Search

This paper proposes a framework combing a fully convolutional network with broad learning system for license plate recognition. The fully convolutional network, which is designed as a pixel-level two-class classification method, is proposed for random-positioned object detection by the fusion of multi-scale and hierarchical features. For character segmentation, a trained AdaBoost cascade classifier is employed to locate a key character representing for an administrative area. We design a symmetric region horizontal projection method to estimate the license plate slant angles, and an approach based on vertical projection without hyphens to solve the problem of touching characters. For character recognition, the broad learning system with stacked auto-encoder of mapped feature nodes is proposed, and two structures are explored to recognize letters and digits, respectively. Experiments conducted on Macau license plates show that the proposed method outperforms some state-of-the-art approaches. The compatibility and generality can be expected by applying the proposed method to other regions or countries.

Automatic modulation classification is essential in radar emitter identification. We propose a cascade classifier by combining a support vector machine (SVM) and convolutional neural network (CNN), considering that noise might be taken as radar signals. First, the SVM distinguishes noise signals by the main ridge slice feature of signals. Second, the complex envelope features of the predicted radar signals are extracted and placed into a designed CNN, where a modulation classification task is performed. Simulation results show that the SVM‐CNN can effectively distinguish radar signals from noise. The overall probability of successful recognition (PSR) of modulation is 98.52% at 20 dB and 82.27% at −2 dB with low computation costs. Furthermore, we found that the accuracy of intermediate frequency estimation significantly affects the PSR. This study shows the possibility of training a classifier using complex envelope features. What the proposed CNN has learned can be interpreted as an equivalent matched filter consisting of a series of small filters that can provide different responses determined by envelope features.

Indoor localization of human objects has many important applications nowadays. Here, we propose a new multiresolution device-free indoor localization scheme using wireless radio frequency (RF) fingerprints. In our proposed new device-free approach, all transceiver devices are fixed in an indoor environment, so human targets do not need to carry any transceiver device with them. First, the indoor geometry is divided into several zones, and then, the received signal strength indicators measured by the receiving antennas are input features to our designed innovative machine learning model to identify within which zone the target is. Our proposed machine learning model, i.e., a multiresolution random forest classifier, is composed of a cascade architecture, which integrates and distills learned results over various zoning resolutions. The proposed new multiresolution approach greatly outperforms the existing random forest classifier. The average Euclidean distance error resulting from our proposed new technique is 1.0376 m.

Nowadays we are using cab system very much because it can fulfil our needs quickly. The most priorable problems are accidents by drowsiness of the drivers, environmental conditions, vehicle conditions etc. Improper cab driver behaviour leads to increase the probability of accidents occurring. Detecting the driver’s behaviour is one of interested research area. An idea of this project is proposed by AI, it is an three level process i.e., face detection and face recognition using viola jones algorithm.

Emotion recognition is the most regularly reviewed concept in examining behavioural perception and human-computer interface. The proposed system considers happy, normal and surprise categories of emotions. The dataset comprises mouth images containing emotions in the form of video frames. The Haar-based cascade classifier is used to segment the mouth region in the face images and the video frames are extracted at 20 frames per second. The histogram of oriented gradients (HOG) and local binary pattern (LBP) are applied to capture the edge and local information from the emotion image by extracting the gradient information and finally combined into a single histogram to form the features where each histogram represents a mouth image. The speeded-up robust features (SURF) and scale-invariant feature transform (SIFT) methods are employed to extract unique points in emotion image with varying number of key-points as features. The support vector machine (SVM) and one-dimensional convolutional neural network (1D-CNN) are trained using these texture features. Trained models are utilized to detect emotions in testing video frames. The experimental results depict that SVM and 1D-CNN achieve an accuracy of 97.44% and 98.51% respectively.

Recognition of license plate is the process of wherein photographic video or images of license plates are being captured and then processed using an application that implements series of algorithms that will provide the alpha numeric conversion of the captured data. In this study, the researchers developed a license plate recognition that incorporates image deblurring to accommodate multiple recognition from video sequences. The approach uses Background Subtraction and Connected Component Analysis for the detection of license plates, Image deblurring to enhance the image and reduce the difficulties in recognition, and LBP Cascade Classifier was implemented for recognition of characters. Since multiple detection for license plate produces different difficulties such as motion blur and camera angle view the approach attempts to minimize the effects of these problems while still being applicable to multiple detection. 30 videos with 3 minutes length each of actual traffic situation were gathered and recorded at the footbridge of UP Ayala Technohub, Commowealth Ave. Quezon City, Philippines and 10 of these videos were used as input for the testing and experiment of the system. The accuracy for plate detection were computed using F-measure which yields to 87.32% for both system with image deblurring and none, while the accuracy for character recognition is 62.66% for system with image deblurring and 48.25% for the system without image deblurring. The result shows that there is a significant difference in the accuracy of license plate recognition between the system with image deblurring and without image deblurring

