance of the OKM-CNN model over other compared methods in a considerable way.

3D真的来了吗？— 三维结构光传感器漫谈

Three-dimensional (3D) imaging and sensing technologies, as valuable information acquisition tools for perceiving the real 3D world, provide data bases for the reconstruction of the geometric shape of objects and subsequent 3D modeling, detection, and recognition. Recently the development of computer vision and optoelectronic imaging technology, as well as the growing demand for 3D technologies in consumer electronics and personal authentication, have promoted the thriving growth of 3D imaging and sensing technologies. After the imaging revolution from monochrome to color, low resolution to high resolution, and static image to dynamic video, the transition of the camera from 2D to 3D will become the new fourth imaging revolution. This issue of Infrared and Laser Engineering organizes a special topic on Optical 3D Imaging and Sensing, which contains 20 high-quality articles, including 15 review papers and 5 research papers. These papers systematically introduce the research progress or trends of the cutting-edge research topics in the field of optical 3D imaging and sensing, and their themes comprehensively cover the current hot research directions in the field of 3D optical imaging: structured-light 3D imaging, fringe projection profilometry, interferometry, phase measuring deflectometry, 3D display technologies (such as holographic display, and integral/light field display), and the interdisciplinary fields of 3D sensing technologies and computational imaging technologies (such as 3D ghost imaging). As the preface of this issue, this paper summarizes the typical 3D sensing technologies and focuses on the current status, key technologies, and typical applications of the 3D structured-light sensor technologies, discusses its existing challenges, and looks forward to its future development directions.

Utilizing Knowledge Distillation in Deep Learning for Classification of Chest X-Ray Abnormalities

Automatic screening and diagnosis of lung abnormalities from chest X-ray images has been recently drawing attention from the computer vision and medical imaging communities. Previous studies of deep neural networks have predominantly demonstrated the effectiveness of lung disease binary classification procedures. However, large numbers of medical images-which can be labeled with a variety of existing or suspected pathologies-are required to be interpreted and reported upon daily by an individual radiologist; this poses a challenge in maintaining a consistently high diagnosis accuracy. In this paper, we present a competitive study of knowledge distillation (KD) in deep learning for classification of abnormalities in chest X-ray images. This method aims to either distill knowledge from cumbersome teacher models into lightweight student models or to self-train these student models, to generate weakly supervised multi-label lung disease classifications. Our approach was based on multi-task deep learning architectures that, in addition to multi-class classification, supported the visualizations utilized in saliency maps of the pathological regions where an abnormality was located. A self-training KD framework, in which the model learned from itself, was shown to outperform both the well-established baseline training procedure and the normal KD, achieving the AUC improvements of up to 6.39% and 3.89%, respectively. Through application to the publicly available ChestX-ray14 dataset, we demonstrated that our approach efficiently overcame the interdependency of 14 weakly annotated thorax diseases and facilitated the state-of-the-art classification compared with the current deep learning baselines.

Generators of Maximal Subgroups of Harada-Norton and some Linear Groups

AbstractGroup theory, the ultimate theory for symmetry, is a powerful tool that has a direct impact on research in robotics, computer vision, computer graphics and medical image analysis. Symmetry is very important in chemistry research and group theory is the tool that is used to determine symmetry. Usually, it is not only the symmetry of molecule but also the symmetries of some local atoms, molecular orbitals, rotations and vibrations of bonds, etc. that are important. Harada-Norton group is an example of a sporadic simple group. There are 14 maximal subgroups of Harada-Norton group. Generators (also known as words) of 11 maximal subgroups are already known. The aim of this note is to give generators of the remaining 3 maximal subgroups, which is an open problem mentioned on A World-wide-web Atlas of Group Representations (http://brauer.maths.qmul.ac.uk/Atlas) [1]. In this report we compute the generators of A6 × A6.D8, 23+2+6.(3 × L3(2)) and 34 : 2.(A4 × A4).4. Moreover we also compute the generators for the Maximal subgroups of some linear groups.

