Generalized Hough Transform Research Articles

An online learned hough forest model based on improved multi-feature fusion matching for multi-object tracking

Object tracking has been one of the most important and active research areas in the field of computer vision. In order to solve low accuracy in object occlusion and deformation for multi-object tracking, an online learned Hough forest model based on improved multi-feature fusion matching for multi-object tracking is proposed in this paper. Firstly, positive and negative samples are selected online according to low-level association among detection responses and construct the feature model of the object with color histogram, histogram of oriented gradient (HOG) and optical flow information. Secondly, longer trajectory associations are generated based on the online learned Hough forest framework. Finally, a trajectory matching algorithm based on multi-feature fusion is proposed, and we introduce two methods of similarity measure in color histogram and feature points matching based on the Gabor filter to generate the probability matrix with the weighted factor. Therefore, it can further form the complete trajectories of the objects by associating them gradually. We evaluate our approach on three public data sets, and show significant improvements compared with state-of-art methods.

Protein-RNA Docking Using ICM.

Protein-RNA interactions play an important role in many biological processes. Computational methods such as docking have been developed to complement existing biophysical and structural biology techniques. Computational prediction of protein-RNA complex structures includes two steps: generating candidate structures from the individual protein and RNA parts and scoring the generated poses to pick out the correct one. In this work, we considered three recently developed data sets of protein-RNA complexes to evaluate and improve the performance of the FFT-based rigid-body docking algorithm implemented in the ICM package. An electrostatic term describing interactions between negatively charged phosphate groups and positively charged protein residues was added to the energy function used during the docking step to take into account the greater role that electrostatic interactions play in protein-RNA complexes. Next, the docking results were used to optimize a scoring function including van der Waals, electrostatic, and solvation terms. This optimization yielded a much smaller weight for the solvation term indicating that solvation energy may be less important for the scoring of protein-RNA structures. Rescoring of the generated poses with the new scoring function led to much higher success rates, while pose clustering by contact fingerprints produced further improvements, achieving a success rate of 0.66 for the top 100 structures.

Pengenalan Tulisan Tangan Offline Dengan Algoritma Generalized Hough Transform dan Backpropagation

Offline handwriting recognition is a technique used to recognize handwriting in paper document which converting it to digital form. Each handwriting has a unique style and shape that can be used to identify the owner. This research aims to develop a method to recognize the digital data handwriting. The method combines two algorithms; the first is Generalized Hough Transform in feature extraction process to detect arbitrary objects on the image; the second algorithm is Backpropagation to train the neural network based on feature values from feature extraction process. Artificial Neural Network (ANN) is used to improve the accuracy of the recognition system. The experiments are performed by using 100 handwriting images of 10 different people. The number of hidden units is defined through experiment to obtain optimal neural network. The experiment result shows that the recognition accuracy is up to 80%.
 Index Terms—Artificial Neural Network, Backrpopagation, Generalized Hough Transform, Offline handwiritng recognition

Fight Recognition in video using Hough Forests and 2D Convolutional Neural Network.

While action recognition has become an important line of research in computer vision, the recognition of particular events such as aggressive behaviors, or fights, has been relatively less studied. These tasks may be extremely useful in several video surveillance scenarios such as psychiatric wards, prisons or even in personal camera smartphones. Their potential usability has led to a surge of interest in developing fight or violence detectors. One of the key aspects in this case is efficiency, that is, these methods should be computationally fast. "Handcrafted" spatiotemporal features that account for both motion and appearance information can achieve high accuracy rates, albeit the computational cost of extracting some of those features is still prohibitive for practical applications. The deep learning paradigm has been recently applied for the first time to this task too, in the form of a 3D Convolutional Neural Network that processes the whole video sequence as input. However, results in human perception of other's actions suggest that, in this specific task, motion features are crucial. This means that using the whole video as input may add both redundancy and noise in the learning process. In this work, we propose a hybrid "handcrafted/learned" feature framework which provides better accuracy than the previous feature learning method, with similar computational efficiency. The proposed method is compared to three related benchmark datasets. The method outperforms the different state-of-the-art methods in two of the three considered benchmark datasets.

Applying Pose Clustering and MD Simulations To Eliminate False Positives in Molecular Docking.