The large number of citizens in an organization makes the development of an attendance system or citizen detection in a place important in the running of work activities in the organization. Utilization of an IP Camera which is only used for regular monitoring without further detection of the needs of citizens in the organization made the development of personnel detection developed for monitoring the presence of personnel. With the development of a face detection system, it is hoped that the facial algorithm development system will be developed using an IP Camera. Face detection has been developed which has many and special features which aim to determine whether or not a face has been detected in an image. With image management that is developed in face detection, detection will be faster and more accurate because the color is processed into gray degrees so that there are fewer color pixels than those with colors. By using the Python programming language and an image detection library called OpenCV, less code will be designed. This study uses the Viola Jones method, which is a fast and accurate face detection method developed by Paul Viola and Michael Jones. In this study, the Viola Jones method uses the Haar Cascade algorithm which functions as a detection feature in the system and is combined with the internal image process and the AdaBoost Learning and Cascade Classifier so that the detected face object will easily classify whether the object is a face or not. In this case the Cascade Classfier used in this study is the face and eyes. The development of this algorithm is carried out for face detection and recognition. The detection is done by taking pictures with the process taken using a webcam. The system will take several pictures and then the image data will be stored in a folder called dataSet. After that, all data is trained so that it can be recognized by the system. With retrieval, detection and recognition limitations that can only be taken from a distance of less than three meters, face detection on the IP Camera can still read objects other than faces. With recognition and accuracy on the webcam camera, about 80,5% this system can be developed with the Haar Cascade algorithm and the Haar Cascade algorithm precisely to be applied to the development of faced detection and face recognition. By developing the Haar Cascade algorithm for face detection, problems and utilization of an organization's data can be more easily detected and used by IP cameras that can support the performance process of face detection and recognition

With the development of deep learning, the performance of many computer vision tasks has been greatly improved. For visual tracking, deep learning methods mainly focus on extracting better features or designing end-to-end trackers. However, during tracking specific targets most of the existing trackers based on deep learning are less discriminative and time-consuming. In this paper, a cascade based tracking algorithm is proposed to promote the robustness of the tracker and reduce time consumption. First, we propose a novel deep network for feature extraction. Since some pruning strategies are applied, the speed of the feature extraction stage can be more than 50 frames per second. Then, a cascade tracker named DCCT is presented to improve the performance and enhance the robustness by utilizing both texture and semantic features. Similar to the cascade classifier, the proposed DCCT tracker consists of several weaker trackers. Each weak tracker rejects some false candidates of the tracked object, and the final tracking results are obtained by synthesizing these weak trackers. Intensive experiments are conducted in some public datasets and the results have demonstrated the effectiveness of the proposed framework.

In this paper, an efficient framework for counting pedestrians crossing a line of interest is proposed. Nowadays, the convolutional neural networks have very good results on pedestrian detection and tracking. However, the major drawback of the neural networks is that they require heavy computing resources. This limits the application of neural networks in low-cost systems. Thus, the low power consuming pedestrian counting systems with comparable performance are still important. To achieve this goal, the proposed method distils the pedestrian detection knowledge from a neural network to train the local binary patterns (LBP) cascade classifier model to detect pedestrians. Then a matching and tracking algorithm is used to count the number of pedestrians. An automaton was developed to eliminate the bouncing position of the detected pedestrians. The experimental comparisons show that, compared to Ma et al. and Felzenszwalb et al.’s methods, the quality of the line of interest counting of the proposed method is about the same and, at the same time, the execution time of the proposed method is much less.

The article addresses the task of improving target recognition in onboard information support system for search and rescue operations in the Arctic region. One of the tasks performed by the complex is recognition of objects in twodimensional camera images, which suffer from the loss of the image brightness, being formed by constant brightness principle, with its direct impact on the probability of target recognition. To preserve the brightness of the image, the authors propose to process the primary signals of the camera according to the principle of constant color brightness. The proposed processing can increase the probability of correct target recognition. The paper analyzes the principles of encoding the primary signals of the television camera. For the object recognition problem, the cascades were trained based on the cascade classifier using the principles of constant brightness and constant color brightness. The output of the trained cascades has confirmed that the processing of the primary signals of a television camera based on the principle of constant brightness improves target recognition and therefore will increase the object recognition performance of the complex under development.

Background/Objectives: The deep learning approaches have paved their way to construct various artificial intelligence products and the proposed system uses a convolutional neural network for detecting real-time emotions of mankind. The objective of the study is to develop a real-time application for emotion recognition using convolutional neural network and transfer learning methods. Methods/Statistical analysis: The proposed system considers happy, normal and surprised categories of emotions. The system consists of four major steps: dataset collection, training, validation, and real-time testing. The dataset is comprised of face images containing emotions such as happy,normal and surprised in the form of video frames. The face and mouth regions are detected using the Haar-Based Cascade classifier at 20 frames per second. Findings: The convolutional neural network (CNN) is trained using mouth images and the pre-trained models VGG16 and VGG19 are trained with face images. The trained model is used to detect the emotions in the live webcam video. The experimental results show that the CNN model trained using mouth images gives an accuracy of 85.71% and the pre-trained models trained with face images using transfer learning method achieves an accuracy of 77.78%. The proposed system using CNN outperforms the pre-trained models for recognizing the emotions in real-time video. Novelty/Applications:The proposed system is entirely based on the mouth region video frames and the real-time emotion recognition system is developed. This work can detect the three emotions in an unconstrained laboratory environment. Keywords: Convolutional neural network; mouth detection; pre-trained models; real-time emotion recognition; transfer learning

Warning signs are of great significance to traffic safety. In this paper, a real-time recognition method for Chinese Traffic Warning Signs (CTWS) is proposed. CTWS are all triangles with yellow background, black border and black pattern. Their similarity is conducive to the localization task of object detection but adverse to the classification task of object detection. After analyzing the characteristics of these signs, real-time recognition for CTWS is carried out by employing Cascade classifier and Convolutional Neural Network (CNN). A Cascade classifier with 9 layers is trained with local binary patterns to locate the CTWS in frames. And a 10-layer CNN model is built to determine the specific category of the signs located by the Cascade classifier. We evaluate the method on CCTSDB-based dataset and GTSDB, and experiments show that the proposed method can perform accurate recognition at an average speed of 81.79fps without GPU. Since the proposed method only needs to call CNN that requires vast computing power in a small number of frames containing CTWS while performing real-time recognition, it can effectively save the valuable on-board computing resources compared with other object detection algorithms that is purely based on CNN such as YOLOv3, YOLOv3-tiny and Faster R-CNN.

