Search

Medical image segmentation is essential in medical image analysis since it can provide reliable assistance in computer-aided clinical diagnosis, treatment planning, and intervention. Although deep learning algorithms based on CNNs and Transformers have made notable progress in medical image segmentation, it is still challenging owing to the objects with complex structures, low discrimination and differences between individuals. To alleviate the problems, we propose a novel 3D medical image segmentation network based on Transformers and CNNs combining morphological information and reticular mechanism. Firstly, the morphological constraint stream is designed to learn the prior shape information based on the CNN model for enhancing the interpretability of the ultimate trained model and accelerating the convergence. Secondly, the Reticular Transformer is utilized to obtain multi-scale information based on the Transformer, which can bind the local texture information and underlying semantic information to further acquire the feature maps with sufficient details and receptive field. The experiments demonstrate that our proposed method outperforms many existing segmentation models in terms of the performance in metrics DSC and HD (80.46\% in DSC on the Synapse dataset and 90.83\% in DSC on the ACDC dataset). The code will be released at https://github.com/rocklijun/MCRformer. Our proposed method can not only achieve superior performance compared with most of the current state-of-the-art methods, but also enhance the robustness and interpretability of the model. Furthermore, the proposed morphological constraint stream has the potential to be transferred to other frameworks for different medical image analysis tasks.

In this letter, we propose a remote sensing image matching method that is simple yet efficient to deal with different deformations. Inspired by the region growing strategy used in image segmentation, we integrate the motion consistency into the general region growing pipeline from a novel perspective. Specifically, we first obtain a subset with a high ratio inlier as the seed correspondence set. Then, to find more reliable correspondences, we formulate the motion consistency into the correspondence growing criterion, which is general to be suitable to many remote sensing applications. Extensive experimental results on the public available remote sensing data set show that our method achieves the best performance compared with state-of-the-art methods.

Medical image segmentation is essential in medical image analysis since it can provide reliable assistance in computer-aided clinical diagnosis, treatment plan-ning, and intervention. Although deep learning algorithms based on CNNs and Transformers have made notable progress in conventional image segmentation, medical image segmentation is still challenging owing to limited ability to learn the prior information about the target shapes, especially for large target areas, which have the problems of rough boundaries in the segmentation results, such as over-segmentation, under-segmentation and internal holes. To alleviate these issues, we propose a novel 3D medical image segmentation network based on Transformer Blocks and the CNN model by fusing morphological information, and design a Reticular Transformer (RT) by the reticular mechanism. Specifically, firstly, the morphological constraint stream is designed to learn the prior shape structure information based on the CNN model. Secondly, the Reticular Transformer is utilized to obtain multi-scale information based on the Transformer, which can further utilize the local texture information and underlying semantic information to obtain sufficient details and receptive field. The experiments on the datasets demonstrate that our proposed method outperforms many existing segmentation models in terms of the performance in metrics DSC and HD (80.46\% in DSC on the Synapse dataset and 90.83\% in DSC on the ACDC dataset). The code will be released at https://github.com/rocklijun/MCRformer.

Surrogate-assisted evolutionary algorithms (SAEAs) are potential approaches to solve computationally expensive optimization problems. The critical idea of SAEAs is to combine the powerful searching capabilities of evolutionary algorithms with the predictive capabilities of surrogate models . In this study, an efficient surrogate-assisted hybrid optimization (SAHO) algorithm is proposed via combining two famous algorithms, namely, teaching–learning-based optimization (TLBO) and differential evolution (DE). The TLBO is focused on global exploration and the DE is concentrated on local exploitation. These two algorithms are carried out alternately when no better candidate solution can be found. Meanwhile, a new prescreening criterion based on the best and top collection information is introduced to choose promising candidates for real function evaluations. Besides, two evolution control (i.e., the generation-based and individual-based) strategies and a top-ranked restart strategy are integrated in the SAHO. Moreover, a local RBF surrogate which does not need too many training samples is employed to model the landscapes of the target function. Sixteen benchmark functions and the tension/compression spring design problem are adopted to compare the proposed SAHO with other state-of-the-art approaches. Extensive comparison results demonstrate that the proposed SAHO has superior performance for solving expensive optimization problems.

