International Journal of Pattern Recognition and Artificial Intelligence

Fire detection technology aroused people’s attention increasingly. The main challenge of the fire detection systems is how to reduce false alarms caused by objects like fire’s colors. Most existing algorithms used only features of fire in visual field. In this work, we put forward a new algorithm to detect dynamic fire from the surveillance video based on the combination of radiation domain features model. First, a fire color model is used to extract flame-like pixels as candidate areas in YCbCr space. Second, we convert the candidate regions from the traditional color space into radiation domain in advance by camera calibration. And we use seven features to model the spectral spatio-temporal model of the fire to more accurately characterize the physical and optical properties of the fire. Finally, we choose a two-class SVM classifier to identify the fire from the candidate areas and use a radial basis function kernel to improve the accuracy of the recognition. Two different sets of data are used to validate the algorithm we proposed. And the experimental results indicate that our method performs well in video fire surveillance.

In this paper, we proposed an all-weather flame detection algorithm which could make full use of active infrared cameras presently installed in many public places for surveillance purposes. Firstly, according to the different spectral imaging results in day and night, we propose a video type classification algorithm (VTCA) via imaging clues. VTCA could help us select different flame visual features in color image and infrared image. Secondly, we use a generic YCbCr-color-space-based chrominance model to extract regions of interest (ROI) of flame. Thirdly, two flame dynamic features are used to verify the candidate ROIs, which are common flame flicker feature and an improved block-based PCA in consecutive frames. The experimental results show that the proposed flame detection model has been successfully applied to various situations, including day and night, indoor and outdoor on our test video datasets, and it gives a better performance compared with other state-of-the-art methods.

Camera can sensor the environment on the lane by extracting the lane lines, but such detection is limited to a short distance with effect of illumination and other factors; radar can detect objects a long distance away but cannot detect the lane conditions. This paper combined machine vision with millimeter-wave radar and extracted the nearby distinct lane line through images; at the same time, the radar obtained the motion trajectory information of distant vehicles, then the least-square method was used to make curve fitting on those motion trajectory information in order to reconstruct the lane line information. Finally, in the stage of fusing two segments of lane lines, the goodness of fit was applied to complete the matching of corresponding lane lines. While, for areas between two segments of lane lines that neither camera or radar can detect, we established a lane model, utilized probabilistic neural network to select the corresponding lane model for matching, and then used approximate mathematics expression according to the selected lane model, thus obtaining the final front road information of current vehicle.

A new method based on eye-tracking data — visual momentum (VM) — was introduced to quantitatively evaluate a dynamic interactive visualization interface. We extracted the dimensionless factors from the raw eye-tracking data, including the fixation time factor $T$, the saccade amplitude factor D and the fixation number factor $N$. A predictive regression model of VM was deduced by eye movement factors and the performance response time (RT). In Experiment 1, the experimental visualization materials were designed with six effectiveness levels according to design techniques proposed by Woods to improve VM (total replacement, fixed format data replacement, long shot, perceptual landmark and spatial representation) and were tested in six parallel subject groups. The coefficients of the regression model were calculated from the data of 42 valid subjects in Experiment 1. The mean VM of each group exhibited an increasing trend with an increase in design techniques. The data of the performance and eye tracking among the combined high VM group, middle VM group and low VM group indicated significant differences. The data analysis indicates that the results were consistent with the previous qualitative research of VM. We tested and verified this regression model in Experiment 2 with another dynamic interactive visualization. The results indicated that the VM calculated by the regression model was significantly correlated with the performance data. Therefore, the virtual parameter VM can be a quantitative indicator for evaluating dynamic visualization. It could be a useful evaluation method for the dynamic visualization in general working environments.

In independent component analysis (ICA) and its diverse algorithms, uniqueness is the most essential requirement and rationality problem compared with performances of existence, stability and convergence. For ICA’s maximum non-Gaussianity estimation (MNE), many achievements have been made in recent twenty years based on uniqueness assumption, which has been taken for granted all along except for some intuitive interpretation. From the perspective of constrained cost function optimization, the paper is to provide a mathematical proof for uniqueness principle in MNE. The research focuses on skewless assumption and kurtosis-based cost function with basic linear ICA model. The relationship between the Kuhn–Tucker (K–T) points of cost function and the local maxima of non-Gaussianity are derived with the help of constrained optimization theory, and then a conclusion is drawn that there is a one-to-one correspondence between independent components and the local maxima, i.e. maximum non-Gaussianity is the sufficient and necessary condition for independent sources recovery. Moreover, the result also leads to an alternative and straightforward approach to the proof of Xu’ one-bit-matching conjecture for the availability of multi-unit approaches.