Traditional iris recognition methods, which are still preferred against artificial intelligence (AI) approaches in practical applications, are often required to capture high-grade iris samples by an iris scanner for accurate subsequent processing. To reduce the system cost for mass deployment of iris recognition, pricey scan devices can be replaced by the average quality cameras combined with additional processing algorithm. In this paper, we propose a Haar-like-feature-based iris localization method to quickly detect the location of human iris in the images captured by low-cost cameras for the ease of post-processing stages. The AdaBoost algorithm was chosen as a learning method for training a cascade classifier using Haar-like features, which was then utilized to detect the iris position. The experimental results have shown acceptable accuracy and processing speed for this novel cascade classifier. This achievement stimulates us to implement this novel capturing device in our iris recognition.

Online learning algorithms have been widely used to address vision-related issues such as object detection and tracking. However, a robust online learning object detection system that can continuously improve performance through self-learning continues to elude designers. This study proposes a novel online learning framework, which combines detection and verification modules to train a scene-specific head detector on a fly. For the detection module, a proposed online bootstrap cascade classifier is employed as the object detector of the framework. The cascade decision strategy is used to integrate a number of weak online classifiers. The resulting system contains sufficient weak classifiers and maintains a low computation cost. During the online learning process, the complexity of the cascade structure adapts to the difficulty of the detection task. For the verification module, a simple yet effective particle filter tracking algorithm, based on information fusion, is used to automatically label online learning samples produced by detection responses. With this method, the object detector improves detection performance by autonomously learning the samples. The online head detection framework is ported to the NVIDIA Jetson TK1 embedded platform, which enables the platform to recognize different head postures through self-learning. Experimental results on three video datasets demonstrate the effectiveness of the framework.

This paper describes the classification of four different varieties of rice grain based on four sets of features, namely morphology, colour, texture and wavelet. The classification is carried out on single rice kernel using image pre-processing steps followed by a cascade network classifier. The performance of the classifiers based on the above feature sets is also compared. It is found that morphological feature is more suitable for the classification of rice kernels, as compared to other features. The number of input features is reduced by a feature selection process using statistical analysis system (SAS) software. The classification accuracy based on selected features is compared with that of original features using different classifiers. It is found that the selected features are able to provide classification accuracy very close to the original features. The performance of the proposed cascade classifier is also tested against standard datasets from the University of California, Irvine (UCI), and the results are compared with other classifiers. The results show that the proposed classifier provides better classification accuracy as compared to other classifiers.

To address the low accuracy problem of playground detection under complex background, the accurate playground localization based on multi-feature extraction and cascade classifier is proposed in th...

In this paper, a new approach is introduced upgrading the complex object recognizing monitoring system up to the image processing system capable of operating both in the visible and the infrared wavelength ranges. For this purpose, both new algorithmic software and user interface are provided that require from the operator neither special knowledge, nor specific competencies in the fields of object detection, tracking and recognition, while allowing determining the thermal imager parameters necessary for constructing a high-quality image of an object. There are formulated the conditions required for obtaining such image that, by its quality, would make the satisfactory detection of the desired object possible. By means of the conducted tests, it is demonstrated that the application of the proposed mathematical and algorithmic support of the complex monitoring and control system provides the solution for the problem of the highly accurate individual recognition.

In recent decades, the traffic on road increased in a huge number. It is very important to manage the safety of the humans as well as to make an efficient flow of the traffic. To manage the traffic flow and to overcome from the situation of the traffic congestion the vehicle detection and counting needs a greater amount of accuracy. In this work, two different techniques are proposed that provides better performance in terms of F-Measure score and Error Ratio . The first technique is based on the foreground estimation while the second proposed technique is based on the training of dataset using a cascade classifier which is based on the Histogram of Oriented Gradients (HOG). Furthermore, four images are provided at once to the proposed system to count the vehicles and generate a signal that shows a greater number of vehicles in that image. The priority of each image will be set on the basics of greater number of vehicles present. The proposed techniques showed outstanding performance on a sunny and a cloudy day which is verified from the experimental results.

This paper proposes an improved AdaBoost classifier for sonar images with low resolution ratio and noise. First, Histogram of oriented gradient (HOG) is used to perform feature extraction, and a weak classifier is obtained by Support vector machine (SVM) at the same time. Then, multiple SVM models are constructed for target classification based on the AdaBoost cascade classification framework. A new function for updating sample weights has been designed in this paper to improve the accuracy of the classifier. And new iteration rules of classifier have been made to reduce the training time of the proposed method. The experimental results on the sonar dataset which are proposed for improving the generalization ability in this paper show that the classification accuracy of the proposed algorithm is about 92%, and the accuracy on Cifar-10 dataset is also higher than general methods.

