PeerJ Computer Science Preprints: Computer Visionhttps://peerj.com/preprints/index.atom?journal=cs&subject=9300Computer Vision articles published in PeerJ Computer Science PreprintsMice tracking using the YOLO algorithmhttps://peerj.com/preprints/278802019-08-012019-08-01Helton Maia PeixotoRichardson Santiago TelesJohn Victor Alves LuizAron Miranda Henriques-AlvesRossana Moreno Santa Cruz
The development of computational tools is essential for the development of new technologies, including experimental designs needed for behavioral neuroscience research. The computational tool developed in this study is based on the convolutional neural networks and the You Only Look Once (YOLO) algorithm for detecting and tracking mice in videos recorded during behavioral neuroscience experiments. The task of mice detection consists of determining the location in the image where the animals are present, for each frame acquired. In this work, we propose mice tracking using the YOLO algorithm, running on an NVIDIA GeForce GTX 1060 GPU. We analyzed a set of data composed of 13622 images, made up of behavioral videos of three important researches in this area. The training set used 50% of the images, 25% for validation and 25% for the tests. The results show that the mean Average Precision (mAP) reached by the developed system was 90.79% and 90.75% for the Full and Tiny versions of YOLO, respectively. It has also been found that the use of the Tiny version is a good alternative for experimental designs that require real-time response. Considering the high accuracy of the results, the developed work allows the experimentalists to perform mice tracking in a reliable and non-evasive way, avoiding common system errors that require delimitations of regions of interest (ROI) or even evasive luminous identifiers such as LED for tracking the animals.
The development of computational tools is essential for the development of new technologies, including experimental designs needed for behavioral neuroscience research. The computational tool developed in this study is based on the convolutional neural networks and the You Only Look Once (YOLO) algorithm for detecting and tracking mice in videos recorded during behavioral neuroscience experiments. The task of mice detection consists of determining the location in the image where the animals are present, for each frame acquired. In this work, we propose mice tracking using the YOLO algorithm, running on an NVIDIA GeForce GTX 1060 GPU. We analyzed a set of data composed of 13622 images, made up of behavioral videos of three important researches in this area. The training set used 50% of the images, 25% for validation and 25% for the tests. The results show that the mean Average Precision (mAP) reached by the developed system was 90.79% and 90.75% for the Full and Tiny versions of YOLO, respectively. It has also been found that the use of the Tiny version is a good alternative for experimental designs that require real-time response. Considering the high accuracy of the results, the developed work allows the experimentalists to perform mice tracking in a reliable and non-evasive way, avoiding common system errors that require delimitations of regions of interest (ROI) or even evasive luminous identifiers such as LED for tracking the animals.Improving the resolution of microscope by deconvolution after dense scanhttps://peerj.com/preprints/278492019-07-292019-07-29Yaohua Xie
Super-resolution microscopes (such as STED) illuminate samples with a tiny spot, and achieve very high resolution. But structures smaller than the spot cannot be resolved in this way. Therefore, we propose a technique to solve this problem. It is termed “Deconvolution after Dense Scan (DDS)”. First, a preprocessing stage is introduced to eliminate the optical uncertainty of the peripheral areas around the sample’s ROI (Region of Interest). Then, the ROI is scanned densely together with its peripheral areas. Finally, the high resolution image is recovered by deconvolution. The proposed technique does not need to modify the apparatus much, and is mainly performed by algorithm. Simulation experiments show that the technique can further improve the resolution of super-resolution microscopes.
Super-resolution microscopes (such as STED) illuminate samples with a tiny spot, and achieve very high resolution. But structures smaller than the spot cannot be resolved in this way. Therefore, we propose a technique to solve this problem. It is termed “Deconvolution after Dense Scan (DDS)”. First, a preprocessing stage is introduced to eliminate the optical uncertainty of the peripheral areas around the sample’s ROI (Region of Interest). Then, the ROI is scanned densely together with its peripheral areas. Finally, the high resolution image is recovered by deconvolution. The proposed technique does not need to modify the apparatus much, and is mainly performed by algorithm. Simulation experiments show that the technique can further improve the resolution of super-resolution microscopes.Improving the quality of low SNR images using high SNR imageshttps://peerj.com/preprints/278002019-06-142019-06-14Yaohua Xie
It is important to get data with Signal-Noise-Ratios (SNR) as high as possible. Compared to other techniques, filtering methods are fast. But they do not make full use of the characteristics of sample structure which reflected by relevant high SNR images. In this study, we propose a technique termed “TransFiltering”. It transplants the characteristics of a high SNR image to the frequency spectrum of a low SNR image by filtering. Usually, the high SNR and the low SNR image should have similar structure pattern. For example, they all come from the same image sequence. In the proposed method, Fourier transform is first performed on both of the images. Then, the frequency spectrum of the low SNR image is filtered according to that of the high SNR image. Finally, inverse Fourier transform is performed to get the image with improved SNR. Experiment results show that the proposed method is both effective and efficient.
