HOG Method Research Articles

Elderly Signature Identification Using HOG Method and K-Nearest Neighbors Classifier

Signature is used to legally approve an agreement, treaty, and state administrative activities. Identification of the signature is needed to ensure the ownership of a signature so that things do not happen that harm the owner of the signature, such as forgery. In this study, the authors identified signatures for people over 50 years of age. The obtained signature is scanned to produce a signature image, the image is pre-processed so that it is ready for feature extraction that can characterize the signature image. The HOG method is used to perform feature extraction to produce a dataset that has 4,536 feature vectors for each signature image. The K-NN classification method is used to identify the signature image, the highest accuracy is obtained when using the value of K = 3, 4 which is 98.6%. The value of K = 7, 8, 10 obtained the lowest accuracy value of 96%. The decrease in accuracy from the highest to the lowest which is not significant shows that the dataset from feature extraction using the HOG method can characterize one signature ownership with other signatures quite well. The HOG method obtains a large feature vector for each signature, so it is recommended to optimize the feature extraction to reduce the feature size and speed up computing.

Food Packaging Target Location and Recognition Technology Based on Machine Vision

In the 21st century, people have an increased demand for food in their lives. Packaging manufacturers product a mass of outer packaging in order to satisfy the demand of food before they access to market for sale. What’s more, food packaging needs to be identified and tested, sorted, and shipped to the food producers. Machine vision has many advantages over than human eyes: high detection efficiency, high accuracy, no controlled by fatigue and so on. So, article takes the outer packaging of food as the research object and research on target location and recognition technology of food packaging based on machine vision was carried out. This article has certain practical application value. The research content is as follows. Firstly, the function of edge detection is recommended. Then three categories of edge detection are introduced. Through the five edge detections are introduced and the experimental results are compared, Canny algorithm is the best scheme. Secondly, this section learns the function and advantages of the HOG. The food packaging image is exacted features with HOG method. Thirdly, this section also introduces the function and the advantage of SIFT. The food packaging image features are taken experiments and recognized. Fourthly, this section recommends the experimental procedure and advantages of single target image location technique. In the end, this article concludes all the experiments and points out the shortcomings of the previous mentioned experimental steps.

Parameter Optimization of Support Vector Machine using River Formation Dynamic on Brain Tumor Classification

Brain tumor is a condition that can interfece with brain function due to abnormal cell growth in the brain. MRI is used as a diagnostic tool when a patient has a brain tumor. The images obtained through MRI will be analyzed by the doctor to determine the type of tumor. Therefore, it is necessary to have a system that can classify tumor types based on MRI images. The image will be extracted using the HOG method, then classified using SVM. Certain measures can be taken to improve SVM performance, such as optimizing its parameters. This research develops a system that uses a novel combination, the SVM with the River Formation Dynamic (RFD) algorithm. RFD is being used to optimize parameters of SVM (C and gamma). The basic idea of RFD is to imitate the movement of water droplets flowing from high to low areas. This research compares the accuracy produced by SVM with the accuracy produced by SVM-RFD. The result is that SVM-RFD provides the better accuracy than only using SVM. The accuracy result by SVM on the MRI dataset is 74.37%. When we compared it with SVM-RFD, the accuracy increased by 13.19% to 87.56%. Further work will be carried out on the implementation of RFD on other SVM parameters to find other parameter combinations that can improve the accuracy of SVM.

Signature Identification using Digital Image Processing and Machine Learning Methods

Signature is used to legally approve an agreement, treaty, and state administrative activities. Identification of the signature is required to ensure ownership of a signature and to prevent things like forgery from happening to the owner of the signature. In this study, data signatures were obtained from 25 people over the age of 50. The signers provided 20 signatures and were free to choose the stationery used to write the signature on white paper. The total data obtained in this study was 500 signature data. The obtained signature was scanned to create a signature image, which was then pre-processed to prepare it for feature extraction, which can characterize the signature images. The HOG method was used to extract features, resulting in a dataset with 4,536 feature vectors for each signature image. To identify the signature image, the classification methods SVM, Decision Tree, Nave Bayes, and K-NN were compared. SVM achieved the highest accuracy, which is 100%. When K=5, the K-NN method achieved a fairly good accuracy of 97.3%. Meanwhile, Naive Bayes and Decision Tree achieved accuracy significantly lower than K-NN (61%). Because the HOG method produced a large feature vector for each signature, it is recommended that important features that represent signatures be optimized or extracted to produce smaller features to speed up computation without sacrificing accuracy, and that the HOG method be compared to other extraction feature methods to obtain a better model in future research.