A number of vision-based algorithms designed to detect and track unmanned aerial vehicles (UAVs) from on board a second UAV have been researched, implemented, and experimentally validated over the last decade. However, the successful methods have tended to rely on characteristics such as color or shape, meaning they require the target UAV to have particular markings or geometries. This paper uses the Viola–Jones cascade classifier, a computer vision algorithm originally designed to detect human faces in video streams, and demonstrates its capability for detecting and tracking an arbitrary type of UAV with excellent performance in either indoor or outdoor environments and with a variety of backgrounds. The Viola–Jones algorithm is applied to two specific quadrotor UAV models, the Solo from 3D Robotics and the AR.Drone 2.0 from Parrot. Experimental testing demonstrates that the resulting system achieves very good detection and tracking performance in real time on each UAV type for both indoor and outdoor flight tests.

Melanoma is the most dangerous form of skin cancer, which is responsible for the majority of skin cancer-related deaths. Early diagnosis of melanoma can significantly reduce mortality rates and treatment costs. Therefore, skin cancer specialists are using image-based diagnostic tools for detecting melanoma earlier. We aim to develop a handheld device featured with low cost and high performance to enhance early detection of melanoma at the primary healthcare. But, developing this device is very challenging due to the complicated computations required by the embedded diagnosis system. Thus, we aim to exploit the recent hardware technology in reconfigurable computing to achieve a high-performance embedded system at low cost. Support vector machine (SVM) is a common classifier that shows high accuracy for classifying melanoma within the diagnosis system and is considered as the most compute-intensive task in the system. In this paper, we propose a dynamic hardware system for implementing a cascade SVM classifier on FPGA for early melanoma detection. A multi-core architecture is proposed to implement a two-stage cascade classifier using two classifiers with accuracies of 98% and 73%. The hardware implementation results were optimized by using the dynamic partial reconfiguration technology, where very low resource utilization of 1% slices and power consumption of 1.5 W were achieved. Consequently, the implemented dynamic hardware system meets vital embedded system constraints of high performance and low cost, resource utilization, and power consumption, while achieving efficient classification with high accuracy., Journal paper, 9 pages, 4 figures, 4 tables

Segmenting complex 3D geometry is a challenging task due to rich structural details and complex appearance variations of target object. Shape representation and foreground-background delineation are two of the core components of segmentation. Explicit shape models, such as mesh based representations, suffer from poor handling of topological changes. On the other hand, implicit shape models, such as level-set based representations, have limited capacity for interactive manipulation. Fully automatic segmentation for separating foreground objects from background generally utilizes non-interoperable machine learning methods, which heavily rely on the off-line training dataset and are limited to the discrimination power of the chosen model. To address these issues, we propose a novel semi-implicit representation method, namely Non-Uniform Implicit B-spline Surface (NU-IBS), which adaptively distributes parametrically blended patches according to geometrical complexity. Then, a two-stage cascade classifier is introduced to carry out efficient foreground and background delineation, where a simplistic Naive-Bayesian model is trained for fast background elimination, followed by a stronger pseudo-3D Convolutional Neural Network (CNN) multi-scale classifier to precisely identify the foreground objects. A localized interactive and adaptive segmentation scheme is incorporated to boost the delineation accuracy by utilizing the information iteratively gained from user intervention. The segmentation result is obtained via deforming an NU-IBS according to the probabilistic interpretation of delineated regions, which also imposes a homogeneity constrain for individual segments. The proposed method is evaluated on a 3D cardiovascular Computed Tomography Angiography (CTA) image dataset and Brain Tumor Image Segmentation Benchmark 2015 (BraTS2015) 3D Magnetic Resonance Imaging (MRI) dataset.

In this research, the scientific novelty is the development and testing of the method. This method will allow combining the high speed of the cascade classifier and the quality of the artificial neural network. As an approbation of the work in this article, we study the set of image files. The set of image files are available in the directory on the disk. Next, we select a set of files. The set of files is similar to the object of interest image files. It means, that we can't identical any objects. Then we rewrite files to the target directory. This set should later be used for training a neural network. Images of the object of interest can be in any orientation and on any background. This non-identity must be recognized by a trained neural network. For the purposes of forming such a set, we will need an identifier that has sufficient accuracy for these purposes. Here we do not need very high accuracy. The very identification error of this preliminary analyzer helps us of introducing some dissimilarity between the objects of the images. The task is complicated by the fact that objects can have different sizes (scale) and orientation in space. Our approach based on cascade classifiers and neural networks combines the high speed of the cascade classifier and the quality of the artificial neural network. And the stage of preparing training data for the network is of great importance.

It is important problem to pick up the objects that is arranged at random automatically. And it is desirable to adapt the problem such that the mobile robot gathers spread objects in the field in the real world. In this paper, the method that consists of an approaching process using the cascade classifier and a picking process using the deep reinforcement learning is proposed to solve the problem. First, in the approaching process, the method to detect the object and to estimate the position using two camera images is proposed. Second, in the picking process, the reward matrix to learn it using the deep reinforcement learning is introduced newly. The effectiveness of each process is investigated by the simulations built by three-dimensional physical simulation using PyBullet. Final, we confirm the realization of the method that consists of the approaching process and the picking process by the experiments using the real robot at real-time.

