Average Running Time Research Articles

Multi-objective optimisation of species distribution models for river management

Environmental and measure implementation costs are two key factors to be considered by river managers in decision making. To balance effects and costs of an action, practitioners can rely on diagnostic analysis of presence/absence freshwater species distribution models (SDMs) trained to over- or underestimating species presence. Prevalence-adjusted model training aims to balance under- and overestimation depending on study objectives and training data characteristics. The objective of minimising under- and overestimation is a typical example of multi-objective optimisation (MOO). The aim of this paper is to address, for the first time, the practice of MOO-based prevalence-adjusted SDM training for freshwater decision management. In a numerical experiment, the use of Pareto-based MOO, specifically the non-dominated sorting genetic algorithm II (NSGA-II), is compared to commonly-used single-objective optimisation. SDMs for 11 pollution-sensitive freshwater macroinvertebrate species are trained with a subset of the Limnodata, a large data set holding records in the Netherlands over 30 years at 20,000 locations. An increase of two to four times is observed for the ability to identify a large range distribution of the solutions in the Pareto space, when using NSGA-II counter to repeated single-objective optimisation, this by increasing the average runtime with only four percent for a single run. In addition, the use of NSGA-II is found to be effective to identify reliable SDMs useful for diagnostic analysis. By applying and comparing a broad range of MOO methodologies for prevalence-adjusted model training, we believe a closer collaboration between model developers and freshwater managers can be facilitated and environmental standard limits can be set on a more objective basis. In conclusion, the use of MOO for prevalence-adjusted model training is assessed as a valuable tool to support river - and potentially all environmental - decision making.

A fast and simple algorithm for calculating flow accumulation matrices from raster digital elevation models

Abstract. Flow accumulation is an essential input for many hydrological and topographic analyses such as stream channel extraction, stream channel ordering and sub-watershed delineation. Flow accumulation matrices can be derived directly from DEMs and general have O(NlogN) time complexity (Arge, 2003; Bai et al., 2015). It is more common to derive the flow accumulation matrix from a flow direction matrix. This study focuses on calculating the flow accumulation matrix from the flow direction matrix that is derived using the single-flow D8 method (Barnes et al., 2014; Garbrecht & Martz, 1997; Nardi et al., 2008; O'Callaghan & Mark, 1984). In this study, we find give an overview of algorithms for flow accumulation calculation that have O(N) time complexity. These algorithms include algorithms are based on the concept of the number of input drainage paths (Wang et al.2011, Jiang et al. 2013), the algorithm based on the basin tree indices (Su et al. 2015), and the recursive algorithm (Choi, 2012; Freeman, 1991).We propose a fast and simple algorithm to calculate the flow accumulation matrix. Compared with the existing algorithms that have O(N) time complexity, our algorithm runs faster and generally requires less memory. Our algorithm is also simple to implement. In our algorithm, we define three types of cells within a flow direction matrix: source cells, interior cells and intersection cells. A source cell does not have neighboring cells that drain to it and its NIDP value is zero. An interior cell has only one neighboring cell that drains to it and its NIDP value is one. An intersection cell has more than one neighboring cell that drains to it and its NIDP value is greater than one. The proposed algorithm initializes the flow accumulation matrix with the value of one. Our algorithm first calculates the NIDP matrix from the flow direction matrix. The algorithm then traverses each cell within the flow direction matrix row by row and column by column, similar to the traversal algorithm. When a source cell c is encountered, the algorithm traces all downstream cells of c until it encounters an intersection cell i. During the tracing, the accumulation value of a cell is added to the accumulation value of its immediate downstream cell. An interior cell has only one neighboring cell that drains to it and its final accumulation value is obtained when the tracing is done. The accumulation value of the intersection cell i is updated from this drainage path. However, cell i has other unvisited neighboring cells that drain to it and its final accumulation value cannot be obtained after this round of tracing. The algorithm decreases the NIDP value of i by one. Cell i is visited again when other drainage paths that pass through it are traced. When all of the drainage paths that pass through it are traced, cell i is treated as an interior cell and the final accumulation value of i is obtained correctly and the last tracing process can continue the tracing after cell i is treated as an interior cell. A worked example of the proposed algorithm is shown in Figure 1.The five flow accumulation algorithms with O(N) time complexity, including Wang’s algorithm, Jiang’s algorithm, the BTI-based algorithm, the recursive algorithm and our proposed algorithm, are implemented in C++. The 3-m LiDAR-based DEMs of thirty counties in the state of Minnesota, USA, are downloaded from the FTP site operated by the Minnesota Geospatial Information Office. The first 30 counties in Minnesota in alphabetic order are chosen for the experiments to avoid selection bias. We use the algorithm proposed by Wang and Liu (2006) to fill the depressions and derive the flow direction matrices for all tested counties. The running times on the Windows system are listed in Figure 2.The average running times per 100 million cells are 14.42 seconds for Wang’s algorithm, 15.90 seconds for Jiang’salgorithm, 18.95 seconds for the BTI-based algorithm, 10.87 seconds for the recursive algorithm, and 5.26 seconds forour proposed algorithm. Our algorithm runs the fastest for all tested DEM. The speed-up ratios of our proposedalgorithm over the second fastest algorithm is about 51%.