Feature matching is one of the key techniques in many vision-based tasks, which aims to establish reliable correspondences between two sets of features. In this paper, we present a new feature matching method, which formulates the matching of two feature sets as a mathematical model based on two common consistency constraints. We first propose an advanced consensus of neighborhood topology, which can better exploit the consensus of topological structures to identify inliers. In order to have reliable neighborhood information for the feature points, a subset with high percentage inliers obtained by a guided matching strategy from the putative matches for the neighborhood construction is used. We demonstrate the advantages of our proposed method on various real image pairs. The results demonstrate that the proposed method is superior to the state-of-the-art feature matching methods.

In this paper, a new robust model fitting method is proposed to efficiently segment multistructure data even when they are heavily contaminated by outliers. The proposed method is composed of three steps: first, a conventional greedy search strategy is employed to generate (initial) model hypotheses based on the sequential "fit-and-remove" procedure because of its computational efficiency. Second, to efficiently generate accurate model hypotheses close to the true models, a novel global greedy search strategy initially samples from the inliers of the obtained model hypotheses and samples subsequent data subsets from the whole input data. Third, mutual information theory is applied to fuse the model hypotheses of the same model instance. The conventional greedy search strategy is used to generate model hypotheses for the remaining model instances, if the number of retained model hypotheses is less than that of the true model instances after fusion. The second and the third steps are performed iteratively until an adequate solution is obtained. Experimental results demonstrate the effectiveness and efficiency of the proposed method for model fitting.

The water supply network is of great importance for a city, and it faces various risks such as leaks and breaks, so its risk level needs to be evaluated regularly. To propose a new and effi...

With the development of Internet of vehicles, the information exchange between vehicles and the outside world results in a higher risk of external network attacks to the vehicles. The attack modes to the most widely used vehicle-mounted CAN bus are complex and diverse, but most of the intrusion detection approaches proposed by now can only detect one type of attack at a time. Aiming at detecting multi-types of attacks using a single model, we proposed a detection method based on the Mosaic-coded convolution neural network for intrusions containing various combinations of attacks with multi-classification capability. A Mosaic-like two-dimensional data grid was created from the one-dimensional CAN ID for the CNN to effectively extract the data features and maintain the time connections between the CAN IDs. Four types of attacks and all possible combinations of them were used to train and test our model. The autoencoder was also used to reduce the dimensionality of the data so as to cut down the model’s complexity. Experimental results showed that the proposed method was effective in detecting all types of attack combinations with high and stable multi-classification ability.

Robust model fitting is an important task for modern electronic industries. In this paper, an efficient robust model-fitting method is proposed to estimate model hypotheses for multistructure data with high outlier rates. The proposed method consists mainly of two steps. First, an improved greedy search strategy is used to generate model hypotheses. Different from the conventional greedy search strategy that always initializes its model hypotheses randomly, the improved greedy search strategy may initialize its model hypotheses by using the inliers of the current best hypotheses for generating more accurate hypotheses. Second, on the basis of the improved greedy search strategy and a specified inlier threshold, a novel parameter detector is used to detect whether the parameters of the generated hypotheses are correct. If they are correct, then the proposed method finishes its fitting process. Otherwise, the first and second steps are performed again. Experimental results on the AdelaideRMF and Hopkins 155 datasets revealed that the proposed method outperformed several state-of-the-art model-fitting methods, including the method based on the conventional greedy search strategy.

We present a novel algorithm (DAM) for deterministic motion trajectory segmentation by using epipolar geometry and adaptive kernel-scale voting. DAM is based on geometric models and exploits the information derived from superpixels to deterministically construct a set of initial correlation matrices. Then DAM introduces a novel adaptive kernel-scale voting scheme to measure each initial correlation matrix. After that, based on the voting scores, the set of initial correlation matrices is accumulated to generate a discriminant affinity matrix, which is utilized for final grouping. The key characteristic of the DAM is its deterministic nature, i.e., DAM is able to achieve reliable and consistent performance for motion trajectory segmentation without randomness. Experimental results on both several traditional datasets (i.e., Hopkins155, Hopkins12, and MTPV62 datasets) and a more realistic and challenging dataset (i.e., KT3DMoSeg) show the significant superiority of the proposed DAM over several state-of-the-art motion trajectory segmentation algorithms with respect to segmentation accuracy.