Although machine learning and artificial intelligence have been widely applied, noise and interference are still major disturbances to degrade the quality of image transmission and processing efficiency in multimedia data transmission. Based on the face detection system utilizing the machine learning algorithms and artificial intelligence, this paper examines, analyzes and compares the performances of the AdaBoost machine learning algorithm and Convolutional Neural Network (CNN)-based algorithm in processing face information, which is disturbed by channel noise and fading effect encountered in the transmission of face image. The face detection system used is based on HAAR feature extraction. The extracted HAAR features are subjected to classification training and learning of the cascade classifier. Then the face detection is performed on the picture information outside the database. Results of computer simulation show that for the image data affected by the fading and AWGN, the face recognition system still marked the positions of the eyes and mouth with high accuracy. However for fading and higher AWGN, using a machine learning algorithm with a convolutional neural network is better than the ADABOOST algorithm. It can be concluded that the machine learning algorithm can effectively reduce the adverse effect of multimedia data transmission without increasing the SNR and use of higher level of modulation scheme.

Traffic sign is important as they provide visual information to drivers. Traffic signal detection is a technique used detects traffic signs put on a road and inform it to a driver and it is an integral part of intelligent driver assistance. This paper proposed as hardware accelerator using cascade classifiers based on FPGA to detect the traffic signs. Cascade classifier consists of number of stages and adaptive boosting technique is trainer for the cascade classifier. FPGA is reconfigurable, cost efficient and more flexible than other hardware platforms. The system is implemented using model-based design (Simulink) and hardware software co-design. Proposed design is analyzed on a Xilinx board.

Optical Music Recognition (OMR) has become a study trend with the increasing demand for digital sheet music. In this paper, we explore techniques and algorithms to implement optical music recognition. This paper aims to encourage people who just begin and enjoy learning object detection by using a simple and comprehensible framework called Haar-like Feature to detect the music notation. Furthermore, it also assists beginner musicians who have a difficult time in memorizing the music theory and rules by generating musical alphabets. The paper will include the process of how to generate the cascade classifier model and how to imply them to detect the target object.

By considering skin color, occasion and dress color as input data, this paper proposes a hierarchical cascade classifier to develop a guideline of personalized facial makeup. Although the makeup recommendation system was previously studied in many researches, but the suggestion cannot be applied for a person accurately in real situation. Color tone based on color wheel theory for facial makeup and color selection from individual skin tone were employed in this study. There were two stages of the hierarchical cascade classifier. The first stage was relied on a rule-based classification procedure, in which rules can be generated by investigating input data within the scope of this research together with the data from a professional makeup artist and 250 face images with makeup originated by makeup experts, resulting in primary color of eye shadow, cheek blush color, and lipstick color. Next, machine learning concept was used as the second stage of the hierarchical cascade classifier to indicate secondary color of eye shadow and alternative lipstick color corresponding to a feature vector. Six classification models, which are Multi-Layer Perceptron, Logistic Regression classifier, Support Vector Machine, Decision Tree, k-nearest neighbor classifier, and Naive Bayes classifier were selected in this study. From the experimental results, the mixture of rule-based classifier and Multi-Layer Perceptron was suitable to be used as a guideline of personalized facial makeup.

To improve the accuracy of lane detection in complex environment, The lane detection algorithm based on Haar feature coupled cascade classifier was proposed. The input image was scaled, and the Regions of Interest (ROI) was extracted according to the position of the vanishing line. The Haar feature of the lane was extracted from the ROI, and a cascade Lane classifier was introduced to roughly detect the lane in the ROI. The line segment detector (LSD) method was used to fit roughly detected lanes. The growth strategy and geometric checking were combined to optimize fitting results and complete target detection. The tests were conducted on multiple datasets, and results show compared with current lane detection methods, the proposed algorithm has higher robustness and accuracy, which was up to 96.5%.

In the era of smart digital transformation, automatic face recognition is the way of identification and verification of a person in many applications of security and authentication. If the person's 2D still image or 3D video frame is taken under controlled lighting and frontal face poses, then today automatic face recognition is a solved problem with more than 98 % accuracy. Under ideal condition of images, automatic face recognition outperforms manual face recognition rate. But the problem of uncontrolled illumination, occlusion, tilting of faces, different expressions of faces, use of accessories and hair color, growth of facial hair, aging effect of a person and low-resolution images makes automatic face recognition underperform; it gets defeated by human. Unless this problem is solved on real time, automatic face recognition system cannot be trusted for crucial security applications like e-passport, fraud detection, counter terrorism and mug-shot verification etc. Extensive research is underway to improve this technology. Traditionally Harr cascade classifier methods, histogram of oriented gradients (HoG), principal component analysis (PCA), Eigen-faces, and support vector machine (SVM) classifier are used in face recognition. Now modern deep learning innovation has superseded those traditional machine learning techniques with more computing power in GPUs and TPUs to tackle all variations in input image quality and ever-growing face database. The goal of this paper is to report the journey of face recognition system from traditional to modern techniques to increase precision and recall of the system.