Real-Time Dense Reconstruction of Tissue Surface From Stereo Optical Video.

We propose an approach to reconstruct dense three-dimensional (3D) model of tissue surface from stereo optical videos in real-time, the basic idea of which is to first extract 3D information from video frames by using stereo matching, and then to mosaic the reconstructed 3D models. To handle the common low-texture regions on tissue surfaces, we propose effective post-processing steps for the local stereo matching method to enlarge the radius of constraint, which include outliers removal, hole filling, and smoothing. Since the tissue models obtained by stereo matching are limited to the field of view of the imaging modality, we propose a model mosaicking method by using a novel feature-based simultaneously localization and mapping (SLAM) method to align the models. Low-texture regions and the varying illumination condition may lead to a large percentage of feature matching outliers. To solve this problem, we propose several algorithms to improve the robustness of the SLAM, which mainly include 1) a histogram voting-based method to roughly select possible inliers from the feature matching results; 2) a novel 1-point RANSAC-based [Formula: see text] algorithm called as DynamicR1PP [Formula: see text] to track the camera motion; and 3) a GPU-based iterative closest points (ICP) and bundle adjustment (BA) method to refine the camera motion estimation results. Experimental results on ex- and in vivo data showed that the reconstructed 3D models have high-resolution texture with an accuracy error of less than 2 mm. Most algorithms are highly parallelized for GPU computation, and the average runtime for processing one key frame is 76.3 ms on stereo images with 960×540 resolution.

Failure Recovery in Resilient X10

Cloud computing has made the resources needed to execute large-scale in-memory distributed computations widely available. Specialized programming models, e.g., MapReduce, have emerged to offer transparent fault tolerance and fault recovery for specific computational patterns, but they sacrifice generality. In contrast, the Resilient X10 programming language adds failure containment and failure awareness to a general purpose, distributed programming language. A Resilient X10 application spans over a number of places. Its formal semantics precisely specify how it continues executing after a place failure. Thanks to failure awareness, the X10 programmer can in principle build redundancy into an application to recover from failures. In practice, however, correctness is elusive, as redundancy and recovery are often complex programming tasks. This article further develops Resilient X10 to shift the focus from failure awareness to failure recovery, from both a theoretical and a practical standpoint. We rigorously define the distinction between recoverable and catastrophic failures. We revisit the happens-before invariance principle and its implementation. We shift most of the burden of redundancy and recovery from the programmer to the runtime system and standard library. We make it easy to protect critical data from failure using resilient stores and harness elasticity—dynamic place creation—to persist not just the data but also its spatial distribution. We demonstrate the flexibility and practical usefulness of Resilient X10 by building several representative high-performance in-memory parallel application kernels and frameworks. These codes are 10× to 25× larger than previous Resilient X10 benchmarks. For each application kernel, the average runtime overhead of resiliency is less than 7%. By comparing application kernels written in the Resilient X10 and Spark programming models, we demonstrate that Resilient X10’s more general programming model can enable significantly better application performance for resilient in-memory distributed computations.

