Key Feature Points Research Articles

Research on Improving the Reliability of Vision Positioning System by Deep Learning

Vision positioning system is widely used in robotics, industrial testing, servo control and other autonomous systems. Precise object positioning needs to accurately match the feature key points of the object, so as to solve the object position and attitude precisely. Simple image or feature matching method cannot meet the requirements of both reliability and accuracy for positioning system. Recently, deep learning has aroused widespread concern, Caffe as a framework for deep learning, performs particularly well in image recognition. In this paper, we propose a method of using convolutional neural networks to recognize target in complex scenes, and then to obtain the information of the target position by precise image matching. The experimental results show that the proposed method effectively eliminates the false target recognition and improves the image matching accuracy.

Reducing dense local feature key-points for faster iris recognition

Iris recognition has gained much attention in research and commercialization during the last decade. For a large population, the matching time of iris biometric system is much slower than the requirement. More the enrolled population size, higher the identification delay. To combat the delay without compromising accuracy of the system, the proposed method introduces a density-based spatial clustering and key point reduction to be applied on Phase Intensive Local Pattern (PILP) based dense feature extracted from the image. The reduction technique can also work with other dense local features. The reduction method is investigated whether it harms the accuracy of iris biometric system with respect to PILP. Widely used databases: BATH and CASIAv3 are used for experimentation. The technique is found successful in reducing representative key-points, thereby speeding up the match time up to five times. This improvement in 1:1 match-time is significant, and becomes more meaningful in identification for a large population.

Feature Detection and Description in Remote Sensing Images using a Hybrid Feature Detector

Feature detection plays a crucial role in image registration. There exists quite a few feature detection algorithms in literature like BRISK, FAST, SURF etc [14]. Each of these algorithms has its own advantages and disadvantages. BRISK is rotation and scale invariant, but it takes more time to detect the feature points. On the other hand FAST, as the name suggests, takes less time to detect the key points, but it is not scale invariant. To overcome the demerits of BRISK and FAST feature detection algorithms, this paper proposes a hybrid feature detection algorithm, which consumes less time to detect the feature key points and it is also rotation and scale invariant. This paper also focus on a comparative analysis of BRISK, FAST and proposed algorithm in terms of time to detect feature points. This paper has taken five feature key points in every Remote-sensing images and also deals with feature detection using the above three algorithms. It can be observed from the results and tables that in case of hybrid feature detector, it takes less time to detect five feature points.

HASHING OF STRUCTURAL DESCRIPTIONS AT BUILDING OF THE CLASS IMAGE DESCRIPTOR, COMPUTING OF RELEVANCE AND CLASSIFICATION OF THE VISUAL OBJECTS

The problem of classification of visual objects on the space of the descriptor features of the image key points is solved with application of data hashing to the image description. The efficiency of application of options of the hashing functions is analyzed at building of the class descriptor as a generalized image of the etalon and at determining the level of relevance for the binary descriptions. Experimental estimation of the time for computing the hash-functions and determining the relevance for the hashed images as compared to the traditional voting approach is performed by means of the program simulation. The efficiency of application of hash-processing of data is confirmed in terms of a substantial increase of the performance parameter.

Feature Extracting from Video Encoded for Searching Using Improved Methods

There is a need for the transportation of visual content from a particular node to a computation unit, on a network limited by bandwidth for distributed visual analysis applications, like mobile visual search or Visual Sensor Networks (VSNs). The proposed system uses the video compression technique to transfer information over a communication network. The system is used for key point detection, descriptor calculation and feature matching. The proposed system provides key points to the server as side data to retrieve only the relevant points from the encoded video. In the proposed encoding approach, the User chooses the video from the dataset, extracts frames from the video and then extracts the key point of the selected frame and encodes the string or the key point on the extracted frame to then reconstruct or compress the video and upload it on the server. For decoding, whenever the user enters the search query on the browser, the server receives the search query. The server then selects the video, which is stored on the server, and extracts frames from the video, then key point or encoded data from the selected frame using decoding, and then compares the search query and the decoded data. If it is true, the video is sent to the user. The results show that the proposed approach offers significantly improved feature matching and image retrieval performance at given bitrates.