In this work, we developed a computational protocol that employs multiple molecular docking experiments, followed by pose clustering, molecular dynamic simulations (10 ns), and energy rescoring to produce reliable 3D models of antibody-carbohydrate complexes. The protocol was applied to 10 antibody-carbohydrate co-complexes and three unliganded (apo) antibodies. Pose clustering significantly reduced the number of potential poses. For each system, 15 or fewer clusters out of 100 initial poses were generated and chosen for further analysis. Molecular dynamics (MD) simulations allowed the docked poses to either converge or disperse, and rescoring increased the likelihood that the best-ranked pose was an acceptable pose. This approach is amenable to automation and can be a valuable aid in determining the structure of antibody-carbohydrate complexes provided there is no major side chain rearrangement or backbone conformational change in the H3 loop of the CDR regions. Further, the basic protocol of docking a small ligand to a known binding site, clustering the results, and performing MD with a suitable force field is applicable to any protein ligand system.

Semi- and weakly-supervised human pose estimation

For human pose estimation in still images, this paper proposes three semi- and weakly-supervised learning schemes. While recent advances of convolutional neural networks improve human pose estimation using supervised training data, our focus is to explore the semi- and weakly-supervised schemes. Our proposed schemes initially learn conventional model(s) for pose estimation from a small amount of standard training images with human pose annotations. For the first semi-supervised learning scheme, this conventional pose model detects candidate poses in training images with no human annotation. From these candidate poses, only true-positives are selected by a classifier using a pose feature representing the configuration of all body parts. The accuracies of these candidate pose estimation and true-positive pose selection are improved by action labels provided to these images in our second and third learning schemes, which are semi- and weakly-supervised learning. While the first and second learning schemes select only poses that are similar to those in the supervised training data, the third scheme selects more true-positive poses that are significantly different from any supervised poses. This pose selection is achieved by pose clustering using outlier pose detection with Dirichlet process mixtures and the Bayes factor. The proposed schemes are validated with large-scale human pose datasets.

Detection of Curvilinear Structure in Images by a Multi-Centered Hough Forest Method

Automatic detection of curvilinear structure in images is a challenging task in computer vision applications. In this paper, we present a novel supervised learning method to identify curvilinear structure based on a modified Hough forest framework. We regard the curvilinear structure as a special object which has multiple object centers at centerline points, and construct a multi-centered Hough forest (MCHF) to cast confidence votes for each image point being a local curvilinear center in generalized Hough space. Considering the specific properties of curvilinear structure, we modify the classical Hough forest method in three aspects. First, the features are composed of a set of filter responses, and each filter is a base which is learned by sparse presentation techniques. Second, the offset of an image patch against its reference center is represented by a displacement function instead of Euclidean distance used in the classical Hough forest. Third, the centerline orientation is incorporated into the forest to reflect the local trend, which is considered in Hough voting. We conduct experiments on retinal vessel, neuron, and aerial road images. Both visualized and quantitative results demonstrate the good performance of MCHF in detecting complex curvilinear structures in various images.

Wrong Matching Points Elimination after Scale Invariant Feature Transform and Its Application to Image Matching

When images are rotated and the scale varies or there are similar objects in the images, wrong matching points appear easily in the scale invariant feature transform (SIFT). To address the problem, this paper proposes a SIFT wrong matching points elimination algorithm. The voting mechanism of Generalized Hough Transform (GHT) is introduced to find the rotation and scaling of the image and locate where the template image appears in the scene in order to completely reject unmatched points. Through a discovery that the neighborhood diameter ratio and direction angle difference of correct matching pairs have a quantitative relationship with the image’s rotation and scaling information, we further remove the mismatching points accurately. In order to improve image matching efficiency, a method for finding the optimal scaling level is proposed. A scaling multiple is obtained through training of sample images and applied to all images to be matched. The experimental results demonstrate that the proposed algorithm can eliminate wrong matching points more effectively than the other three commonly used methods. The image matching tests have been conducted on images from the Inria BelgaLogos database. Performance evaluation results show that the proposed method has a higher correct matching rate and higher matching efficiency.

Person Identification Using Pose-Based Hough Forests from Skeletal Action Sequence

Person Identification Using Pose-Based Hough Forests from Skeletal Action Sequence

Hough Forest With Optimized Leaves for Global Hand Pose Estimation With Arbitrary Postures.

Vision-based hand pose estimation is important in human-computer interaction. While many recent works focus on full degree-of-freedom hand pose estimation, robust estimation of global hand pose remains a challenging problem. This paper presents a novel algorithm to optimize the leaf weights in a Hough forest to assist global hand pose estimation with a single depth camera. Different from traditional Hough forest, we propose to learn the vote weights stored at the leaf nodes of a forest in a principled way to minimize average pose prediction error, so that ambiguous votes are largely suppressed during prediction fusion. Experiments show that the proposed method largely improves pose estimation accuracy with optimized leaf weights on both synthesis and real datasets and performs favorably compared to state-of-the-art convolutional neural network-based methods. On real-world depth videos, the proposed method demonstrates improved robustness compared to several other recent hand tracking systems from both industry and academy. Moreover, we utilize the proposed method to build virtual/augmented reality applications to allow users to manipulate and examine virtual objects with bare hands.