An Analysis of Multirules for Monitoring Assay Quality Control

Multirules are often employed to monitor quality control (QC). The performance of multirules is usually determined by simulation and is difficult to predict. Previous studies have not provided computer code that would enable one to experiment with multirules. It would be helpful for analysts to have computer code to analyze rule performance. To provide code to calculate power curves and to investigate certain properties of multirule QC. We developed computer code in the R language to simulate multirule performance. Using simulation, we studied the incremental performance of each rule and determined the average run length and time to signal. We provide R code for simulating multirule performance. We also provide a Microsoft Excel spreadsheet with a tabulation of results that can be used to create power curves. We found that the R4S and 10x rules add very little power to a multirule set designed to detect shifts in the mean. QC analysts should consider using a limited-rule set.

Relationship Of Sex And Physical Activity On Vertical Jump Power Changes In U.s. Army Trainees

One goal of Basic Combat Training (BCT) is to improve the general fitness of recruits to successfully meet the demands of military activities. Previous studies have focused on the aerobic fitness responses to BCT. In contrast few investigations have examined the changes in whole body power production following BCT and if these changes may be modified by various pre-BCT factors. More specifically, do all recruits show similar training responses in terms of muscular power production as measured by vertical jump (VJ) testing? PURPOSE: To determine the relationship of sex and physical activity history on changes in VJ power output following BCT. METHODS: Four hundred fourteen recruits (298 men; 116 women; (mean ± SD) age: 21 ± 3 y; height: 172 ± 9 cm; body mass: 73.0 ± 13.4 kg) performed maximal VJ testing before and after 8 weeks of U.S. Army BCT. Body mass and VJ height were used to estimate VJ peak power using the Harman equation. Recruits filled out a survey on their prior physical activity during the 2 months prior to entering BCT. Logistic regression was used to calculate odds ratios showing whether sex or physical activity prior to BCT is predictive of changes in a recruit’s VJ power output. RESULTS: Females were 2.1 times more likely to show an increase in VJ power than males (p=0.01). Recruits that performed running training 3-4 times per week were 1.8 times more likely to improve their VJ power than recruits that only ran 0-2 times per week (p=0.01). Recruits that had an average running mile time from 7:00-7:59 (min:sec) were 1.6 times more likely to generate more VJ power than those running at a sub 7:00 mile time (p=0.03). CONCLUSION: Recruits showed differential changes in VJ performance following BCT, depending on sex, prior running training experience, and average 1-mile running time. DISCLAIMER: The views expressed in this abstract are those of the authors and do not reflect the official policy of the Department of Army, Department of Defense, or the U.S. Government. Supported by the U.S. Army Medical Research and Materiel Command.

Colored rice quality inspection system using machine vision

The machine vision-based quality evaluation of colored rice is an important component of achieving automatic rice production and processing. This study took red indica rice as the research object and developed a machine vision system for inspecting flawed rice kernels that are broken, chalky, damaged or spotted. Near infrared images of colored rice samples were collected by the machine vision system. A support vector machine (SVM) classifier, with the input of the invariant moment ellipse major axis, was applied to identify the broken rice kernels in the images. Then, the head rice images were obtained, and another SVM performed the gray-level segmentation. The segmented areas were doubly examined using the centroid distance constraint and the pixel search positioning method, which allowed the chalky areas to be accurately extracted. Finally, the damaged and the spotted areas on the rice kernels were detected by using edge detection and morphological methods. The experimental results show that the recognition accuracy for broken kernels, chalkiness, and damaged and spotted areas reached 99.3%, 96.3% and 93.6%, respectively. In addition, the average running time of the proposed method was 0.15 s, with four types of defects detected at one-time. Hence, it was concluded that the proposed method has significant potential to be applied for rapid and accurate colored rice quality detection and to provide technical support for the machine vision-based inspection of the automated grading of rice.