Toward New Retail: A Benchmark Dataset for Smart Unmanned Vending Machines

Deep learning is a popular direction in computer vision and digital image processing. It is widely utilized in many fields, such as robot navigation, intelligent video surveillance, industrial inspection, and aerospace. With the extensive use of deep learning techniques, classification and object detection algorithms have been rapidly developed. In recent years, with the introduction of the concept of “unmanned retail,” object detection, and image classification play a central role in unmanned retail applications. However, open-source datasets of traditional classification and object detection have not yet been optimized for application scenarios of unmanned retail. Currently, classification and object detection datasets do not exist that focus on unmanned retail solely. Therefore, in order to promote unmanned retail applications by using deep learning-based classification and object detection, in this article we collected more than 30 000 images of unmanned retail containers using a refrigerator affixed with different cameras under both static and dynamic recognition environments. These images were categorized into ten kinds of beverages. After manual labeling, images in our constructed dataset contained 155 153 instances, each of which was annotated with a bounding box. We performed extensive experiments on this dataset using ten state-of-the-art deep learning-based models. Experimental results indicate great potential of using these deep learning-based models for real-world smart unmanned vending machines.

Temporal phase unwrapping using deep learning

The multi-frequency temporal phase unwrapping (MF-TPU) method, as a classical phase unwrapping algorithm for fringe projection techniques, has the ability to eliminate the phase ambiguities even while measuring spatially isolated scenes or the objects with discontinuous surfaces. For the simplest and most efficient case in MF-TPU, two groups of phase-shifting fringe patterns with different frequencies are used: the high-frequency one is applied for 3D reconstruction of the tested object and the unit-frequency one is used to assist phase unwrapping for the wrapped phase with high frequency. The final measurement precision or sensitivity is determined by the number of fringes used within the high-frequency pattern, under the precondition that its absolute phase can be successfully recovered without any fringe order errors. However, due to the non-negligible noises and other error sources in actual measurement, the frequency of the high-frequency fringes is generally restricted to about 16, resulting in limited measurement accuracy. On the other hand, using additional intermediate sets of fringe patterns can unwrap the phase with higher frequency, but at the expense of a prolonged pattern sequence. With recent developments and advancements of machine learning for computer vision and computational imaging, it can be demonstrated in this work that deep learning techniques can automatically realize TPU through supervised learning, as called deep learning-based temporal phase unwrapping (DL-TPU), which can substantially improve the unwrapping reliability compared with MF-TPU even under different types of error sources, e.g., intensity noise, low fringe modulation, projector nonlinearity, and motion artifacts. Furthermore, as far as we know, our method was demonstrated experimentally that the high-frequency phase with 64 periods can be directly and reliably unwrapped from one unit-frequency phase using DL-TPU. These results highlight that challenging issues in optical metrology can be potentially overcome through machine learning, opening new avenues to design powerful and extremely accurate high-speed 3D imaging systems ubiquitous in nowadays science, industry, and multimedia.

Generic Isolated Cell Image Generator.

Building automated cancer screening systems based on image analysis is currently a hot topic in computer vision and medical imaging community. One of the biggest challenges of such systems, especially those using state‐of‐the‐art deep learning techniques, is that they usually require a large amount of training data to be accurate. However, in the medical field, the confidentiality of the data and the need for medical expertise to label them significantly reduce the amount of training data available. A common practice to overcome this problem is to apply data set augmentation techniques to artificially increase the size of the training data set. Classical data set augmentation methods such as geometrical or color transformations are efficient but still produce a limited amount of new data. Hence, there has been interest in data set augmentation methods using generative models able to synthesize a wider variety of new data. VitaDX is actually developing an automated bladder cancer screening system based on the analysis of cell images contained in urinary cytology digital slides. Currently, the number of available labeled cell images is limited and therefore exploitation of the full potential of deep learning techniques is not possible. In an attempt to increase the number of labeled cell images, a new generic generator for 2D cell images has been developed and is described in this article. This framework combines previous works on cell image generation and a recent style transfer method referred to as doodle‐style transfer in this article. To the best of our knowledge, we are the first to use a doodle‐style transfer method for synthetic cell image generation. This framework is quite modular and could be applied to other cell image generation problems. A statistical evaluation has shown that features of real and synthetic cell images followed roughly the same distribution. Finally, the realism of the synthetic cell images has been assessed through a visual evaluation performed with the help of medical experts. © 2019 The Authors. Cytometry Part A published by Wiley Periodicals, Inc. on behalf of International Society for Advancement of Cytometry.

