Comparative Study of Corner and Feature Extractors for Real-Time Object Recognition in Image Processing

Interpreting medical images is certainly a complex task which requires extensive knowledge [1]. According to Computer Aided Diagnosis (CAD) serves as a second opinion that will help radiologists in diagnosis and on the other hand Content-based Image Retrieval uses visual content to help users browse, search and retrieve similar medical images from a database based on the user’s interest [2-4]. The competency of the CBMIR system depends on feature extraction methods [5]. The textural features are very important to determine the content of a medical image. Textural features provide scenic depth, the spatial distribution of tonal variation, and surface orientation [6]. Therefore, this study seeks to compare and evaluate some of the hand-crafted texture feature extraction techniques in CBMIR. This is to help those concerned in enhancing CBIR systems to make informed decisions concerning the selection of the best textural feature extraction techniques. Since there is no clear indication of which of the various texture feature extraction techniques is best suited for a given performance metric when considering which of the techniques to choose for a particular study in CBMIR systems. The objective of this work, therefore, is to comparatively evaluate the performance of the following texture feature extraction techniques; Local Binary Pattern (LBP), Gabor Filter, Gray-Level Co-occurrence Matrix (GLCM), Haralick Descriptor, Features from Accelerated Segment Test (FAST) and Features from Accelerated Segment Test and Binary Robust Independent Elementary Features (FAST &BRIEF) using the metrics; precision, recall, f1-score, mean squared error (MSE), accuracy and time. These techniques are coupled with specific similarity measure to obtain results. The results showed that LBP, Haralick Descriptor, FAST, and GLCM had the best results in terms of (Precision and Accuracy), Time, F1-Score, and Recall respectively. LBP had 82.05% and 88.23% scores for precision and accuracy respectively. The following scores represent the performance of the Haralick descriptor, FAST, and GLCM models respectively; 0.88s, 38.7%, and 44.82%. These test scores are obtained from datasets ranging from 1k-10.5k. Aside from LBP outperforming the other 5 models mentioned, it still outperformed the following proposed models ' [7]’, ‘Tamura texture feature and wavelet transform combined with Hausdorff distance - [8]’, ‘ [9]’ in terms of (precision, accuracy, and recall) and (precision and recall) respectively and probably f1-score (since f1-score is the weighted average of precision and recall). It is believed that an ensemble of LBP, Haralick descriptors, and Support Vector Machine (SVM) can represent a robust system for both medical image retrieval and classification.

Optical Coherence Tomography (OCT) is widely used for endoscopic imaging in endoluminal organs because of its high imaging accuracy and resolution. However, OCT endoscopic imaging suffers from Non-Uniform Rotational Distortion (NURD), which can be caused by many factors, such as irregular motor rotation and changes in friction between the probe and the sheath. Correcting this distortion is essential to obtaining high-quality Optical Coherence Tomography Angiography (OCTA) images. There are two main approaches for correcting NURD: hardware-based methods and algorithm-based methods. Hardware-based methods can be costly, challenging to implement, and may not eliminate NURD. Algorithm-based methods, such as image registration, can be effective for correcting NURD but can also be prone to the problem of NURD propagation. To address this issue, we process frames by coarse and fine registration, respectively. The new reference frame is generated by filtering out the A-scan that may have the NURD problem by coarse registration. And the fine registration uses this frame to achieve the final NURD correction. In addition, we have improved the Features from Accelerated Segment Test (FAST) algorithm and put it into coarse and fine registration process. Four evaluation functions were used for the experimental results, including signal-to-noise ratio (SNR), peak signal-to-noise ratio (PSNR), mean squared error (MSE), and structural similarity index measure (SSIM). By comparing with Scale-invariant feature transform (SIFT), Speeded up robust features (SURF), Oriented FAST and Rotated BRIEF (ORB), intensity-based (Cross-correlation), and Optical Flow algorithms, our algorithm has a higher similarity between the corrected frames. Moreover, the noise in the OCTA data is better suppressed, and the vascular information is well preserved. Our image registration-based algorithm reduces the problem of NURD propagation between B-scan frames and improves the imaging quality of OCT endoscopic images.