Spatio-temporal elastic cuboid trajectories for efficient fight recognition using Hough forests

While action recognition has become an important line of research in computer vision, the recognition of particular events such as aggressive behaviors, or fights, has been relatively less studied. These tasks may be exceedingly useful in some video surveillance scenarios such as psychiatric centers, prisons or even in personal camera smartphones. Their potential usability has caused a surge of interest in developing fight or violence detectors. The key aspect in this case is efficiency, that is, these methods should be computationally very fast. In this paper, spatio-temporal elastic cuboid trajectories are proposed for fight recognition. This method is based on the use of blob movements to create trajectories that capture and model the different motions that are specific to a fight. The proposed method is robust to the specific shapes and positions of the individuals. Additionally, the standard Hough forests classifier is adapted in order to use it with this descriptor. This method is compared to other nine related methods on four datasets. The results show that the proposed method obtains the best accuracy for each dataset and is also computationally efficient.

Identification and Assessment of Selected Handwritten Function Graphs Using Least Square Approximation Combined with General Hough Transform

Abstract The paper provides a comparison of three variants of algorithms for automatic assessment of some examination tasks involving sketching a function graph based on image processing. Three types of functions have been considered: linear, quadratic, and trigonometric. The assumption adopted in the design of the algorithm is to map the way the examiner assesses the solutions and to achieve the evaluation quality close to the one obtained in manual evaluation. In particular, the algorithm should not reject a partly correct solution and also extract the correct solution from other lines, deletions and corrections made by a student. Essential subproblems to solve in our scheme concern image segmentation, object identification and automatic understanding. We consider several techniques based on Hough Transform, least square fitting and nearest neighbor based classification. The most reliable solution is an algorithm combining least square fitting and Hough Transform.

Performance Evaluation of the Generalized Hough Transform

The generalized Hough transform(GHough) can be used effectively for detecting and extracting an arbitrary-shaped 2-D model in an input image. However, the main drawbacks of the GHough are both heavy computation and an excessive storage requirement. Thus, most of the researches so far have focused on reducing both the time and space requirement of the GHough. But it is still not clear how well their improved algorithms will perform under various noise in an input image. Thus, this paper proposes a new framework that can measure the performance of the GHough quantitatively. For this purpose, we view the GHough as a detector in signal detection theory and the ROC curve will be used to specify the performance of the GHough. Finally, we show that we can evaluate the GHough under various noise conditions in an input image.

A Positioning Lockholes of Container Corner Castings Method Based on Image Recognition

Abstract This article proposes a method of locating and recognizing lockholes in shipping container corner castings. This method converts the original image of the containers captured by a camera into the HSV (Hue, Saturation, Value) color space. To reduce the influence of the surface color of the containers and lights from the environment on the locating and recognizing algorithm, most noisy points of the image are filtered by binarization and a morphology opening operation to make the features of the containers clearer in the image. Thus, the container body can be separated from the total image. Then, the position and size of the corner castings are defined through calculation based on the international standard of the shipping container size. Lastly, by using this method, we can locate the corner casting in the image by using the General Hough Transform fitting algorithm onto ellipses.

Aeroplane detection in very-high-resolution images using deep feature representation and rotation-invariant Hough forests

ABSTRACTThis letter proposes a processing chain for detecting aeroplanes from very high-resolution (VHR) remotely sensed images with the fusion of deep feature representation and rotation-invariant Hough forests. First, superpixel segmentation is used to generate meaningful and non-redundant patches. Second, deep learning techniques are exploited to construct a multi-layer feature encoder for representing high-order features of patches. Third, a set of multi-scale rotation-invariant Hough forests are trained to detect aeroplanes of varying orientations and sizes. Experiments show that the proposed method is a promising solution for detecting aeroplanes from VHR remotely sensed images, with a completeness, correctness, and F-measure of 0.956, 0.970, and 0.963, respectively. Comparative studies with four existing methods also demonstrate that the proposed method outperforms the other existing methods in accurately detecting aeroplanes of varying appearances, orientations, and sizes.