It is important to get data with Signal-Noise-Ratios (SNR) as high as possible. Compared to other techniques, filtering methods are fast. But they do not make full use of the characteristics of sample structure which reflected by relevant high SNR images. In this study, we propose a technique termed “TransFiltering”. It transplants the characteristics of a high SNR image to the frequency spectrum of a low SNR image by filtering. Usually, the high SNR and the low SNR image should have similar structure pattern. For example, they all come from the same image sequence. In the proposed method, Fourier transform is first performed on both of the images. Then, the frequency spectrum of the low SNR image is filtered according to that of the high SNR image. Finally, inverse Fourier transform is performed to get the image with improved SNR. Experiment results show that the proposed method is both effective and efficient.A new algorithm for band detection and pattern extraction on pulsed-field gel electrophoresis imageshttps://peerj.com/preprints/277712019-06-022019-06-02Mohammad RezaeiNaser ZohorianNemat SoltaniParviz Mohajeri
This paper presents a new approach for band detection and pattern recognition for molecule types. Although a few studies have examined band detection, but there is still no automatic method that can perform well despite the high noise. The band detection algorithm was designed in two parts, including band location and lane pattern recognition. In order to improve band detection and remove undesirable bands, the shape and light intensity of the bands were used as features. One-hundred lane images were selected for the training stage and 350 lane images for the testing stage to evaluate the proposed algorithm in a random fashion. All the images were prepared using PFGE BIORAD at the Microbiology Laboratory of Kermanshah University of Medical Sciences. An adaptive median filter with a filter size of 5x5 was selected as the optimal filter for removing noise. The results showed that the proposed algorithm has a 98.45% accuracy and is associated with less errors compared to other methods. The proposed algorithm has a good accuracy for band detection in pulsed-field gel electrophoresis images. Considering the shape of the peaks caused by the bands in the vertical projection profile of the signal, this method can reduce band detection errors. To improve accuracy, we recommend that the designed algorithm be examined for other types of molecules as well.
This paper presents a new approach for band detection and pattern recognition for molecule types. Although a few studies have examined band detection, but there is still no automatic method that can perform well despite the high noise. The band detection algorithm was designed in two parts, including band location and lane pattern recognition. In order to improve band detection and remove undesirable bands, the shape and light intensity of the bands were used as features. One-hundred lane images were selected for the training stage and 350 lane images for the testing stage to evaluate the proposed algorithm in a random fashion. All the images were prepared using PFGE BIORAD at the Microbiology Laboratory of Kermanshah University of Medical Sciences. An adaptive median filter with a filter size of 5x5 was selected as the optimal filter for removing noise. The results showed that the proposed algorithm has a 98.45% accuracy and is associated with less errors compared to other methods. The proposed algorithm has a good accuracy for band detection in pulsed-field gel electrophoresis images. Considering the shape of the peaks caused by the bands in the vertical projection profile of the signal, this method can reduce band detection errors. To improve accuracy, we recommend that the designed algorithm be examined for other types of molecules as well.Extracting super-resolution details directly from a diffraction-blurred image or part of its frequency spectrumhttps://peerj.com/preprints/275912019-04-282019-04-28Edward Y Sheffield
It is usually believed that the low frequency part of a signal’s Fourier spectrum represents its profile, while the high frequency part represents its details. Conventional light microscopes filter out the high frequency parts of image signals, so that people cannot see the details of the samples (objects being imaged) in the blurred images. However, we find that in a certain “resolvable condition”, a signal’s low frequency and high frequency parts not only represent profile and details respectively. Actually, any one of them also contains the full information (including both profile and details) of the sample’s structure. Therefore, for samples with spatial frequency beyond diffraction-limit, even if the image’s high frequency part is filtered out by the microscope, it is still possible to extract the full information from the low frequency part. On the basis of the above findings, we propose the technique of Deconvolution Super-resolution (DeSu-re), including two methods. One method extracts the full information of the sample’s structure directly from the diffraction-blurred image, while the other extracts it directly from part of the observed image’s spectrum (e.g., low frequency part). Both theoretical analysis and simulation experiment support the above findings, and also verify the effectiveness of the proposed methods.