Automating the customer verification process in a car sharing system based on machine learning methods

Convenient and accurate verification of the user of a car sharing system is one of the key components of the successful functioning of the car sharing system as a whole. The machine learning-based KYC (Know your customer) process algorithm makes it possible to improve the accuracy of customer data validation and verification. This makes it possible to eliminate possible losses and reputational losses of the company in case of unforeseen situations while using the client's car sharing services. The object of this study is to find a solution to the problem of user verification in a car sharing system based on the KYC process using deep learning methods with a combination of OCR (Optical Character Recognition) methods. The statement of the user verification problem in the car sharing system was formalized and the key parameters for the KYC process have been determined. The algorithm of the KYC process was constructed. The algorithm includes six successive stages: separating a face in the photograph, comparing faces, checking documents and their validity period, establishing and recognizing ROI (region of interest), formulating a verification decision. To separate the face in the client's photograph and compare faces, methods based on deep learning, as well as the quick HoG method (Histogram of oriented gradients), were considered and implemented. Verification of these methods on a test dataset, which includes images of documents of two thousand clients, showed that the recognition accuracy was 91 % according to Jaccard's metric. The average time of face separation using the HoG method was 0.2 seconds and when using trained models – 3.3 seconds. Using a combination of ROI and ORC separation methods makes it possible to significantly improve the accuracy of verification. The proposed client verification algorithm is implemented as an API on an ML server and integrated into the car sharing system.

Diagnosis of anomalies based on hybrid features extraction in thyroid images.

Diagnosing benign and malignant glands in thyroid ultrasound images is considered a challenging issue. Recently, deep learning techniques have significantly resulted in extracting features from medical images and classifying them. Convolutional neural networks ignore the hierarchical structure of entities within images and do not pay attention to spatial information as well as the need for a large number of training samples. Capsule networks consist of different hierarchical capsules equivalent to the same layers in the convolutional neural networks. We propose a feature extraction method for ultrasound images based on the capsule network. Then, we combine those deep features with conventional features such as Histogram of Oriented Gradients and Local Binary Pattern together to form a hybrid feature space. We increase the accuracy percentage of a support vector machine (SVM) by balancing and reducing the data dimensions of samples. Since the SVM provides different training kernels according to the sample distribution method, the extracted textural features were categorized using each of these kernels to obtain the result. The parameters of classification evaluation using the researcher-made model have outperformed the other methods in this field. Experimental results showed that the combination of HOG, LBP, and CapsNet methods outperformed the others, with 83.95% accuracy in the SVM with a linear kernel.

Artificial Intelligence Based Social Distance Monitoring in Public Areas

COVID-19 is an infectious disease caused by a newly discovered coronavirus called SARS-CoV-2. There are two ways of contamination risk, namely spreading through droplets or aerosol-type spreading into the air with people's speech in crowded environments. The best way to prevent the spread of COVID-19 in a crowd public area is to follow social distance rules. Violation of the social distance is a common situation in areas where people frequently visit such as hospitals, schools and shopping centers. In this study, an artificial intelligence-based social distance determination study was developed in order to detect social distance violations in crowded areas. Within the scope of the study, a new dataset was proposed to determine social distance between pedestrians. The YOLOv3 algorithm, which is very successful in object detection, was compared with the SSD-MobileNET, which is considered to be a light weighted model, and the traditionally handcrafted methods Haar-like cascade and HOG methods. Inability to obtain depth information, which is one of the biggest problems encountered in monocular cameras, has been tried to be eliminated by perspective transformation. In this way, the social distance violation detected in specific area is notified by the system to the relevant people with a warning.

Moving Object Tracking Based on Sparse Optical Flow with Moving Window and Target Estimator.