As we known, most of convolution neural architectures are manually designed. However, they cannot obtain the optimal structures. To address this problem, based on Weighted Feature Pyramid Networks (WFPN), in this paper, we use gaussian kernel to calculate the weight to design a novel method called the Neural Architecture Search Weighted Feature Pyramid Networks (i.e., NAS-WFPN). NAS-WFPN mainly consists of three parts (i.e., top-down pathway, bottom-up pathway and lateral connections) to fuse features across different scales. Experimental results show that NAS-WFPN achieves higher accuracy compared with the existing object detection methods. Specifically, NAS-WFPN increases accuracy by 2.3 AP compared to SSDLite with MobileNetV2 model and gets 49.1 AP, which exceeds NAS-FPN and Mask R-CNN.

Feature Pyramid Networks (FPN) is a popular feature extraction. However, FPN and its variants do not investigate the influence of resolution information and semantic information in the object detection. Thus, FPN and its variants cannot detect some objects on challenging images. In this paper, based on FPN, we propose to use gaussian kernel function to assign different weight values to semantic information and resolution information for different images in the object detection. The proposed method, is called a Weighted Feature Pyramid Network (WFPN), and shows significant improvement over the traditional feature pyramids in several applications. Using WFPN in Faster R-CNN system, the proposed method achieves better performance on the PASCAL detection benchmark.

The key to computer-based image recognition is to distinguish salient objects from the image background. However, it is still challenging to detect salient region when an object significantly touches the image boundaries. In this study, we present a novel salient region detection method based on a color space volume and a novel weighted saliency probability to address the issue. First, we propose a novel color space volume, and it was regarded as the foreground based on the L^*a^*b^* color space. Second, we present a new background measure called background probability to find background regions by exploiting a background prior and centroid distance weights. Moreover, we propose a new foreground measure called foreground probability to detect foreground by utilizing the brightness of color space volume. Finally, we propose a novel weighted saliency probability to obtain a clean and uniform salient map based on the background probability and the foreground probability. Experiments on three benchmark image datasets demonstrated that the proposed method outperformed several well-known saliency detection methods.

Non-rigid point set registration is a fundamental problem in many fields related to computer vision, medical image processing, and pattern recognition. In this paper, we develop a new point set registration method by using an adaptive weighted objective function, which formulates the alignment of two point sets as a mixture model estimation problem. The correspondences and the transformation are jointly recovered by using the expectation–maximization algorithm to obtain the promising results. First, the correspondences are established using local feature descriptors, and the adaptation parameters for the mixture model are computed from these correspondences. Then, the underlying transformation is recovered by minimizing the adaptive weighted objective function deduced from the mixture model. We demonstrate the advantages of the proposed method on various types of synthetic and real data and compare the results against those obtained using the state-of-the-art methods. The experimental results show that the proposed method is robust and outperforms the other registration approaches.

Motion segmentation is an important task for intelligent transportation systems. In this paper, inspired by the fact that a feature point trajectory can be sparsely represented as a combination of several feature point trajectories that share coherent transformations, an efficient and effective motion segmentation method with a sparsity constraint is proposed. Specifically, we first propose an accumulated scheme to efficiently integrate motion information from all the frames of a video sequence to construct a correlation matrix. Then, a sparse affinity matrix is built on the correlation matrix by using information-theoretic principles, where the nonzero elements in the same row of the sparse affinity matrix correspond to the feature point trajectories more likely belonging to the same motion. Thereafter, a segment and merge procedure is proposed to effectively estimate the number of motions via the sparse affinity matrix. Finally, by applying spectral clustering on the sparse affinity matrix, different motions in the video sequence are accurately segmented based on the estimated number of motions. Experimental results on the Hopkins 155 and the 62-clip datasets demonstrate that the proposed method achieves superior performance compared with several state-of-the-art methods.

Generating promising hypotheses plays a critical role in the success of multi-structure model fitting methods. However, conventional multi-structure hypothesis generation strategies do not exploit the information derived from the results of model selection to guide the subsequent hypothesis generation process. This leads to the problem that these hypothesis generation strategies are often computationally expensive for generating promising hypotheses, especially for heavily contaminated multi-structure data. To address this problem, we first propose a guided sampling strategy to accelerate promising hypothesis generation process by using information derived from the results of model selection on the fly. Then we present a Unified Hypothesis Generation (UHG) framework, which effectively combines the conventional multi-structure hypothesis generation strategy with the proposed guided sampling strategy by using a Markov Chain Monte Carlo process based on a cooling schedule. Experimental results on public databases demonstrate that the proposed UHG achieves significant superiority over several state-of-the-art sampling methods in terms of accuracy and efficiency, especially on multi-structure data. HighlightsA unified sampling framework UHG is proposed for multi-structure model fitting.UHG combines the merits of a global search strategy and a local search strategy.The global search strategy is guided by the conventional sampling strategy.The local search strategy is guided by using the results of model selection.Experimental results show the effectiveness of UHG on public datasets.