It is usually believed that the low frequency part of a signal’s Fourier spectrum represents its profile, while the high frequency part represents its details. Conventional light microscopes filter out the high frequency parts of image signals, so that people cannot see the details of the samples (objects being imaged) in the blurred images. However, we find that in a certain “resolvable condition”, a signal’s low frequency and high frequency parts not only represent profile and details respectively. Actually, any one of them also contains the full information (including both profile and details) of the sample’s structure. Therefore, for samples with spatial frequency beyond diffraction-limit, even if the image’s high frequency part is filtered out by the microscope, it is still possible to extract the full information from the low frequency part. On the basis of the above findings, we propose the technique of Deconvolution Super-resolution (DeSu-re), including two methods. One method extracts the full information of the sample’s structure directly from the diffraction-blurred image, while the other extracts it directly from part of the observed image’s spectrum (e.g., low frequency part). Both theoretical analysis and simulation experiment support the above findings, and also verify the effectiveness of the proposed methods.Event-driven industrial robot control architecture for the Adept V+ platformhttps://peerj.com/preprints/275522019-02-272019-02-27Oleksandr SemeniutaPetter Falkman
Modern industrial robotic systems are highly interconnected. They operate in a distributed environment and communicate with sensors, computer vision systems, mechatronic devices, and computational components. On the fundamental level, communication and coordination between all parties in such distributed system are characterized by discrete event behavior. The latter is largely attributed to the specifics of communication over the network, which, in terms, facilitates asynchronous programming and explicit event handling. In addition, on the conceptual level, events are an important building block for realizing reactivity and coordination. Event-driven architecture has manifested its effectiveness for building loosely-coupled systems based on publish-subscribe middleware, either general-purpose or robotic-oriented. Despite all the advances in middleware, industrial robots remain difficult to program in context of distributed systems, to a large extent due to the limitation of the native robot platforms. This paper proposes an architecture for flexible event-based control of industrial robots based on the Adept V+ platform. The architecture is based on the robot controller providing a TCP/IP server and a collection of robot skills, and a high-level control module deployed to a dedicated computing device. The control module possesses bidirectional communication with the robot controller and publish/subscribe messaging with external systems. It is programmed in asynchronous style using pyadept, a Python library based on Python coroutines, AsyncIO event loop and ZeroMQ middleware. The proposed solution facilitates integration of Adept robots into distributed environments and building more flexible robotic solutions with event-based logic.
Modern industrial robotic systems are highly interconnected. They operate in a distributed environment and communicate with sensors, computer vision systems, mechatronic devices, and computational components. On the fundamental level, communication and coordination between all parties in such distributed system are characterized by discrete event behavior. The latter is largely attributed to the specifics of communication over the network, which, in terms, facilitates asynchronous programming and explicit event handling. In addition, on the conceptual level, events are an important building block for realizing reactivity and coordination. Event-driven architecture has manifested its effectiveness for building loosely-coupled systems based on publish-subscribe middleware, either general-purpose or robotic-oriented. Despite all the advances in middleware, industrial robots remain difficult to program in context of distributed systems, to a large extent due to the limitation of the native robot platforms. This paper proposes an architecture for flexible event-based control of industrial robots based on the Adept V+ platform. The architecture is based on the robot controller providing a TCP/IP server and a collection of robot skills, and a high-level control module deployed to a dedicated computing device. The control module possesses bidirectional communication with the robot controller and publish/subscribe messaging with external systems. It is programmed in asynchronous style using pyadept, a Python library based on Python coroutines, AsyncIO event loop and ZeroMQ middleware. The proposed solution facilitates integration of Adept robots into distributed environments and building more flexible robotic solutions with event-based logic.A fast iris recognition system through optimum feature extractionhttps://peerj.com/preprints/273632019-02-232019-02-23Humayan Kabir RanaMd. Shafiul AzamMst. Rashida AkhtarJulian M.W. QuinnMohammad Ali Moni
With an increasing demand for stringent security systems, automated identification of individuals based on biometric methods has been a major focus of research and development over the last decade. Biometric recognition analyses unique physiological traits or behavioral characteristics, such as an iris, face, retina, voice, fingerprint, hand geometry, keystrokes or gait. The iris has a complex and unique structure that remains stable over a person's lifetime, features that have led to its increasing interest in its use for biometric recognition.