An image vision and automatic calibration system for universal robots

The rapid development of technology is causing the replacement of many traditional manufacture industries by automation. Robotic arms are now commonly used in many sectors. The requirements for robotic arms in different sectors are quite different; but, in general, robotic arms not only save cost and man power, they also improve safety. The aim of this study was an investigation of the integration of image identification with robotic arms. To do this, the Denavit–Hartenberg transformation matrix was used to analyze the mechanical kinematics of the joints of each robotic arm axis. This allowed the spatial relationships between the Cartesian 3 D coordinate system and the joint of each axis to be determined and communication between the robotic arm and image identification to be established. A robotic arm prototype platform with automatic image identification and calibration was constructed using a quick and robust method. Several of the variables that exist in real robotic arm applications have been solved in this study: primarily the accuracy error that occurs when repeated gripping of workpieces is done and a movement and placement track is set up. Deviations occur frequently and if image identification can be applied to offset repeated inaccuracy, the quality of finished products will be more consistent. The Universal Robots 5 six-axis robot and cameras, which use a filter, Sobel computation, and Hough transformation computer vision image processing technology, together comprise a system that can automatically identify the workpiece and carry out loading and unloading accurately.

Image Segmentation Using Multilevel Thresholding: A Research Review

Image segmentation is a basic problem in computer vision and various image processing applications. Over the years, commonly used image segmentation has become quite challenging because of its utilization in many applications. Image thresholding is one of the most exploited techniques to accomplish image segmentation. Multilevel thresholding is found to be most appropriate and well known among all the image segmentation techniques. The segmented image quality is based on the techniques incorporated to choose the threshold value. In this paper, an exhaustive survey has been carried out, considering both the general purpose and satellite images to cover the performance comparison of various image segmentation approaches based on meta-heuristics optimization algorithms, present in the literature for multilevel image thresholding. In addition, this paper also focuses on information theoretic approach-based objective criterion using different statistical properties such as between-class variance, entropy, moment and maximum likelihood for selecting multilevel thresholds. A list of 157 publications on the subject is also appended for quick reference.

Biologically-inspired image processing in computational retina models

Retinal Prosthesis (RP) is an approach to restore vision, using an implanted device to electrically stimulate the retina. A fundamental problem in RP is to translate the visual scene to retina neural spike patterns, mimicking the computations normally done by retina neural circuits. Towards the perspective of improved RP interventions, we propose a Computer Vision (CV) image preprocessing method based on Retinal Ganglion Cells functions and then use the method to reproduce retina output with a standard Generalized Integrate & Fire (GIF) neuron model. “Virtual Retina” simulation software is used to provide the stimulus-retina response data to train and test our model. We use a sequence of natural images as model input and show that models using the proposed CV image preprocessing outperform models using raw image intensity (interspike-interval distance 0.17 vs 0.27). This result is aligned with our hypothesis that raw image intensity is an improper image representation for Retinal Ganglion Cells response prediction.

Inception Modules Enhance Brain Tumor Segmentation.