As a promising method for robotic catheter therapeutics, a controllable wireless microrobot system with a less invasive intraoperative procedure has been presented for thrombosis surgery. The controllable wireless microrobot utilizes an electromagnetic actuator system for microrobot actuation and biplane X-ray system for microrobot imaging. We propose the new 3D tracking method of microrobot for this system. As a learning-based system, cascade classifier was adapted for real-time microrobot tracking. Combination of cascade classifier and contour-based system made accurate microrobot detection method on 2D projected X-ray images. Kalman filter interpolated lost frames, and triangulation method reconstructed 3D pose (position and orientation) from 4 endpoints of microrobot pairs. Also, the semi-automatic calibration method for bi-plane C-arm devices was proposed by using modified shortest path algorithm. To find the optimal rescaling setting, 3D tracking accuracy was investigated according to the rescaling of original intensity values. The experiment results showed a good performance with tracking errors of 2.37±9.91mm in position and 6.53±13.80° in orientation under 2200 optimal width of rescaling. If tracking evaluation is constrained under frames detected by cascade classifier, 3D tracking errors improved sincerely by 0.28±1.13mm in position and 3.48±2.89° in orientation. The optimal width of the rescaling setting needed bigger value than the mean of intensity values. The proposed tracking technique accomplished a fast frame rate of 34.72 frames/sec under OpenCL implementation of OpenCV. Learning based system is robust to the change of graphics setting since training can adapt rapidly to this change. Also, proposed framework can cooperate with the change of microrobot shape if a contour-based method is optimized. Hence, the method can be used for therapeutic millimeter- or micron-sized manipulator recognition in vascular, as well as implanted objects in the human body.

Face recognition is one of current biometrics identification methods, that based on the measuring to one of human biological characteristics and utilize them to recognize individuals. these characteristics which are called biometric they are hard to fake because they identify a person by measuring one of its biological characteristics such as (finger print, iris print and face print). With the rapid improvement of mobile technologies that happen in last decade face recognition process can make using mobile phone, this paper explains the building of mobile face recognition system using Eigen face approach, Experimental results have been tested on a local data-set that has been created to analyze the efficiency of the application in various cases including different illumination conditions, variation of view, and orientation, the recognition rate of the application when testing on Galaxy Grand Prime + was 78.4. while The recognition rate when testing on Galaxy Note 5 was 82.4. The accuracy of this application can reach to 100% if we use camera with high accuracy and on good light condition.

In this article, the authors propose a concise corner detection algorithm, which is called CCDA. A cascade classifier concept is used to derive a corner detector, which can quickly discard the most non-corner pixels. The ruler of gradient direction is used to get the corner, which can avoid the influence of the light change. The method of second derivative non-maximum suppression is used to get the location of the corner and can get the exact corner point. As a result, CCDA is compare-tested with classical corner detection algorithms by using the same images which include synthetic corner patterns and real images. The result shows that CCDA has a similar speed to the FAST algorithm and better accuracy and robustness than the HARRIS algorithm.

Background Acetylation on lysine is a widespread post-translational modification which is reversible and plays a crucial role in some biological activities. To better understand the mechanism, it is necessary to identify acetylation sites in proteins accurately. Computational methods are popular because they are more convenient and faster than experimental methods. In this study, we proposed a new computational method to predict acetylation sites in human by combining sequence features and structural features including physicochemical property (PCP), position specific score matrix (PSSM), auto covariation (AC), residue composition (RC), secondary structure (SS) and accessible surface area (ASA), which can well characterize the information of acetylated lysine sites. Besides, a two-step feature selection was applied, which combined mRMR and IFS. It finally trained a cascade classifier based on SVM, which successfully solved the imbalance between positive samples and negative samples and covered all negative sample information. Results The performance of this method is measured with a specificity of 72.19% and a sensibility of 76.71% on independent dataset which shows that a cascade SVM classifier outperforms single SVM classifier. Conclusions In addition to the analysis of experimental results, we also made a systematic and comprehensive analysis of the acetylation data.

In general, human are given the mind and mind to be able to determine or be able to didtinguish individuals who appear either human, animal, plant, and other objects that are known or unknown. And it is possible for human to recognize these object from their sight and from their brain memory. Especially on the human face, human can recognize whether the object is human or not human, and can recognize the object very well through his own eyes.face detection system in human becomes very important in the development of science of digital image processing. The research has been done with many advantages and disadvantages. From a face many information features that can be read, such as eyes, nose, and mouth. The detection system uses Viola Jones method as an object detection method. The Viola Jones method is known to have considerable Speed and accuracy as it combines several concepts (Haar feature, Integral image, Adaboost, Cascade classifier) into a main method for detecting objects.Based on tests conducted on face identification under conditions that may affect face detection results, the results show an accuracy of 67,6 % to detect the face.

Face detection is considered as a one of the most important issues in the identification and authentication systems which use biometric features. Face detection is not straightforward as long as it has lots of dissimilarity of image appearance. Some challenges are required to be resolved to improve the detection process. These challenges include environmental constraints, device specific constraints and the facial feature constraints. Here in our paper we present a modified method to enrich face detection by using combination of Haar cascade files using skin detection, eye detection and nose detection. Our proposed system has been evaluated using Wild database. The experimental results have shown that the proposed system can achieve accuracy of detection up to 96%. Also, here we compared the proposed system with the other face detection systems and the success rate of the proposed system is better than the considered systems.