In this study, we proposed a technique incorporating Principal Component Analysis (PCA) based on Discrete Wavelet Transformation (DWT) for the extraction of the optimum features of an iris and reducing the runtime needed for iris templates classification. The idea of using DWT behind PCA is to reduce the resolution of the iris template. DWT converts an iris image into four frequency sub-bands. One frequency sub-band instead of four has been used for further feature extraction by using PCA. Our experimental evaluation demonstrates the efficient performance of the proposed technique.
With an increasing demand for stringent security systems, automated identification of individuals based on biometric methods has been a major focus of research and development over the last decade. Biometric recognition analyses unique physiological traits or behavioral characteristics, such as an iris, face, retina, voice, fingerprint, hand geometry, keystrokes or gait. The iris has a complex and unique structure that remains stable over a person's lifetime, features that have led to its increasing interest in its use for biometric recognition.In this study, we proposed a technique incorporating Principal Component Analysis (PCA) based on Discrete Wavelet Transformation (DWT) for the extraction of the optimum features of an iris and reducing the runtime needed for iris templates classification. The idea of using DWT behind PCA is to reduce the resolution of the iris template. DWT converts an iris image into four frequency sub-bands. One frequency sub-band instead of four has been used for further feature extraction by using PCA. Our experimental evaluation demonstrates the efficient performance of the proposed technique.Implementation and validity of the long jump knowledge-based system: Case of the approach run phasehttps://peerj.com/preprints/275242019-02-082019-02-08Teerawat KamnardsiriWorawit JanchaiPattaraporn KhuwuthyakornWacharee Rittiwat
This study aimed to propose the method of implementation of the Knowledge-Based System (KBS) in the case of approach-run phase. The proposed method was implemented for improving the long jump performance of athletes in the approach-run phase. Moreover, this study aimed to examine KBS concurrent validity in distinguishing between professional and amateur populations and then KBS convergent validity against a Tracker video analysis tool. Seven running professionals aged 19 to 42 years and five amateurs aged 18 to 38 years had captured with ten conditions of different movements (C1 to C10) using a standard video camera (60 fps, 10 mm lens). The camera was fixed on the tripod. The results showing an age-related difference in a speed measurement of ten conditions were evidently using the KBS. Good associations were found between KBS and Tracker 4.94 video analysis tool across various conditions of three variables that were the starting position (r=0.926 and 0.963), the maximum velocity (r=0.972 and 0.995) and the location of maximum velocity (r=0.574 and 0.919). In conclusion, the proposed method is a reliable tool for measuring the starting position, maximum speed and position of maximum speed. Furthermore, the proposed method can also distinguish speed performance between professional and amateur across multiple movement conditions.
This study aimed to propose the method of implementation of the Knowledge-Based System (KBS) in the case of approach-run phase. The proposed method was implemented for improving the long jump performance of athletes in the approach-run phase. Moreover, this study aimed to examine KBS concurrent validity in distinguishing between professional and amateur populations and then KBS convergent validity against a Tracker video analysis tool. Seven running professionals aged 19 to 42 years and five amateurs aged 18 to 38 years had captured with ten conditions of different movements (C1 to C10) using a standard video camera (60 fps, 10 mm lens). The camera was fixed on the tripod. The results showing an age-related difference in a speed measurement of ten conditions were evidently using the KBS. Good associations were found between KBS and Tracker 4.94 video analysis tool across various conditions of three variables that were the starting position (r=0.926 and 0.963), the maximum velocity (r=0.972 and 0.995) and the location of maximum velocity (r=0.574 and 0.919). In conclusion, the proposed method is a reliable tool for measuring the starting position, maximum speed and position of maximum speed. Furthermore, the proposed method can also distinguish speed performance between professional and amateur across multiple movement conditions.An improved human skin detection and localization by using machine learning techniques in RGB and YCbCr color spaceshttps://peerj.com/preprints/274882019-01-152019-01-15Mohammad Mortazavi T.Omid Mahdi Ebadati E.