A novel Gravity-FREAK feature extraction and Gravity-KLT tracking registration algorithm based on iPhone MEMS mobile sensor in mobile environment.

Based on the traditional Fast Retina Keypoint (FREAK) feature description algorithm, this paper proposed a Gravity-FREAK feature description algorithm based on Micro-electromechanical Systems (MEMS) sensor to overcome the limited computing performance and memory resources of mobile devices and further improve the reality interaction experience of clients through digital information added to the real world by augmented reality technology. The algorithm takes the gravity projection vector corresponding to the feature point as its feature orientation, which saved the time of calculating the neighborhood gray gradient of each feature point, reduced the cost of calculation and improved the accuracy of feature extraction. In the case of registration method of matching and tracking natural features, the adaptive and generic corner detection based on the Gravity-FREAK matching purification algorithm was used to eliminate abnormal matches, and Gravity Kaneda-Lucas Tracking (KLT) algorithm based on MEMS sensor can be used for the tracking registration of the targets and robustness improvement of tracking registration algorithm under mobile environment.

A method for the registration of spectral images of paintings and its evaluation

Abstract This work concerns the automatic registration of spectral images of paintings upon planar, or approximately planar, surfaces. An approach that capitalizes upon this planarity is proposed, which estimates homography transforms that register the spectral images into an aligned spectral cube. Homography estimation methods are comparatively evaluated for this purpose. A non-linear, robust estimation method that is based on keypoint features is adopted, as the most accurate. A marker-based, quantitative evaluation method is proposed for the measurement of multispectral image registration accuracy and, in turn, utilized for the comparison of the proposed registration method to the state of the art. For the same purpose, characteristic for this application domain, benchmark datasets that are annotated with correctly corresponding points have been compiled and are publicly availed.

Rancang Bangun Aplikasi Pengenalan Ukiran Bali dengan Metode ORB

Balinese carving is one of the arts and culture in Bali and have various types in its application, but the lack of information resulted in people do not really know the kinds of Balinese carving. Lack of information about Balinese carvings can be solved by utilizing advances in information technology in image processing, which is to make application that can recognize Balinese carvings. Applications that made is a mobile application that can be used on smartphones with Android operating system. The method that applied is Oriented FAST and Rotated BRIEF (ORB) that able to recognize the image of Bali carving based on the features keypoints. Carved image matching using Bruteforce Hamming Distance to find the similarity between features carved images so that they can be identified. The results of the accuracy of the research process recognition Balinese carving method ORB reaches the percentage of 48% Balinese carvings recognizable and 52% did not recognize, so it can be concluded the result method ORB is quite good to recognize Balinese carving.

An experimental evaluation of the accuracy of keypoints-based retinal image registration.

This work regards an investigation of the accuracy of a state-of-the-art, keypoint-based retinal image registration approach, as to the type of keypoint features used to guide the registration process. The employed registration approach is a local method that incorporates the notion of a 3D retinal surface imaged from different viewpoints and has been shown, experimentally, to be more accurate than competing approaches. The correspondences obtained between SIFT, SURF, Harris-PIIFD and vessel bifurcations are studied, either individually or in combinations. The combination of SIFT features with vessel bifurcations was found to perform better than other combinations or any individual feature type, alone. The registration approach is also comparatively evaluated against representative methods of the state-of-the-art in retinal image registration, using a benchmark dataset that covers a broad range of cases regarding the overlap of the acquired images and the anatomical characteristics of the imaged retinas.

Enhanced image similarity analysis system in digital pathology

In digital pathology, image similarity algorithms are used to find cancer in tissue cells from medical images. However, it is very difficult to apply image similarity algorithms used in general purpose system. Because in the medical field, accuracy and reliability must be perfect when looking for cancer cells by using image similarity techniques to pathology images. To cope with this problem, this paper proposes an efficient similar image search algorithm for digital pathology by applying leveling and tiling scheme on OpenSlide format. Furthermore, we apply image sync method to extract feature key points during image similarity processing. In the experiment, to prove the efficiency of the proposed system, we conduct several experiments including algorithm performance, algorithm accuracy and computation time. The experiments result shows that the proposed system efficiently retrieves similar cell images from pathology images.