Crescent sand dunes called barchans are the fastest moving sand dunes in the desert, causing disturbance for infrastructure and threatening human settlements. Their study is of great interest for urban planners and geologists interested in desertification (Hugenholtz et al., 2012). In order to study them at a large scale, the use of remote sensing is necessary. Indeed, barchans can be part of barchan fields which can be composed of thousands of dunes (Elbelrhiti et al.2008). Our region of interest is located in the south of Morocco, near the city of Laayoune, where barchans are stretching over a 400 km corridor of sand dunes. We used image processing techniques based on machine learning approaches to detect both the location and the outlines of barchan dunes. The process we developed combined two main parts: The first one consists of the detection of crescent shaped dunes in satellite images using a supervised learning method and the second one is the mapping of barchans contours (windward, brink and leeward) defining their 2D pattern. For the detection, we started by image enhancement techniques using contrast adjustment by histogram equalization along with noise reduction filters. We then used a supervised learning method: We annotated the samples and trained a hierarchical cascade classifier that we tested with both Haar and LBP features (Viola et Jones, 2001; Liao et al., 2007). Then, we merged positive bounding boxes exceeding a defined overlapping ratio. The positive examples were then qualified to the second part of our approach, where the exact contours were mapped using an image processing algorithm: We trained an ASM (Active Shape Model) (Cootes et al., 1995) to recognize the contours of barchans. We started by selecting a sample with 100 barchan dunes with 30 landmarks (10 landmarks for each one of the 3 outlines). We then aligned the shapes using Procrustes analysis, before proceeding to reduce the dimensionality using PCA. Finally, we tested different descriptors for the profiles matching: HOG features were used to construct a multivariate Gaussian model, and then SURF descriptors were fed an SVM. The result was a recursive model that successfully mapped the contours of barchans dunes. We experimented with IKONOS high resolution satellite images. The use of IKONOS high resolution satellite images proved useful not only to have a good accuracy, but also allowed to map the contours of barchans sand dunes with a high precision. Overall, the execution time of the combined methods was very satisfying.

Kecelakaan lalu lintas sering terjadi disebabkan kendaraan yang melaju dengan kecepatan tinggi. Penelitian ini mengembangkan sistem penegakan speed bump (polisi tidur) yang dapat memberikan peringatan bagi pengemudi dalam memperlambat laju kendaraan dan memberikan kenyamanan saat melaju dengan kecepatan rendah. Penegakan speed bump dilakukan berdasarkan kecepatan kendaraan yang terdeteksi menggunakan metode Haar Cascade Classifier, yang merupakan gabungan beberapa konsep yaitu Haar Features, Integral Image, AdaBoost Learning, dan Cascade Classifier. Pengujian dilakukan menggunakan video jalan raya pada satu jalur. Sistem dibuat menggunakan interpreter python dengan library OpenCV. Pendeteksian didapatkan hasil yang cukup baik apabila dilakukan pada intensitas cahaya tinggi, dan didapatkan tingkat akurasi deteksi sebesar 97,92%. Perhitungan kecepatan kendaraan didapatkan dengan membandingkan hasil kecepatan pada sistem dengan video dalam keadaan real time, yang dibuktikan dari tingkat error dengan nilai MSE yaitu 2,88.

Objective Develop an approach, One-class-at-a-time, for triaging psychiatric patients using machine learning on textual patient records. Our approach aims to automate the triaging process and reduce expert effort while providing high classification reliability. Materials and Methods The One-class-at-a-time approach is a multistage cascading classification technique that achieves higher triage classification accuracy compared to traditional multiclass classifiers through 1) classifying one class at a time (or stage), and 2) identification and application of the highest accuracy classifier at each stage. The approach was evaluated using a unique dataset of 433 psychiatric patient records with a triage class label provided by “I2B2 challenge,” a recent competition in the medical informatics community. Results The One-class-at-a-time cascading classifier outperformed state-of-the-art classification techniques with overall classification accuracy of 77% among 4 classes, exceeding accuracies of existing multiclass classifiers. The approach also enabled highly accurate classification of individual classes—the severe and mild with 85% accuracy, moderate with 64% accuracy, and absent with 60% accuracy. Discussion The triaging of psychiatric cases is a challenging problem due to the lack of clear guidelines and protocols. Our work presents a machine learning approach using psychiatric records for triaging patients based on their severity condition. Conclusion The One-class-at-a-time cascading classifier can be used as a decision aid to reduce triaging effort of physicians and nurses, while providing a unique opportunity to involve experts at each stage to reduce false positive and further improve the system’s accuracy.

High-frequency surface wave radar (HFSWR) has a vital civilian and military significance for ship target detection and tracking because of its wide visual field and large sea area coverage. However, most of the existing ship target detection methods of HFSWR face two main difficulties: (1) the received radar signals are strongly polluted by clutter and noises, and (2) it is difficult to detect ship targets in real-time due to high computational complexity. This paper presents a ship target detection algorithm to overcome the problems above by using a two-stage cascade classification structure. Firstly, to quickly obtain the target candidate regions, a simple gray-scale feature and a linear classifier were applied. Then, a new error self-adjustment extreme learning machine (ES-ELM) with Haar-like input features was adopted to further identify the target precisely in each candidate region. The proposed ES-ELM includes two parts: initialization part and updating part. In the former stage, the L1 regularizer process is adopted to find the sparse solution of output weights, to prune the useless neural nodes and to obtain the optimal number of hidden neurons. Also, to ensure an excellent generalization performance by the network, in the latter stage, the parameters of hidden layer are updated through several iterations using L2 regularizer process with pulled back error matrix. This process yields appropriate output weights and the appropriate hidden weights. Experimental results show that (1) compared with standard ELM, our proposed ES-ELM has higher classification accuracy and training efficiency, and the generalization performance is not sensitive to regularization parameter, (2) the proposed ship target detection algorithm based on ES-ELM outperforms most of the state-of-the-art methods for detection accuracy and computational efficiency.