A sampling method EGHG is proposed for multi-structure geometry estimation.EGHG combines the benefits of a global and a local sampling strategy.The global sampling strategy is designed to rapidly obtain promising solutions.The local sampling strategy is designed to efficiently achieve accurate solutions.Experimental results show the effectiveness of EGHG on public real image pairs. We propose an Efficient Guided Hypothesis Generation (EGHG) method for multi-structure epipolar geometry estimation. Based on the Markov Chain Monte Carlo process, EGHG combines two guided sampling strategies: a global sampling strategy and a local sampling strategy. The global sampling strategy, guided by using both spatial sampling probabilities and keypoint matching scores, rapidly obtains promising solutions. The spatial sampling probabilities are computed by using a normalized exponential loss function. The local sampling strategy, guided by using both Joint Feature Distributions (JFDs) and keypoint matching scores, efficiently achieves accurate solutions. In the local sampling strategy, EGHG updates a set of current best hypothesis candidates on the fly, and then computes JFDs between the input data and these candidates. Experimental results on public real image pairs show that EGHG significantly outperforms several state-of-the-art sampling methods on multi-structure data.

The performance of many robust model fitting techniques is largely dependent on the quality of the generated hypotheses. In this paper, we propose a novel guided sampling method, called accelerated guided sampling (AGS), to efficiently generate the accurate hypotheses for multistructure model fitting. Based on the observations that residual sorting can effectively reveal the data relationship (i.e., determine whether two data points belong to the same structure), and keypoint matching scores can be used to distinguish inliers from gross outliers, AGS effectively combines the benefits of residual sorting and keypoint matching scores to efficiently generate accurate hypotheses via information theoretic principles. Moreover, we reduce the computational cost of residual sorting in AGS by designing a new residual sorting strategy, which only sorts the top-ranked residuals of input data, rather than all input data. Experimental results demonstrate the effectiveness of the proposed method in computer vision tasks, such as homography matrix and fundamental matrix estimation.

This paper proposes a robust model fitting method, called Outliers Removed via Spectral Clustering (ORSC), to estimate multiple inlier structures in the presence of a large number of outliers. The basic idea is to cast each data point to the conceptual space, where the distance distribution of inliers and outliers from the origin is significantly different. Therefore, all the points can be classified into inliers and outliers according to the distribution of points of each subspace, which is obtained by a spectral clustering algorithm. Furthermore, we can use the clustering result to guide the follow-up sampling to get more clean data points for hypotheses generation when handling the complex multi-structure model with a large proportion of outliers. Experimental results show that our method achieves superior performance compared with some state-of-the-art methods in terms of line fitting, circle fitting, and real images.

Efficient hypothesis generation plays an important role in robust model fitting. In this study, based on the combination of residual sorting and local constraints, we propose an efficient guided hypothesis generation method, called Rapid Hypothesis Generation (RHG). By exploiting the local constraints to guide the hypothesis generation process, RHG raises the probability of generating promising hypotheses and reduces the computational cost during hypotheses generation. Experimental results on homography and fundamental matrix estimation show that RHG can effectively guide hypothesis generation process and rapidly generate promising hypotheses for heavily contaminated multi-structure data.

In this paper, we propose a novel hypergraph based method (called HF) to fit and segment multi-structural data. The proposed HF formulates the geometric model fitting problem as a hypergraph partition problem based on a novel hypergraph model. In the hypergraph model, vertices represent data points and hyperedges denote model hypotheses. The hypergraph, with large and "data-determined" degrees of hyperedges, can express the complex relationships between model hypotheses and data points. In addition, we develop a robust hypergraph partition algorithm to detect sub-hypergraphs for model fitting. HF can effectively and efficiently estimate the number of, and the parameters of, model instances in multi-structural data heavily corrupted with outliers simultaneously. Experimental results show the advantages of the proposed method over previous methods on both synthetic data and real images., Comment: Pattern Recognition, 2016