Research of Adaptive Definition Model of Frontier Map Based on Feature Points

With increasingly tense social development resources, once involved in territorial issues between the countries, there must be fight for every inch of land. Therefore, it is particularly important to develop a frontier map. This paper combines with the adaptive principle and extraction of feature points to establish an adaptive definition model of frontier map based on the special purpose of the frontier map. And this paper first obtains a scatter diagram of the relationship between the scale limit and negative load by calculating the negative load of the electronic map, and then obtains a curvilinear equation through the differential fitting, so as to find out key feature points to reflect the variation trend of the curve. This paper finally establishes an adaptive definition model of the frontier map, and establishes seven key scales developed by the frontier defining map, so as to make a significant contribution to develop a perfect electronic map of the national boundary.

A Robust and Efficient Copy-Move Forgery Detection Technique based on SIFT and SVD

Objective: To detect copy-move forgery from a given digital image by reducing false detection thus increasing precision rate and F1 score. It is also invariant to scaling, rotation, noise attack. Methods/Analysis: The paper describes an effective and novel method to detect the copy-move forgery detection using Singular Value Decomposition (SVD) and Scale-Invariant Feature Transform (SIFT) features. It divides the given image into equal sized blocks and applies keypoint detection and feature descriptor for each block of the image. Then again SVD is calculated from128 SIFT descriptor to detect the forgery part of the given image. In this approach, a correlation is calculated for different images under different attacks like scaling, rotation, and noise. Finding: The proposed system is tested using standard image data against various types of image attacks like scaling, rotation, noise, blur etc. From the result, it is found that the proposed system is robust and invariant to most of the image attacks. The proposed system easily detects the copied part of the image more efficiently compared to other block based method since SIFT feature is invariant to scaling, rotation, noise, blur etc. Novelty/Improvement: The proposed methods uses block based technique and apply SIFT and SVD technique for detecting forged region in the given image. False detection is reduced in the proposed system which increases the precision and F1 score.

Improved Entropy and Key Point Feature Extractor for Effective Multimedia Information Retrieval

Improved Entropy and Key Point Feature Extractor for Effective Multimedia Information Retrieval

IMAGE FUSION FOR MULTIFOCUS IMAGES USING SPEEDUP ROBUST FEATURES

The multi-focus image fusion technique has emerged as major topic in image processing in order to generate all focus images with increased depth of field from multi focus photographs. Image fusion is the process of combining relevant information from two or more images into a single image. The image registration technique includes the entropy theory. Speed up Robust Features (SURF), feature detector and Binary Robust Invariant Scalable Key points (BRISK) feature descriptor is used in feature matching process. An improved RANDOM Sample Consensus (RANSAC) algorithm is adopted to reject incorrect matches. The registered images are fused using stationary wavelet transform (SWT).The experimental results prove that the proposed algorithm achieves better performance for unregistered multiple multi-focus images and it especially robust to scale and rotation translation compared with traditional direct fusion method.

Adapting Dynamic Appearance for Robust Visual Tracking

This paper designs a robust discriminative tracking method based on correlation filter and key point matching, the highlight of which is the ability to adapt appearance changes. Toward the complicated and changeable real-world scene, a multi-channel of optimal selection correlation filter (MOSCF) is proposed, and it can model the holistic appearances by learning short-term dynamic features. Meantime, a key point feature set is maintained for representing the long-term stable appearances. In order to combine the similarity evaluation scores of two appearance features, a confidence fusion framework is established to obtain the final output confidence map. Particularly, partially occluded object is located by a kind of local key point voting mechanism. To alleviate error accumulation caused by inappropriate learning, an adaptive learning ratio is set to independently update the filter template corresponding to each channel of MOSCF. The results of the qualitative and quantitative experiments on the OTB-100 benchmark suggest that the proposed tracker outperforms several state-of-the-art methods.

Uyghur Printed Document Image Retrieval Based on SIFT Features

Image retrieval is an attractive topic in the field of information retrieval in electronic library and computer vision. This paper proposed a research in the field of Uyghur document image retrieval that using 128-dimensional SIFT features for Uyghur printed document images. Through the Euclidean distance ratio with different threshold to statistics the number of right and wrong matched points of SIFT feature key points to achieve the local feature based document image retrieval for 1000 Uyghur printed document images. Through experiment acquired around 995 right matched points of each Uyghur printed document image, where the total key points are 1418 and the 100% matching percentage, respectively. The experimental results indicate that the 128-dimension SIFT based image retrieval method is a kind of effective method for Uyghur document image.