Objective: Cataract, which is the clouding of the crystalline lens, is the most prevalent eye disease accounting for 51% of all eye diseases in the U.S. Cataract is a progressive disease, and its early detection is critical for preventing blindness. In this paper, an efficient approach to identify cataract disease by adopting luminance features using a smartphone is proposed. Methods: Initially, eye images captured by a smartphone were cropped to extract the lens, and the images were preprocessed to remove irrelevant background and noise by utilizing median filter and watershed transformation. Then, a novel luminance transformation from pixel brightness algorithm was introduced to extract lens image features. The luminance and texture features of different types of cataract disease images could be obtained accurately in this stage. Finally, by adopting support vector machines (SVM) as the classification method, cataract eyes were identified. Results: From all the images that we fed into our system, our method could diagnose diseased eyes with 96.6% accuracy, 93.4% specificity, and 93.75% sensitivity. Conclusion: The proposed method provides an affordable, rapid, easy-to-use, and versatile method for detecting cataracts by using smartphones without the use of bulky and expensive imaging devices. This methodcan be used for bedside telemedicine applications or in remote areas that have medical shortages. Previous smartphone-based cataract detection methods include texture feature analysis with 95 % accuracy, Gray Level Co-occurrence Matrix (GLCM) method with 89% accuracy, red reflex measurement method, and RGB color feature extraction method using cascade classifier with 90% accuracy. The accuracy of cataract detection in these studies is subject to changes in smartphone models and/or environmental conditions. However, our novel luminance-based method copes with different smartphone camera sensors and chroma variations, while operating independently from sensors’ color characteristics and changes in distances and camera angle. Clinical and Translational Impact—This study is an early/pre-clinical research proposing a novel luminance-based method of detecting cataract using smartphones for remote/at-home monitoring and telemedicine application.

In this paper, a system to regulate the traffic with the help of real-time vehicle density using Haar feature-based cascade classifier is proposed. Based on the traffic densities on all roads, the model allocates smartly the duration for which the green light glows. Image processing has been chosen for calculation of traffic density as cameras are very much cheaper than other devices such as sensors. The proposed model is constructed as follows: A Raspberry Pi that is connected to 4 sets of LEDs that represent the traffic lights is used. It does the process of monitoring the traffic density of each side and changes the signal according to the density in every direction. Therefore the side with higher traffic density will have green signal turned on for a longer duration. Thus it makes the traffic signaling system automatic, by providing the vehicle density measurement to the system. It makes sure that there are no difficulties in traffic handling.

Natural scene images often contain a lot of important texts, which carry the information we need, so it has important practical value to locate text information. This paper proposes a Cascade Classifier Text Localization (CCTL) algorithm. Firstly, a cascade classification algorithm based on Real AdaBoost is proposed to improve the accuracy of text localization. Secondly, a perceptual-based grouping algorithm is proposed to establish a perceptual organization framework. The use of adjacency and similar rules to group texts can improve the effect of text grouping. The proposed algorithm is verified and evaluated on the ICDAR dataset. At the same time, it is compared with other algorithms. The experimental results show that the proposed method has superiority in natural scene text detection.

Novel coronavirus pneumonia (NCP) has become a global pandemic disease, and computed tomography-based (CT) image analysis and recognition are one of the important tools for clinical diagnosis. In order to assist medical personnel to achieve an efficient and fast diagnosis of patients with new coronavirus pneumonia, this paper proposes an assisted diagnosis algorithm based on ensemble deep learning. The method combines the Stacked Generalization ensemble learning with the VGG16 deep learning to form a cascade classifier, and the information constituting the cascade classifier comes from multiple subsets of the training set, each of which is used to collect deviant information about the generalization behavior of the data set, such that this deviant information fills the cascade classifier. The algorithm was experimentally validated for classifying patients with novel coronavirus pneumonia, patients with common pneumonia (CP), and normal controls, and the algorithm achieved a prediction accuracy of 93.57%, sensitivity of 94.21%, specificity of 93.93%, precision of 89.40%, and F1-score of 91.74% for the three categories. The results show that the method proposed in this paper has good classification performance and can significantly improve the performance of deep neural networks for multicategory prediction tasks.

Early cataract detection is a non-common discussed topic for research, considering that it is a common disease, and is among the main causes of blindness. By detection on time, it is possible to detect cataracts in its initial stage. In this paper, the recently proposed method for cataract detection will be discussed. A tool is being used for cataract detection in OpenCV using python script. This is the new technique for this generation for cataract detection.

Face Recognition plays an important role in most of the disciplines in the recent world. One of the most common discipline is security and fraud detection. Facial alignment is a method of arranging the facial landmarks points on the face and those landmarks gives fine points for face recognition. Face detection and identification is important in the fraud detection. Therefore, detection of profile and semi-profile faces plays a vital role in security purpose.The face alignment by utilizing Hourglass model gives better accuracy for face recognition by using Haar- Cascade Classifier can be obtained by facial aligned dataset. The performance is measured by the accuracy rate and precision. It gives better results when compared to face recognition by using Support Vector Machine[SVM] and Principal Component Analysis[PCA]. The model gives alignment of facial points with 68 landmark points and the aligned data is sent as an input for the face recognition. The face alignment and recognition gives most accurate and precise results.

In order to enhance the recognition rate of small object in UAV reconnaissance images, this paper proposes a detection algorithm based on MobileNet-SSD, soft-NMS and haar-like features cascade classifier. First, input the batch reconnaissance images into the improved SSD detection network based on MobileNet v1 to obtain the initial detection frame; then use the soft-NMS algorithm to remove redundant duplicate frames to improve the detection accuracy; finally, all the detected images obtained after non-maximum suppression are input into the Haar-cascade classifier in batches, and are jointly verified with the detection results of the SSD model to further boost the recognition accuracy.