Moving object detection and tracking are technologies applied to wide research fields including traffic monitoring and recognition of workers in surrounding heavy equipment environments. However, the conventional moving object detection methods have faced many problems such as much computing time, image noises, and disappearance of targets due to obstacles. In this paper, we introduce a new moving object detection and tracking algorithm based on the sparse optical flow for reducing computing time, removing noises and estimating the target efficiently. The developed algorithm maintains a variety of corner features with refreshed corner features, and the moving window detector is proposed to determine the feature points for tracking, based on the location history of the points. The performance of detecting moving objects is greatly improved through the moving window detector and the continuous target estimation. The memory-based estimator provides the capability to recall the location of corner features for a period of time, and it has an effect of tracking targets obscured by obstacles. The suggested approach was applied to real environments including various illumination (indoor and outdoor) conditions, a number of moving objects and obstacles, and the performance was evaluated on an embedded board (Raspberry pi4). The experimental results show that the proposed method maintains a high FPS (frame per seconds) and improves the accuracy performance, compared with the conventional optical flow methods and vision approaches such as Haar-like and Hog methods.

Traffic sign recognition based on HOG feature extraction

The substantial increase in the number of motor vehicles in recent years has caused many traffic safety problems and has aroused widespread concern. As the basis of intelligent vehicle environment perception and a necessary condition for realizing the functions of assisted driving system, traffic sign recognition is of great significance for realizing automatic driving of vehicles, improving intelligent transportation systems, and promoting the development of smart cities.This paper mainly identifies traffic signs, using histogram of gradient feature extraction method. The image is collected and preprocessed by a vision sensor. The color threshold segmentation method and morphological processing are used to reduce the interference of the background area and enhance the contour of the sign area. Finally, HOG method is used to collect the gradient of each pixel point in the cell unit or the direction histogram of the edge to identify traffic signs. Through MATALB simulation, it is obtained that the HOG image feature extraction method has high accuracy, small error and short recognition time, which shows the effectiveness of the algorithm.

A New Pedestrian Feature Description Method Named Neighborhood Descriptor of Oriented Gradients

In view of the issue of the mechanism for enhancing the neighbourhood relationship of blocks of HOG, this paper proposes neighborhood descriptor of oriented gradients (NDOG), an improved feature descriptor based on HOG, for pedestrian detection. To obtain the NDOG feature vector, the algorithm calculates the local weight vector of the HOG feature descriptor, while integrating spatial correlation among blocks, concatenates this weight vector to the tail of the HOG feature descriptor, and uses the gradient norm to normalize this new feature vector. With the proposed NDOG feature vector along with a linear SVM classifier, this paper develops a complete pedestrian detection approach. Experimental results for the INRIA, Caltech-USA, and ETH pedestrian datasets show that the approach achieves a lower miss rate and a higher average precision compared with HOG and other advanced methods for pedestrian detection especially in the case of insufficient training samples.

Texture Analysis of Breast Cancer via LBP, HOG, and GLCM techniques

Breast cancer is a prevailing reason for death, and it is a particular kind of tumor that is popular among ladies across the world. Till presently, there is no efficient method to stop the appearance of the breast tumor. Accordingly, early detection is the first stage in the diagnosis of breast tumors and reduces mortality. Screening Mammography is the most effective technique for early detection of breast tumors. Great experience and large practices of specialists are wanted when examining breast tissue in a mammogram. In this work, feature extraction techniques are offered as methods to decrease false-positive that occur in breast diagnosis. Mini-MIAS database used to evaluate these approaches. LBP, HOG, and GLCM are feature extraction techniques used for analyzing mass tissue and extract features from the ROI. Contrast, energy, correlation, and homogeneity are used as features properties. These features utilized as the input to the different classifiers which achieved the best results. To improve the diagnosis ability, “contrast limited adaptive histogram equalization” utilized as a preprocessing system. The best results gained in this work by LBP method and logistic regression classifier at ROI (30×30) where the accuracy 92.5%. The HOG method achieved the best results with the SVM classifier where accuracy 90% at ROI (30×30). GLCM provides the best results with the KNN classifier where the accuracy 89.3% at ROI (30×30). The greatest accuracy reached in the case of ROI (30×30) in all feature extraction methods that used in this work.