Magnetic resonance images of brain tumors are routinely used in neuro-oncology clinics for diagnosis, treatment planning, and post-treatment tumor surveillance. Currently, physicians spend considerable time manually delineating different structures of the brain. Spatial and structural variations, as well as intensity inhomogeneity across images, make the problem of computer-assisted segmentation very challenging. We propose a new image segmentation framework for tumor delineation that benefits from two state-of-the-art machine learning architectures in computer vision, i.e., Inception modules and U-Net image segmentation architecture. Furthermore, our framework includes two learning regimes, i.e., learning to segment intra-tumoral structures (necrotic and non-enhancing tumor core, peritumoral edema, and enhancing tumor) or learning to segment glioma sub-regions (whole tumor, tumor core, and enhancing tumor). These learning regimes are incorporated into a newly proposed loss function which is based on the Dice similarity coefficient (DSC). In our experiments, we quantified the impact of introducing the Inception modules in the U-Net architecture, as well as, changing the objective function for the learning algorithm from segmenting the intra-tumoral structures to glioma sub-regions. We found that incorporating Inception modules significantly improved the segmentation performance (p < 0.001) for all glioma sub-regions. Moreover, in architectures with Inception modules, the models trained with the learning objective of segmenting the intra-tumoral structures outperformed the models trained with the objective of segmenting the glioma sub-regions for the whole tumor (p < 0.001). The improved performance is linked to multiscale features extracted by newly introduced Inception module and the modified loss function based on the DSC.

Image retrieval based on convolutional neural network and linear discriminate analysis

Image retrieval is a hot research topic in the field of computer vision image processing, and the user queries the image database for similar images and produces a list of recommendations. The paper firstly sets forth the research status of image retrieval, then the convolution neural network is briefly introduced. Due to the traditional image retrieval and recommendation system use manual extraction of image features is relatively cumbersome, and the retrieval accuracy is not high research status, the paper proposes an image retrieval method based on the improved convolutional neural network and linear discriminate analysis. Caltech256 and CIFAR-10 datasets were trained using the model in this paper, experimental, results show that the proposed method can effectively improve the performance of retrieval.

Predicting furan content in a fried dough system using image analysis

The aim of this paper is to test different models for predicting furan content in a dough system, based on partial least squares regression using colour images. Starch dough systems were fried at five temperatures between 150 and 190 °C and for 5, 7, 9, 11 and 13 min. The furan content was quantified using gas chromatography/mass spectrometry, while the corresponding images were simultaneously obtained and processed in order to extract 2914 features. Good furan content predictions were obtained using computer vision image chromatic features using correlation coefficient of prediction (Rp = 0.86). However, the best prediction correlation was obtained using the image textural features (Rp = 0.93), when the number of features was reduced to 10 by algorithms applications. These results suggest that furan content in fried dough systems can be predicted using features of computer vision images.

Low Rank and Sparse Decomposition for Image and Video Applications

The matrix decomposing into a sum of low-rank and sparse components has found extensive applications in many areas including video surveillance, computer vision, and medical imaging. In this paper, we propose a new algorithm for recovery of low rank and sparse components of a given matrix. We have also proved the convergence of the proposed algorithm. The simulation results with synthetic and real signals such as image and video signals indicate that the proposed algorithm has a better performance with lower run-time than the conventional methods.

Blind image deblurring with reinforced use of edges

Blind image deblurring tries to restore a blurred image to a clear image without the blurring kernel known in advance, which is widely required in applications such as computer vision and medical image processing. With regard to this, the key issues here are to accurately estimate the blurring kernel for deconvolution of a blurred image, and avoid the ringing artifacts in the restored image, which are both related to high-quality detection of edge information in the blurred image. Though much endeavor has been made, it is still difficult to extract edge information well in blurred images and lacks investigation how edge information causes ringing artifacts. In this paper, we make a study on this and develop novel measures to optimize edge extraction and determine suitable width and weights for the extracted edges for reinforcing their use in deblurring, by which image deblurring can be improved with ringing artifacts considerably suppressed. Experimental results demonstrate our improvements over the existing methods.