Road Detection by Using a Generalized Hough Transform

Road detection plays key roles for remote sensing image analytics. Hough transform (HT) is one very typical method for road detection, especially for straight line road detection. Although many variants of Hough transform have been reported, it is still a great challenge to develop a low computational complexity and time-saving Hough transform algorithm. In this paper, we propose a generalized Hough transform (i.e., Radon transform) implementation for road detection in remote sensing images. Specifically, we present a dictionary learning method to approximate the Radon transform. The proposed approximation method treats a Radon transform as a linear transform, which then facilitates parallel implementation of the Radon transform for multiple images. To evaluate the proposed algorithm, we conduct extensive experiments on the popular RSSCN7 database for straight road detection. The experimental results demonstrate that our method is superior to the traditional algorithms in terms of accuracy and computing complexity.

Compressive tracking based on online Hough forest

Existing tracking algorithms often suffer from the drift and lost problems caused by factors such as pose variation, illumination change, occlusion and motion. Therefore, developing a robust and effective tracker is still a challenging task. In this paper, we propose a real-time compressive tracking based on online Hough forest. The gray and texture features of discrete samples are extracted and compressed via the random measurement matrix. Online Hough forest classifier is used to vote the location probability of the target, and it optimizes the confidence map estimation for the target detection. The location of target being tracked is determined by combining the upper frame of the target center location and the probability confidence map of the incremental Hough forest. Finally, the classifier parameters are updated online by introducing the illumination variation and target occlusion feedback mechanism adaptively. The experiments with state-of-the-art algorithms on challenging sequences demonstrated that the proposed algorithm can effectively enhance the robustness and accuracy, and inherit the real-time performance of the compressive tracking algorithm.

A learning-based variable size part extraction architecture for 6D object pose recovery in depth images

State-of-the-art techniques for 6D object pose recovery depend on occlusion-free point clouds to accurately register objects in 3D space. To deal with this shortcoming, we introduce a novel architecture called Iterative Hough Forest with Histogram of Control Points that is capable of estimating the 6D pose of an occluded and cluttered object, given a candidate 2D bounding box. Our Iterative Hough Forest (IHF) is learnt using parts extracted only from the positive samples. These parts are represented with Histogram of Control Points (HoCP), a “scale-variant” implicit volumetric description, which we derive from recently introduced Implicit B-Splines (IBS). The rich discriminative information provided by the scale-variant HoCP features is leveraged during inference. An automatic variable size part extraction framework iteratively refines the object's roughly aligned initial pose due to the extraction of coarsest parts, the ones occupying the largest area in image pixels. The iterative refinement is accomplished based on finer (smaller) parts, which are represented with more discriminative control point descriptors by using our Iterative Hough Forest. Experiments conducted on a publicly available dataset report that our approach shows better registration performance than the state-of-the-art methods.

Incoherent optical generalized Hough transform: pattern recognition and feature extraction applications

Pattern recognition and feature extraction are image processing applications of great interest in defect inspection and robot vision among others. In comparison to purely digital methods, the attractiveness of optical processors for pattern recognition lies in their highly parallel operation and real-time processing capability. This work presents an optical implementation of the generalized Hough transform (GHT), a well-established technique for recognition of geometrical features in binary images. Detection of a geometric feature under the GHT is accomplished by mapping the original image to an accumulator space; the large computational requirements for this mapping make the optical implementation an attractive alternative to digital-only methods. We explore an optical setup where the transformation is obtained, and the size and orientation parameters can be controlled, allowing for dynamic scale and orientation-variant pattern recognition. A compact system for the above purposes results from the use of an electrically tunable lens for scale control and a pupil mask implemented on a high-contrast spatial light modulator for orientation/shape variation of the template. Real-time can also be achieved. In addition, by thresholding of the GHT and optically inverse transforming, the previously detected features of interest can be extracted.

Latent-Class Hough Forests for 6DoF Object Pose Estimation.

In this paper we present Latent-Class Hough Forests, a method for object detection and 6 DoF pose estimation in heavily cluttered and occluded scenarios. We adapt a state of the art template matching feature into a scale-invariant patch descriptor and integrate it into a regression forest using a novel template-based split function. We train with positive samples only and we treat class distributions at the leaf nodes as latent variables. During testing we infer by iteratively updating these distributions, providing accurate estimation of background clutter and foreground occlusions and, thus, better detection rate. Furthermore, as a by-product, our Latent-Class Hough Forests can provide accurate occlusion aware segmentation masks, even in the multi-instance scenario. In addition to an existing public dataset, which contains only single-instance sequences with large amounts of clutter, we have collected two, more challenging, datasets for multiple-instance detection containing heavy 2D and 3D clutter as well as foreground occlusions. We provide extensive experiments on the various parameters of the framework such as patch size, number of trees and number of iterations to infer class distributions at test time. We also evaluate the Latent-Class Hough Forests on all datasets where we outperform state of the art methods.