A HYBRID MACHINE LEARNING TECHNIQUE FOR EARLY PREDICTION OF ALZHEIMER’S DISEASE

In this paper, a new hybrid machine learning technique is proposed to detect the Alzheimer’s disease at a very early stage from MRI images, so that the neurodegenerative disease can be controlled and thereby preventing its progression at a very early stage. In this technique, the feature extraction is carried out using SURF, FAST, BRISK, Harris, Min Eigen and HOG methods and feature selection is carried out using Principal Component Analysis and fisher methods for early prediction of various stages of Alzheimer's disease. An analysis of the proposed method is done by combining it with Support Vector Machine, k Nearest Neighbor and Naive Bayes classifiers and performance parameters are evaluated. From the experimental results, the proposed model yields high accuracy rate and is found to be superior than the methods which uses single feature extraction and feature selection methods which are developed for the prediction and classification of Alzheimer’s disease.

Accurate classification of cherry fruit using deep CNN based on hybrid pooling approach

Abstract The most important quality parameter of a product is its nutritional value, but marketability of agricultural products depends primarily on the overall appearance and shape of the products. This study was carried out with the aim of developing cherry fruit packing methods and thus reducing waste and increasing its exportability and marketability. Therefore, the purpose of research was to use the improved Convolutional Neural Network (CNN) algorithm to detect the appearance of cherries and provide an efficient system for their grading. In order to identify and classify images cherry on two classes (regular and irregular shaped) was prepared. After preprocessing the images, the proposed method utilized its ability to improve generalization in the CNN through a combination of max pooling and average pooling techniques, to grade cherries. In order to compare the proposed method (CNN) with HOG and LBP methods, the properties of the images extracted by KNN, ANN, Fuzzy and Ensemble Decision Trees (EDT) algorithms were categorized. The proposed method based on hybrid pooling is also compared with CNN with baseline pooling method such as average pooling. Comparisons based on the results of simulation demonstrate the superiority of the proposed improved CNN over other methods presenting an accuracy of 99.4 %. Therefore, the CNN and image processing methods are effective in managing the marketability and exportability of the cherry fruit and can replace the traditional methods applied for grading cherries.

Development and testing of a conveyor for detecting various types of vehicles when transporting agricultural products from the field

Research purpose: development of a technology that will detect various types of vehicles in the image. To achieve this purpose it is necessary to solve the following objectives: Highlighting requirements for the technology being developed; Development of the technology for finding an object in an image; The choice of methods and algorithms for the allocation of objects for the development of the technology; The choice of methods and algorithms that allow detecting objects of a certain class; Implementation of the developed technology in the software system. The scientific novelty of the work lies in the application of previously known methods, which have been shown to be effective in other studies, to a new object of study, namely, to detect in images vehicles of various classes used for the transportation of agricultural goods. The practical value of the expected results lies in creating the software based on the developed technology, which makes it possible to detect vehicles of a certain class with a high degree of probability in a static image. The development of such software will automate part of the business processes and reduce labor costs. The following methods are used as the main ones in the developed technology: the HOG method (histograms of oriented gradients) and the support vector method (SVM).

Intelligent Home Automation System for Disabled People

This paper presents the intelligent Home Automation System (HAS) for disabled people, since still now facilities for disabled people are insufficient around the world. Numerous researchers developed different technologies considering Bluetooth technology, ZigBee system, and Wifi technology. However, these topologies lack efficient support for disabled people. Therefore, to engage the disabled people with the modern technology and make their life safe, secure and comfortable, authors have introduced the Rasberry Pi 3 Model B with Blynk application which is able to control the home appliances from their smart android phone. In this research, the motion of the intruder has been considered as the key parameters and HOG method is used to detect the motion. Experimental validation of the proposed model implies that this process is more secure and user-friendly for disabled people compare to other existing technology. Therefore, the main contribution of this research is to develop a secured automated system, which will enable the disabled people to control the home appliances and thus overcome the limitation of the existing technology.