Spherical optimal transportation

Optimal mass transportation (OT) problem aims at finding the most economic way to transform one probability measure to the other, which plays a fundamental role in many fields, such as computer graphics, computer vision, machine learning, geometry processing and medical imaging. Most existing algorithms focus on searching the optimal transportation map in Euclidean space, based on Kantorovich theory or Brenier theory. This work introduces a novel theoretic framework and computational algorithm to compute the optimal transportation map on the sphere. Constructing with a variational principle approach, our spherical OT map is carried out by solving a convex energy minimization problem and building a spherical power diagram.In theory, we prove the existence and the uniqueness of the spherical optimal transportation map; in practice, we present an efficient algorithm using the variational framework and Newton’s method. Comparing to the existing approaches, this work is more rigorous, efficient, robust and intrinsic to the spherical geometry. It can be generalized to the hyperbolic geometry or to higher dimensions.Our experimental results on a variety of models demonstrate efficacy and efficiency of the proposed method. At the same time, our method generates diffeomorphic, area-preserving, and seamless spherical parameterization results.

Image Similarity Searching Use Multi Part Cutting And Grayscale Color Histogram

The use of information technology in everyday life continues to increase so rapidly. This is inseparable from the role of researchers, especially in the field of information technology. Information technology has become a necessity so that it is widely used in the fields of education, trade, livestock and even to the agricultural sector. One of the obstacles that is needed is to check all activities involving information systems, especially when there is data in the form of images. Problems that arise are usually needed by humans to check and sort items that have been carried out by humans. This is the background of this research to help reduce activities involving humans. The process of finding the similarity of images in computer vision can be used in several fields such as education, retail, and other fields. In the field of computer vision education can be utilized for the automatic absence process through face recognition, in terms of retailing, it can be used for sorting through object detection. The process of finding similarities between images that are queries and dataset images will be the subject of research, and the process of calculating similarities between queries and datasets will be discussed step by step. The method used in the search process is by calculating the shortest distance between query images and dataset. The steps taken are the extraction feature and then RGB to gray color conversion. The next stage is to cut the image into four parts which will then be calculated the distance of the ecludian. The final part will calculate the performance of the algorithm using the matrix confusion method, so that the test results are in the form of error rates, precision, and accuracy. The trial process uses 30 data using 1000 datasets. In the test results obtained information in the form of recall of 1, 0.66 accuracy and 0.66precision.Keywords: Image Similarity, Histogram Image Grayscale, Ecludiance Distance.

Target-free vision-based technique for vibration measurements of structures subjected to out-of-plane movements

Vibration measurements have been widely used for structural health monitoring (SHM). Usually, wired sensors are required to attach on the testing structure, which may be arduous, costly and sometimes impossible to install those sensors on the remote and inaccessible part of the structure to be monitored. To overcome the limitations of contact sensors based vibration measurement methods, computer vision and digital image processing based methods have been proposed recently to measure the dynamic displacement of structures. Real-life structure subjected to bi-directional dynamic forces is susceptible to significant out-of-plane movement. Measuring the vibrations of structures under the out-of-plane movements using target-free vision-based methods have not been well studied. This paper proposes a target-free vision-based approach to obtain the vibration displacement and acceleration of structures subjected to out-of-plane movements from minor level excitations. The proposed approach consists of the selection of a region of interest (ROI), key-feature detection and feature extraction, tracking and matching of the features along the entire video, while there is no artificial target attached on the structure. The accuracy of the proposed approach is verified by conducting a number of experimental tests on a reinforced concrete structural column subjected to bi-directional ground motions with peak ground accelerations (PGA) ranging from 0.01 g to 1.0 g. The results obtained by the proposed approach are compared with those measured by using the conventional accelerometer and laser displacement sensor (LDS). It is found that the proposed approach accurately measures the displacement and acceleration time histories of the tested structure. Modal identification is conducted using the measured vibration responses, and natural frequencies can be identified accurately. The results demonstrate that the proposed approach is reliable and accurate to measure the dynamic responses and perform the system modal identification for structural health monitoring.