Digital image recognition based on Fractional-order-PCA-SVM coupling algorithm

This paper use the fractional differential mask operator to describe and deal with the highly self similar digital medical image, then extracting the main features of the these images with principal component analysis, and recognizing these images with SVM algorithm. To provide an efficient algorithm for medical image processing with high degree of self similarity. Based on the analysis of fractional differential theory, principal component analysis theory and support vector machine (SVM) algorithm, designing a Fractional-order-PCA-SVM coupling algorithm for digital image recognition. and taking the digital image in ORL face database as the empirical object, using PCA-NN coupling algorithm, SVM algorithm, PCA-SVM algorithm, Sobel-PCA-NN algorithm, coupling coupling Sobel-SVM coupling algorithm, Sobel-PCA-SVM algorithm, Fractional-order-PCA-NN algorithm and coupling coupling Fractional-order-SVM coupling algorithm as the comparison algorithm, creating four experiments with the same sample, regarding the speed of operation and the accuracy of recognition as the criteria of high evaluation, verifying the superiority of the Fractional-order-PCA-SVM coupling algorithm designed in this paper. The average run time of the Fractional-order-PCA-SVM coupling algorithm only 4.152 s in the four experiments, although the speed of operation is not the highest, it also meets the demand of identification. The reason that the operation speed of the algorithm slightly lower than the three control algorithms is that the three algorithms are monotonic; The average accuracy rate of Fractional-order-PCA-SVM coupling algorithm is 99.2425% in four experiments, significantly better than the eight comparison algorithm; For digital medical image recognition the coupling algorithm designed in this paper is effective.

Customer Critique Analysis System for PT. KCI’s Twitter

Commuter Line is one of the main choices for users of public transport in Jakarta. Commuter Line in Indonesia is managed by PT. Kereta Commuter Indonesia (PT. KCI). PT. KCI has an official Twitter account @CommuterLine. The purpose of this research was built Customer Critique Analysis System for Twitter PT. Kereta Commuter Indonesia using Support Vector Machine (SVM). The software engineering method used in this research is Component-Based Software Engineering (CBSE). Applications built in this research uses some components/libraries, which are Tweetsharp, LibSVM, Metronic, Kendo UI and JQCloud. The results of this research is a useful analytics application to perform grabbing and tweet data analysis from Twitter account @CommuterLine, and a dashboard to display the results of the analysis in graphical form. The test results show that the system meets the functional and nonfunctional requirements from the user. Testing is done by installing the application on the server PT. KCI, then it was tested by user.. Non-functional testing was done by measuring application performance and usability. Performance measured is classification accuracy and running time of the program. Accuracy performance testing resulted in average sentence classification accuracy of 80.4%, and the accuracy of complaint classification type of 82.3%. The average running time is 3 minutes 25 seconds.

A Study on Generic Data Analysis Application the Cloud Using SGX

The analysis itself however in the midst of preparation and post-processing steps. Handling information returning directly from the supply, e.g. a sensor, typically needs preconditioning like parsing and removing orthogonal data before data processing algorithms will be applied to research the info. complete data processing frameworks generally don't offer such parts since they need a currency computer file format. what is more, they're typically restricted to the offered algorithms or a speedy integration of latest algorithms for the aim of fast testing isn't attainable. to deal with this disadvantage, we have a tendency to gift the info analysis framework Knowing, that is well extendable with extra algorithms by exploitation associate OSGi compliant design. the software package and therefore the hypervisor out of the TCB; therefore, confidentiality and integrity square measure preserved albeit these massive parts square measure compromised. VC3 depends on SGX processors to isolate memory regions on individual computers, and to deploy new protocols that secure distributed MapReduce computations. VC3 optionally enforces region self-integrity invariants for all MapReduce code running inside isolated regions, to stop attacks thanks to unsafe memory reads and writes. Experimental results on common benchmarks show that VC3 performs well compared with unprotected Hadoop: VC3's average runtime overhead is negligible for its base security guarantees, 4.5% with write integrity and eight with read/write integrity.