Research on License Plate Image Segmentation and Intelligent Character Recognition

With the accumulation of people’s wealth and the improvement of purchasing power, more and more people are buying cars as a means of travel. Walking and cycling of the past have now become a car trip. License plate recognition technology is especially important in intelligent transportation systems. It has been widely used in large shopping malls or supermarket parking lots, highway toll stations, speeding violation supervision and other fields. However, the accuracy and efficiency of license plate image recognition are insufficient. To solve the above problems, we propose a license plate character recognition method based on local HOG and layered LBP feature fusion from the perspectives of image pre-processing, license location, characters’ segmentation and recognition. First, pre-processing the license image area by highlighting the license plate image; then, the license plate is positioned based on wavelet decomposition and brightness moment; next the tilted license plate image is corrected, the license plate frame is adjusted, and characterization is performed by using the improved projection method based on the fact that the projection of the character is a single peak or a double peak. Finally, the local HOG and hierarchical LBP feature fusion methods are used to identify the license characters. The results show that the license plate’s character recognition rate of the proposed method reaches 99.71%, and the time taken is small. This not only improves the character recognition rate, but also saves recognition time. The results show that the method has important practical significance in license plates’ recognition.

Underwater Image Enhancement using HOG and SIFT Method

Capturing underwater images can be considered as form of art and collecting data of underwater environment which plays major role in the field of marine biology research, marine zoology and ecology study. It also plays a significant role in scientific missions such as analyzing marine life species, taking census of population and monitoring underwater biological environment. Underwater imaging also offers many other attractions such as aquatic plants and animals, different types and species of fishes, shipwrecks, coral reef and beautiful landscapes. But the captured images lack contrast and are hazy due to absorption and scattering. Thus the quality of the image gets deteriorated. The degraded underwater images can be enhanced using different methods. This research work proposed the white balancing technique and image blending process to improve the degraded images. The image blending can be done using the methods of feature matching. The features can be matched using HOG and SIFT methods. The proposed technique is implemented in MATLAB and results are analyzed in the terms of PSNR and MSE.

Development of High Recognition Rate FKP System using Fractional Cuckoo Search Optimization Method

Researchers have recently found that the finger knuckle print is an emerging and promising biometric identifier. In order to get better performance in a biometric system requires fusion of multiple instances within a modality. This paper makes use of fractional theory to propose an efficient feature level of fusion scheme named as a Fractional Cuckoo Search algorithm (FCS) for finger knuckle images. In this work both left index and right index features of an individual are extracted by using HOG method. These feature vectors are fused by optimal weight score selection, which finds out the weights from the proposed fractional cuckoo search algorithm. Thus the performance of the proposed system has been evaluated with the help of multi- layered SVM classifier in terms of False Acceptance Ratio (FAR), False Rejection Ratio (FRR) and Accuracy. From the experimental results, our proposed optimization system obtains a better accuracy of 98.97%.

Traffic sign detection using histogram of oriented gradients and max margin object detection

Traffic signs are important markers in two-wheeled and four-wheeled vehicles. However, there is a change in direction or arrangement on the road that cannot be opened on a map which can cause incorrect information, which can cause traffic jams. In this journal the author uses a camera mounted on a car that provides a solution for drivers who issue problems that occur on the road that show directions or arrangements that are not directly updated using the HOG and MMOD methods. HOGs and MMODs are methods that can refute objects well, and move images and will be recognized immediately. The information received can be sent directly to an electronic map so that it can be accessed automatically by the driver’s information and assistance and other information finds the right path, so that it can help the driver and can avoid traffic jams.

A hybrid fuzzy framework for face detection and recognition using behavioral traits

Face recognition integrated with emotions and behavior can be a salient feature for identification of antisocial/criminal people. In this paper, a hybrid Fuzzy based Behavior Prognostic System for Disparate traits (FBPSDs) has been proposed. FBPSDs can be used in an emergency situation to identify and extract mischievous person(s) involved in any swindling act. The technique uses face recognition to identify the contextual faces. Further, fuzzy rules have been formulated to extricate more features for the culprit identification. The proposed framework has three stages: face detection, feature extraction and behavior analysis. Face detection for the proposed FBPSDs is further implemented using three techniques Haar Cascade, Skin color based and HOG methods on Labeled Faces in the Wild (LFW) and Facial dataset (Face 94, Face 95 and Face 96). Testing accuracy of 91%, 50%, 92% (LFW) and 91%, 60%, 91.5% (facial dataset) is achieved. Result analysis and evaluation prove that HOG method is the most acceptable and appropriate method for face detection.