Human Skin Detection is one of the most applicable methods in human detection, face detection and so many other detections. These processes can be used in a wide spectrum like industry, medicine, security, etc. The objective of this work is to provide an accurate and efficient method to detect human skin in images. This method can detect and classify skin pixels and reduce the size of images. With the use of RGB and YCbCr color spaces, proposed approach can localize a Region Of Interest (ROI) that contains skin pixels. This method consists of three steps. In the first stage, pre-processing an image like normalization, detecting skin range from the dataset, etc. is done. In the second stage, the proposed method detects candidate’s pixels that are in the range of skin color. In the third stage, with the use of a classifier, it decreases unwanted pixels and areas to decrease the accuracy of the region. The results show 97% sensitivity and 85% specificity for support vector machine classifier.
Human Skin Detection is one of the most applicable methods in human detection, face detection and so many other detections. These processes can be used in a wide spectrum like industry, medicine, security, etc. The objective of this work is to provide an accurate and efficient method to detect human skin in images. This method can detect and classify skin pixels and reduce the size of images. With the use of RGB and YCbCr color spaces, proposed approach can localize a Region Of Interest (ROI) that contains skin pixels. This method consists of three steps. In the first stage, pre-processing an image like normalization, detecting skin range from the dataset, etc. is done. In the second stage, the proposed method detects candidate’s pixels that are in the range of skin color. In the third stage, with the use of a classifier, it decreases unwanted pixels and areas to decrease the accuracy of the region. The results show 97% sensitivity and 85% specificity for support vector machine classifier.RobOMP: Robust variants of Orthogonal Matching Pursuit for sparse representationshttps://peerj.com/preprints/274822019-01-142019-01-14Carlos A Loza
Sparse coding aims to find a parsimonious representation of an example given an observation matrix or dictionary. In this regard, Orthogonal Matching Pursuit (OMP) provides an intuitive, simple and fast approximation of the optimal solution. However, its main building block is anchored on the minimization of the Mean Squared Error cost function (MSE). This approach is only optimal if the errors are distributed according to a Gaussian distribution without samples that strongly deviate from the main mode, i.e. outliers. If such assumption is violated, the sparse code will likely be biased and performance will degrade accordingly. In this paper, we introduce five robust variants of OMP (RobOMP) fully based on the theory of M-Estimators under a linear model. The proposed framework exploits efficient Iteratively Reweighted Least Squares (IRLS) techniques to mitigate the effect of outliers and emphasize the samples corresponding to the main mode of the data. This is done adaptively via a learned weight vector that models the distribution of the data in a robust manner. Experiments on synthetic data under several noise distributions and image recognition under different combinations of occlusion and missing pixels thoroughly detail the superiority of RobOMP over MSE-based approaches and similar robust alternatives. We also introduce a denoising framework based on robust, sparse and redundant representations that open the door to potential further applications of the proposed techniques. The five different variants of RobOMP do not require parameter tuning from the user and, hence, constitute principled alternatives to OMP.
Sparse coding aims to find a parsimonious representation of an example given an observation matrix or dictionary. In this regard, Orthogonal Matching Pursuit (OMP) provides an intuitive, simple and fast approximation of the optimal solution. However, its main building block is anchored on the minimization of the Mean Squared Error cost function (MSE). This approach is only optimal if the errors are distributed according to a Gaussian distribution without samples that strongly deviate from the main mode, i.e. outliers. If such assumption is violated, the sparse code will likely be biased and performance will degrade accordingly. In this paper, we introduce five robust variants of OMP (RobOMP) fully based on the theory of M-Estimators under a linear model. The proposed framework exploits efficient Iteratively Reweighted Least Squares (IRLS) techniques to mitigate the effect of outliers and emphasize the samples corresponding to the main mode of the data. This is done adaptively via a learned weight vector that models the distribution of the data in a robust manner. Experiments on synthetic data under several noise distributions and image recognition under different combinations of occlusion and missing pixels thoroughly detail the superiority of RobOMP over MSE-based approaches and similar robust alternatives. We also introduce a denoising framework based on robust, sparse and redundant representations that open the door to potential further applications of the proposed techniques. The five different variants of RobOMP do not require parameter tuning from the user and, hence, constitute principled alternatives to OMP.