Indonesia has a diversity of cultural arts, including art crafts. This craft art covers several fields including the arts of wood crafts, metal crafts, ceramic crafts, leather crafts, and textile crafts. In textile crafts, there are several types, such as weaving, knitting, embroidery, and batik. One area that has a characteristic of batik is Cirebon. Cirebon has a traditional batik center, namely Trusmi Batik. TrusmiBatik is a company that produces various kinds of batik, including mega mendungbatik motifs, paksinagaliman, patrankeris, patran kale, umbrella lion, barong lion, sunyaragian arum garden, banjarbalong, ayam alas, supiturang, nagaseba, sawat bride and groom cymbals, etc. The Trusmi batik company has problems on how to minimize artificial batik and how to distinguish original batik from artificial batik. One of technologiesthat can be used is Augmented Reality by using a QR Code (Quick Response) as a sign especially those which can find out whether the batik is original or imitation. The Feature From Accelerated Segment Test is a high-speed interest point detection algorithm based on pixel considerations in circular areas around interest points. Using the FAST algorithm can be used to detect markers so that they can distinguish original batik from artificial batik. Keywords: Trusmi Batik, QR Code, Augmented Reality, FAST.

Detecting the changes within a single image is significant for applications like planogram compliance where objects of same type are arranged in a sequence. This article discusses on detection of such objects which differ in color, texture and shape from other objects in the given image. In this proposed algorithm, the given image is converted to HSV color space and it is partitioned into regions; Fast Discrete Curvelet Transform (FDCT) coefficients are extracted. In order to identify changed regions in the given image, salient points are detected using the Features from Accelerated Segment Test (FAST) corner detection algorithm. It has been observed that the value channel content of the HSV image is much significant in extracting useful features for change detection. Experimental results show an efficacy of 87% in detecting changes in a single row of a shelf image having same items arranged in a linear fashion.Keywords(FAST) corner detectionHSV color spaceChange detectionFDCTImage processing

The classic feature point image stitching algorithm is time-consuming in feature extraction, and an image sparse matrix method is proposed to determine the feature points. This method first uses Laplace operator to extract the image gradient and set the threshold to obtain the sparse matrix of image segmentation, then using features from accelerated segment test (FAST) detection algorithm for feature point. Finally, the speeded-up robust features (SURF) descriptor is given to increase the stability of the feature points. This method solves the problem of time-consuming feature extraction, making the real-time image mosaic possible. In the process of image mosaic, this paper presents a method using Kalman filter to predict the overlap region, thus reducing the computational complexity. In addition, because there is always a mismatch between feature points matching, this paper proposes a method using perceptual hashing to refine the matching pair, which solves the error caused by mismatching for the next image registration and improves the registration accuracy. Experimental results show that the proposed algorithm improves the speed and precision of image mosaic.

In this study, we designed an autonomous mobile robot based on the rules of the Federation of International Robotsoccer Association (FIRA) RoboSot category, integrating the techniques of computer vision, real-time image processing, dynamic target tracking, wireless communication, self-localization, motion control, path planning, and control strategy to achieve the contest goal. The self-localization scheme of the mobile robot is based on the algorithms featured in the images from its omni-directional vision system. In previous works, we used the image colors of the field goals as reference points, combining either dual-circle or trilateration positioning of the reference points to achieve selflocalization of the autonomous mobile robot. However, because the image of the game field is easily affected by ambient light, positioning systems exclusively based on color model algorithms cause errors. To reduce environmental effects and achieve the self-localization of the robot, the proposed algorithm is applied in assessing the corners of field lines by using an omni-directional vision system. Particularly in the mid-size league of the RobotCup soccer competition, selflocalization algorithms based on extracting white lines from the soccer field have become increasingly popular. Moreover, white lines are less influenced by light than are the color model of the goals. Therefore, we propose an algorithm that transforms the omni-directional image into an unwrapped transformed image, enhancing the extraction features. The process is described as follows: First, radical scan-lines were used to process omni-directional images, reducing the computational load and improving system efficiency. The lines were radically arranged around the center of the omni-directional camera image, resulting in a shorter computational time compared with the traditional Cartesian coordinate system. However, the omni-directional image is a distorted image, which makes it difficult to recognize the position of the robot. Therefore, image transformation was required to implement self-localization. Second, we used an approach to transform the omni-directional images into panoramic images. Hence, the distortion of the white line can be fixed through the transformation. The interest points that form the corners of the landmark were then located using the features from accelerated segment test (FAST) algorithm. In this algorithm, a circle of sixteen pixels surrounding the corner candidate is considered and is a high-speed feature detector in real-time frame rate applications. Finally, the dual-circle, trilateration, and cross-ratio projection algorithms were implemented in choosing the corners obtained from the FAST algorithm and localizing the position of the robot. The results demonstrate that the proposed algorithm is accurate, exhibiting a 2-cm position error in the soccer field measuring 600 cm2 x 400 cm2.