Mapping quantum circuits on 3D nearest-neighbor architectures

In some quantum technologies, an interaction is only allowed between physically adjacent qubits, hence the nearest-neighbor requirement is needed. In such technologies, quantum gates are limited to operate on adjacent qubits. To make a quantum circuit compliant with the nearest-neighbor requirement, SWAP gates are inserted into the circuit to move the interacting qubits of a gate to be adjacent to each other. The mapping of qubits on the physical environment has an important role on reducing the number of SWAP gates and thus the circuit latency. Focusing on this issue, in this paper, a method is proposed that maps a quantum circuit onto a 3D physical hardware such as a 3D optical lattice. A new methodology for this mapping problem is proposed based on a complex network spectral clustering algorithm and graph theory, which are suitable for very large scale networks. It includes three steps: finding the order of qubit mapping, placing physical qubits, and routing. Simulation results show that the proposed mapping approach not only decreases the average number of SWAP gates by about 37% but also improves the average runtime by about 87% for the 2D architecture compared to PAQCS. Moreover, it reduces the average number of SWAP gates for the 3D architecture by 17.1% compared to the best studies in the literatures.

Evolution of star clusters on eccentric orbits: semi-analytical approach

We study the dynamical evolution of star clusters on eccentric orbits using a semi-analytical approach. In particular we adapt and extend the equations of EMACSS code, introduced by Gieles et al. (2014), to work with eccentric orbits. We follow the evolution of star clusters in terms of mass, half-mass radius, core radius, Jacobi radius and the total energy over their dissolution time. Moreover, we compare the results of our semi-analytical models against $N$-body computations of clusters with various initial half-mass radius, number of stars and orbital eccentricity to cover both tidally filling and under-filling systems. The evolution profiles of clusters obtained by our semi-analytical approach closely follow those of $N$-body simulations in different evolutionary phases of star clusters, from pre-collapse to post-collapse. Given that the average runtime of our semi-analytical models is significantly less than that of $N$-body models, our approach makes it feasible to study the evolution of large samples of globular clusters on eccentric orbits.

Lightweight Architecture for Real-Time Hand Pose Estimation with Deep Supervision

The high demand for computational resources severely hinders the deployment of deep learning applications in resource-limited devices. In this work, we investigate the under-studied but practically important network efficiency problem and present a new, lightweight architecture for hand pose estimation. Our architecture is essentially a deeply-supervised pruned network in which less important layers and branches are removed to achieve a higher real-time inference target on resource-constrained devices without much accuracy compromise. We further make deployment optimization to facilitate the parallel execution capability of central processing units (CPUs). We conduct experiments on NYU and ICVL datasets and develop a demo1 using the RealSense camera. Experimental results show our lightweight network achieves an average running time of 32 ms (31.3 FPS, the original is 22.7 FPS) before deployment optimization. Meanwhile, the model is only about half parameters size of the original one with 11.9 mm mean joint error. After the further optimization with OpenVINO, the optimized model can run at 56 FPS on CPUs in contrast to 44 FPS running on a graphics processing unit (GPU) (Tensorflow) and it can achieve the real-time goal.

Masi Entropy for Satellite Color Image Segmentation Using Tournament-Based Lévy Multiverse Optimization Algorithm

A novel multilevel threshold segmentation method for color satellite images based on Masi entropy is proposed in this paper. Lévy multiverse optimization algorithm (LMVO) has a strong advantage over the traditional multiverse optimization algorithm (MVO) in finding the optimal solution for the segmentation in the three channels of an RGB image. As the work advancement introduces a Lévy multiverse optimization algorithm which uses tournament selection instead of roulette wheel selection, and updates some formulas in the algorithm with mutation factor. Then, the proposal is called TLMVO, and another advantage is that the population diversity of the algorithm in the latest iterations is maintained. The Masi entropy is used as an application and combined with the improved TLMVO algorithm for satellite color image segmentation. Masi entropy combines the additivity of Renyi entropy and the non-extensibility of Tsallis entropy. By increasing the number of thesholds, the quality of segmenttion becomes better, then the dimensionality of the problem also increases. Fitness function value, average CPU running time, Peak Signal-to-Noise Ratio (PSNR), Structural Similarity Index (SSIM) and Feature Similarity Index (FSIM) were used to evaluate the segmentation results. Further statistical evaluation was given by Wilcoxon's rank sum test and Friedman test. The experimental results show that the TLMVO algorithm has wide adaptability to high-dimensional optimization problems, and has obvious advantages in objective function value, image quality detection, convergence performance and robustness.