Threshold logic based low-level vision sparse object features

Purpose The real-time generation of feature descriptors for object recognition is a challenging problem. In this research, the purpose of this paper is to provide a hardware friendly framework to generate sparse features that can be useful for key feature point selection, feature extraction, and descriptor construction. The inspiration is drawn from feature formation processes of the human brain, taking into account the sparse, modular, and hierarchical processing of visual information.

Visual concept learning system based on lexical elements and feature key points conjunction

Subject of Research. The paper deals with the process of visual concept building based on two unlabeled sources of information (visual and textual). Method. Visual concept-based learning is carried out with image patterns and lexical elements simultaneous conjunction. Concept-based learning consists of two basic stages: early learning acquisition (primary learning) and lexical-semantic learning (secondary learning). In early learning acquisition stage the visual concept dictionary is created providing background for the next stage. The lexical-semantic learning makes two sources timeline analysis and extracts features in both information channels. Feature vectors are formed by extraction of separated information units in both channels. Mutual information between two sources describes visual concepts building criteria. Main Results. Visual conceptbased learning system has been developed; it uses video data with subtitles. The results of research have shown principal ability of visual concepts building by our system. Practical Relevance. Recommended application area of described system is an object detection, image retrieval and automatic building of visual concept-based data tasks.

Robust Feature Matching in Long-Running Poor-Quality Videos

We describe a methodology that is designed to match key point and region-based features in real-world images, acquired from long-running security cameras with no control over the environment. We detect frame duplication and images from static scenes that have no activity to prevent processing saliently identical images, and describe a novel blur-sensitive feature detection method, a combinatorial feature descriptor, and a distance calculation that efficiently unites texture and color attributes to discriminate feature correspondence in low-quality images. Our methods are tested by performing key point matching on real-world security images such as outdoor closed-circuit television videos that are low quality and acquired in uncontrolled conditions with visual distortions caused by weather, crowded scenes, emergency lighting, or the high angle of the camera mounting. We demonstrate an improvement in accuracy of matching key points between images compared with state-of-the-art feature descriptors. We use key point features from a Harris corner detector, scale-invariant feature transform, speeded-up robust features, binary robust invariant scalable keypoints, and features from accelerated segment test as well as MSER and MSCR region detectors to provide a comprehensive analysis of our generic method. We demonstrate feature matching using a 138D descriptor that improves the matching performance of a state-of-the-art 384D color descriptor with just 36% of the storage requirements.

Cucumber disease recognition based on Global-Local Singular value decomposition

Plant leaf based plant disease recognition is becoming an important research topic in the pattern recognition and image processing. Many conventional methods are not effective for plant disease recognition because of the complexity of the disease leaf images. Singular value decomposition (SVD) has shown its advantage over other classical algorithms in terms of clustering accuracy and robustness to the data representation and feature extraction. Nonetheless, there is still considerable room to improve the practical application performance of SVD. To improve the recognition rate of cucumber disease, based on the Gglobal-Local SVD, a recognition method of cucumber disease is proposed. First, the spot image is segmented by the watershed algorithm from each cucumber disease leaf image. Second, each spot image is divided into a few blocks, and then the combining features of Global-Local singular values are extracted and ordered from each block by SVD. Third, the key-point vectors are constructed and their dimensionalities are adjusted to equal to each other. Finally, a SVM classifier is adopted to recognize the class of the unknown disease leaf image. Comparing with the existing cucumber disease recognition methods, the proposed one can extract the key-point features from the spot image whose dimension is significantly lower than that of the original space. Also, it can effectively avoid the key information loss in the usual stochastic feature extraction and selection process. The approach is tested on three kinds of cucumber disease leaf images. In the experiments, the original singular values are replaced by the nonnegative Global-Local Logarithmic singular values. The experimental results show that the proposed method is effective and feasible for the cucumber disease recognition, and has the highest recognition rate and more practical value.