With the rapid development of artificial intelligence and computer vision, feature point detection algorithms have emerged as a significant area of research, particularly in applications such as SLAM (Simultaneous Localization and Mapping) and image registration. However, the FAST (Features from Accelerated Segment Test) detection method employed in the ORB (Oriented FAST and Rotated BRIEF) algorithm exhibits limitations in detection accuracy. This paper proposes a feature point detection method based on sequential neural networks and knowledge distillation, aiming to enhance the performance of the FAST algorithm. Firstly, the sequential nature of the data points in the FAST algorithm is analyzed, leading to the development of a binary classification and regression model using a sequential neural network. Considering the significant computational overhead of the model, knowledge distillation techniques are introduced to optimize the efficiency of the neural network. Secondly, the constructed model undergoes performance testing and evaluation, with a detailed comparison against the traditional FAST method. Finally, this paper discusses the future prospects of feature point detection, suggesting potential directions for further optimization.

Due to COVID-19's rapid spread, millions of people around the world have been affected and there has been extensive destruction. One of the most effective ways of preventing its spread and transmission during the pandemic period wearing of a mask and was required then in most public areas. As a result, this necessitate the use of programmed real-time mask detection devices in place of manual reminders. Face mask detection requires real-time processing of a huge amount of data with constrained processing resources, hence local descriptors that are quick to calculate, quick to match, and cheap to store are highly sought after. To achieve improved matching and reduction in memory use and computational complexity, the study offers a combination of Features from Accelerated Segment Test (FAST) corner detector with Histogram of Oriented Gradient (HOG) feature descriptor to allow faster matching and minimize memory usage and computation cost. The features obtained were then classified into face mask present and face mask absent using SVM, NB and CNN. Results obtained gives an accuracy of 99.41% which was higher than that previous results of 99.27% and 95% accuracy. Furthermore, it took the system only 48secs to extract the features obtained from face for training and testing. This outcome confirmed the suitability of the suggested method for real-time face mask detection.

Object recognition is a key research area in the field of image processing and computer vision, which recognizes the object in an image and provides a proper label. In the paper, three popular feature descriptor algorithms that are Scale Invariant Feature Transform (SIFT), Speeded Up Robust Feature (SURF) and Oriented Fast and Rotated BRIEF (ORB) are used for experimental work of an object recognition system. A comparison among these three descriptors is exhibited in the paper by determining them individually and with different combinations of these three methodologies. The amount of the features extracted using these feature extraction methods are further reduced using a feature selection (k-means clustering) and a dimensionality reduction method (Locality Preserving Projection). Various classifiers i.e. K-Nearest Neighbor, Naive Bayes, Decision Tree, and Random Forest are used to classify objects based on their similarity. The focus of this article is to present a study of the performance comparison among these three feature extraction methods, particularly when their combination derives in recognizing the object more efficiently. In this paper, the authors have presented a comparative analysis view among various feature descriptors algorithms and classification models for 2D object recognition. The Caltech-101 public dataset is considered in this article for experimental work. The experiment reveals that a hybridization of SIFT, SURF and ORB method with Random Forest classification model accomplishes the best results as compared to other state-of-the-art work. The comparative analysis has been presented in terms of recognition accuracy, True Positive Rate (TPR), False Positive Rate (FPR), and Area Under Curve (AUC) parameters.

Mobile terminal gesture recognition is an extreme challenge, not only because its limited computing resource make it complicated to identify feature points but also the complex background can easily affect the recognition result. This study proposes a gesture recognition method based on improved features from accelerated segment test (FAST) corner detection. First, in order to eliminate the effects of complex background and light, the intersection of the two frame images is obtained through background subtraction and the multi‐colour space to realise the detection of the hand. Second, in order to improve the performance of the algorithm, an improved FAST corner detection method combined with the back propagation neural network (BPNN) is proposed in accordance with the characteristics of fingertips. Subsequently, the feature points are screened by method of non‐maximum suppression. Finally, gesture recognition is realised by matching feature points. Experimental results illustrate that this method has strong anti‐interference ability in complex background, and it is good at performance.