Privacy-Preserving Record Grouping and Consent Management Based on a Public-Private Key Signature Scheme: Theoretical Analysis and Feasibility Study.

BackgroundClinical and social trials create evidence that enables medical progress. However, the gathering of personal and patient data requires high security and privacy standards. Direct linking of personal information and medical data is commonly hidden through pseudonymization. While this makes unauthorized access to personal medical data more difficult, a centralized pseudonymization list can still pose a security risk. In addition, medical data linked via pseudonyms can still be used for data-driven reidentification.ObjectiveOur objective was to propose a novel approach to pseudonymization based on public-private key cryptography that allows (1) decentralized patient-driven creation and maintenance of pseudonyms, (2) 1-time pseudonymization of each data record, and (3) grouping of patient data records even without knowing the pseudonymization key.MethodsBased on public-private key cryptography, we set up a signing mechanism for patient data records and detailed the workflows for (1) user registration, (2) user log-in, (3) record storing, and (4) record grouping. We evaluated the proposed mechanism for performance, examined the potential risks based on cryptographic collision, and carried out a threat analysis.ResultsThe performance analysis showed that all workflows could be performed with an average runtime of 0.057 to 42.320 ms (user registration), 0.083 to 0.606 ms (record creation), and 0.005 to 0.198 ms (record grouping) depending on the chosen cryptographic tools. We expected no realistic risk of cryptographic collision in the proposed system, and the threat analysis revealed that 3 distinct server systems of the proposed setup had to be compromised to allow access to combined medical data and private data. However, this would still allow only for data-driven deidentification. For a full reidentification, all 3 trial servers and all study participants would have to be compromised. In addition, the approach supports consent management, automatically anonymizes the data after trial closure, and provides basic mechanisms against data forging.ConclusionsThe proposed approach has a high security and privacy level in comparison with traditional centralized pseudonymization approaches and does not require a trusted third party. The only drawback in comparison with central pseudonymization is the directed feedback of accidental findings to individual participants, as this is not possible with a quasi-anonymous storage of patient data.

On checking skeptical and ideal admissibility in abstract argumentation frameworks

Abstract argumentation frameworks (afs) are directed graphs with vertices being abstract arguments and edges denoting the attacks between them. Within the context of afs, we implement and evaluate an algorithm for two essential computational problems: checking skeptical and ideal admissibility. We evaluate the implemented algorithms using a widely-known benchmark. In terms of the number of solved problem instances and the average running time, our implementation outperforms two prominent systems.

An average-case sublinear forward algorithm for the haploid Li and Stephens model

BackgroundHidden Markov models of haplotype inheritance such as the Li and Stephens model allow for computationally tractable probability calculations using the forward algorithm as long as the representative reference panel used in the model is sufficiently small. Specifically, the monoploid Li and Stephens model and its variants are linear in reference panel size unless heuristic approximations are used. However, sequencing projects numbering in the thousands to hundreds of thousands of individuals are underway, and others numbering in the millions are anticipated.ResultsTo make the forward algorithm for the haploid Li and Stephens model computationally tractable for these datasets, we have created a numerically exact version of the algorithm with observed average case sublinear runtime with respect to reference panel size k when tested against the 1000 Genomes dataset.ConclusionsWe show a forward algorithm which avoids any tradeoff between runtime and model complexity. Our algorithm makes use of two general strategies which might be applicable to improving the time complexity of other future sequence analysis algorithms: sparse dynamic programming matrices and lazy evaluation.

Fast and fully-automated detection and segmentation of pulmonary nodules in thoracic CT scans using deep convolutional neural networks

Deep learning techniques have been extensively used in computerized pulmonary nodule analysis in recent years. Many reported studies still utilized hybrid methods for diagnosis, in which convolutional neural networks (CNNs) are used only as one part of the pipeline, and the whole system still needs either traditional image processing modules or human intervention to obtain final results. In this paper, we introduced a fast and fully-automated end-to-end system that can efficiently segment precise lung nodule contours from raw thoracic CT scans. Our proposed system has four major modules: candidate nodule detection with Faster regional-CNN (R-CNN), candidate merging, false positive (FP) reduction with CNN, and nodule segmentation with customized fully convolutional neural network (FCN). The entire system has no human interaction or database specific design. The average runtime is about 16 s per scan on a standard workstation. The nodule detection accuracy is 91.4% and 94.6% with an average of 1 and 4 false positives (FPs) per scan. The average dice coefficient of nodule segmentation compared to the groundtruth is 0.793.

Mathematical models for stable matching problems with ties and incomplete lists

We present new integer linear programming (ILP) models for NP-hard optimisation problems in instances of the Stable Marriage problem with Ties and Incomplete lists (SMTI) and its many-to-one generalisation, the Hospitals/Residents problem with Ties (HRT). These models can be used to efficiently solve these optimisation problems when applied to (i) instances derived from real-world applications, and (ii) larger instances that are randomly-generated. In the case of SMTI, we consider instances arising from the pairing of children with adoptive families, where preferences are obtained from a quality measure of each possible pairing of child to family. In this case, we seek a maximum weight stable matching. We present new algorithms for preprocessing instances of SMTI with ties on both sides, as well as new ILP models. Algorithms based on existing state-of-the-art models only solve 6 of our 22 real-world instances within an hour per instance, and our new models incorporating dummy variables and constraint merging, together with preprocessing and a warm start, solve all 22 instances within a mean runtime of a minute. For HRT, we consider instances derived from the problem of assigning junior doctors to foundation posts in Scottish hospitals. Here, we seek a maximum size stable matching. We show how to extend our models for SMTI to HRT and reduce the average running time for real-world HRT instances by two orders of magnitude. We also show that our models outperform by a wide margin all known state-of-the-art models on larger randomly-generated instances of SMTI and HRT.

A Method for Broccoli Seedling Recognition in Natural Environment Based on Binocular Stereo Vision and Gaussian Mixture Model.

Illumination in the natural environment is uncontrollable, and the field background is complex and changeable which all leads to the poor quality of broccoli seedling images. The colors of weeds and broccoli seedlings are close, especially under weedy conditions. The factors above have a large influence on the stability, velocity and accuracy of broccoli seedling recognition based on traditional 2D image processing technologies. The broccoli seedlings are higher than the soil background and weeds in height due to the growth advantage of transplanted crops. A method of broccoli seedling recognition in natural environments based on Binocular Stereo Vision and a Gaussian Mixture Model is proposed in this paper. Firstly, binocular images of broccoli seedlings were obtained by an integrated, portable and low-cost binocular camera. Then left and right images were rectified, and a disparity map of the rectified images was obtained by the Semi-Global Matching (SGM) algorithm. The original 3D dense point cloud was reconstructed using the disparity map and left camera internal parameters. To reduce the operation time, a non-uniform grid sample method was used for the sparse point cloud. After that, the Gaussian Mixture Model (GMM) cluster was exploited and the broccoli seedling points were recognized from the sparse point cloud. An outlier filtering algorithm based on k-nearest neighbors (KNN) was applied to remove the discrete points along with the recognized broccoli seedling points. Finally, an ideal point cloud of broccoli seedlings can be obtained, and the broccoli seedlings recognized. The experimental results show that the Semi-Global Matching (SGM) algorithm can meet the matching requirements of broccoli images in the natural environment, and the average operation time of SGM is 138 ms. The SGM algorithm is superior to the Sum of Absolute Differences (SAD) algorithm and Sum of Squared Differences (SSD) algorithms. The recognition results of Gaussian Mixture Model (GMM) outperforms K-means and Fuzzy c-means with the average running time of 51 ms. To process a pair of images with the resolution of 640×480, the total running time of the proposed method is 578 ms, and the correct recognition rate is 97.98% of 247 pairs of images. The average value of sensitivity is 85.91%. The average percentage of the theoretical envelope box volume to the measured envelope box volume is 95.66%. The method can provide a low-cost, real-time and high-accuracy solution for crop recognition in natural environment.