In recent years, Ambient Intelligence (AmI) has attracted increased attention within the advanced technology industry in an effort to modernize and develop a more intelligent and reliable information system. Technologies to detect a specific object in images are expected to further expand to wide range of applications, such as car detection functions for Intelligent Transport System (ITS) and other systems. Computer vision and pattern recognition are emerging fast and will continue to grow together with local feature detection methods. In this paper, we propose an object detection and tracking system which is based on Scale Invariant Feature Transform (SIFT) and Speed Up Robust Features (SURF) feature extraction methods. From the evaluation results, we observe that the accuracy of matched keypoints of SURF algorithm are higher than SIFT.

The classic feature point image stitching algorithm is time-consuming in feature extraction, and an image sparse matrix method is proposed to determine the feature points. This method first uses Laplace operator to extract the image gradient and set the threshold to obtain the sparse matrix of image segmentation, then using features from accelerated segment test (FAST) detection algorithm for feature point. Finally, the speeded-up robust features (SURF) descriptor is given to increase the stability of the feature points. This method solves the problem of time-consuming feature extraction, making the real-time image mosaic possible. In the process of image mosaic, this paper presents a method using Kalman filter to predict the overlap region, thus reducing the computational complexity. In addition, because there is always a mismatch between feature points matching, this paper proposes a method using perceptual hashing to refine the matching pair, which solves the error caused by mismatching for the next image registration and improves the registration accuracy. Experimental results show that the proposed algorithm improves the speed and precision of image mosaic.

A development of automatic location identification and tracking system for visually impaired/ challenged person is a very challenging task in an indoor environment. In this paper, the comprehensive study of different feature detection and matching techniques namely, Minimum Eigenvalue (MinEigen) algorithm, Harris–Stephens (Harris) algorithm, Speeded Up Robust Features (SURF), Features from Accelerated Segment Test (FAST), Binary Robust Invariant Scalable Keypoints (BRISK) and Maximally Stable Extremal Regions (MSER) is presented. These algorithms are employed to detect and match the features of an image and retrieve the best matched image. Based on our experiments, we compare those algorithms on parameters such as sum of square difference (SDD), precision, recall, number of detected, matched features and processing time. Empirically, we have found that SURF algorithm produce minimum SSD score to achieve best matching. The MSER and MinEign algorithm extracts high and low number of features respectively. In respect of processing time, BRISK takes maximum and FAST method takes minimum time when compared to the other algorithms.

Presently, cognitive Internet of Things (CIoT) with cloud computing (CC) enabled intelligent healthcare models are developed, which enables communication with intelligent devices, sensor modules, and other stakeholders in the healthcare sector to avail effective decision making. On the other hand, Alzheimer disease (AD) is an advanced and degenerative illness which injures the brain cells, and its earlier detection is necessary for suitable interference by healthcare professional. In this aspect, this paper presents a new Oriented Features from Accelerated Segment Test (FAST) with Rotated Binary Robust Independent Elementary Features (BRIEF) Detector (ORB) with optimal artificial neural network (ORB-OANN) model for AD diagnosis and classification on the CIoT based smart healthcare system. For initial pre-processing, bilateral filtering (BLF) based noise removal and region of interest (RoI) detection processes are carried out. In addition, the ORB-OANN model includes ORB based feature extractor and principal component analysis (PCA) based feature selector. Moreover, artificial neural network (ANN) model is utilized as a classifier and the parameters of the ANN are optimally chosen by the use of salp swarm algorithm (SSA). A comprehensive experimental analysis of the ORB-OANN model is carried out on the benchmark database and the obtained results pointed out the promising outcome of the ORB-OANN technique in terms of different measures.

The recognition of human activities through images represents a fundamental research domain in computer vision and pattern recognition, with practical applications in human–computer interaction, video analysis, and surveillance. This research objective is to develop an enhanced neurocomputing network for humans using Deep Neural Networks (DNNs) for precise image-based behaviour classification. Our proposed DNN-based framework unites several preprocessing and feature extraction methods to accomplish this goal. The system begins with Hue, Saturation, and Value (HSV) colour processing to enhance image visibility, followed by Gaussian filtering for noise reduction. The statistical method performs silhouette extraction, and the feature extraction process utilises Local Intensity Order Pattern (LIOP) and Features from accelerated segment test (FAST) algorithms together. Feature discrimination is enhanced through the application of fuzzy optimisation techniques. The optimised features are processed by a DNN, which classifies human activities. The proposed framework demonstrates high effectiveness through its recognition performance, which achieved 94% accuracy on BIT interaction data and 88.25% accuracy on SBU interaction data. This work presents an advanced human activity recognition system that shows promise for real-world applications, such as surveillance systems, video analytics, and interactive technologies, enabling more precise analysis of human behaviour.