Three-dimensional reconstruction of teeth plays an important role in the operation of living dental implants. However, the tissue around teeth and the noise generated in the process of image acquisition bring a serious impact on the reconstruction results, which must be reduced or eliminated. Combined with the advantages of wavelet transform and bilateral filtering, this paper proposes an image denoising method based on the above methods. The method proposed in this paper not only removes the noise but also preserves the image edge details. The noise in high frequency subbands is denoised using a locally adaptive thresholding and the noise in low frequency subbands is filtered by the bilateral filtering. Peak signal-to-noise ratio (PSNR), structural similarity index measure (SSIM) and 3D reconstruction using the iso-surface extraction method are used to evaluate the denoising effect. The experimental results show that the proposed method is better than the wavelet denoising and bilateral filtering, and the reconstruction results meet the requirements of clinical diagnosis.

In the actual cotton planting environment, rapid change of light within a day, reflection from different backgrounds and different weather conditions can affect the imaging of cotton. Therefore, the crop object segmentation is difficult. Images which were captured in 12 natural scenes during cotton planting, including three weather conditions, such as sunny, cloudy and rainy and four soil cover conditions, such as white mulch film, black mulch film, straw and bare soil were regarded as the research objects. This paper presents the cotton leaf segmentation method based on Immune algorithm and pulse coupled neural networks (PCNN). First, 17 color components of white mulch film, black mulch film, straw, bare soil and cotton under the conditions of sunny, cloudy and rainy days were analyzed by using statistical method. Three high feasible and anti-light color components were selected by histogram statistical with mean gray value. Second, the optimal parameters of PCNN model and the optimal number of iterations were determined by using immune algorithm optimization theory, and the method in this paper was tested by using 1200 cotton images which were captured under 12 natural scenes. Finally, the test results showed that this method can distinguish cotton target region from soil and other background regions. Meanwhile, for reflection of mulch film, crop shadow, dark light, complex background, noise, etc. which are often appeared in natural scene, four image segmentation methods of Otsu algorithm, $K$-Means algorithm, FCM algorithm and PCNN were compared with the proposed method in this paper. The segmentation result showed that the proposed method has good resistance to change of light and complex background. The average $M_E$ of the proposed method is 6.5%, significantly lower than that of other four methods and the performance is better than other four methods. This method can segment cotton images in different weather conditions and different backgrounds accurately under complex natural conditions. It will contribute to the subsequent growth status determination and pest diagnosis of cotton.

Image haze removal has been extensively studied, but there has been no such an image database regarding the haze level. It is not convenient for readers to verify the assumptions or priors that are supposed to be useful for haze removal, and meanwhile, it is not fair to compare the performance of haze removal methods, which are effective for images with different haze levels. To solve this problem, we built a database consisting of more than 3464 images of different kinds of outdoor scenes. The images of the database are grouped into four classes regarding the haze level. Along with the database, we also observe a frequency magnitude prior, i.e. the frequency magnitude decreases with the increasing haze level, which can be used as a prior to develop haze removal methods. Our purpose is to help develop image haze removal methods, as well as verify existing statistical priors and discover new ones that can be used for image processing.

In order to improve the accuracy of image segmentation, an improved adaptive level set method is proposed based on level set evolution without re-initialization method and adaptive distance preserving level set evolution method. A new definition of weight coefficient in evolution equations is the main innovation of this paper. The improved method can detect certain object boundaries, interior and exterior contours of an object, edges of multi-objects and weak boundaries of an object by synthetic and real images numerical experiments. Numerical results show that the improved adaptive level set method has faster segmentation speed and higher segmentation accuracy compared with the previous two methods, especially in weak boundaries and edges of multi-objects segmentation problems.

A change and maintenance method is proposed based on the change propagation model and the procedure model information for solving data maintenance problem of 3D machining procedure model change and to help improve the flexibility of 3D machining procedure model and the reliability of the change result. Design Structure Matrix (DSM) is established by analyzing the relationships between machining features in the machining process route to obtain all possible propagation paths. On the basis of obtained paths, machining features that may be affected and machining procedure models related to machining features are stored by the structured method. Algorithms to solve the problems of adding, deleting and modifying machining features are proposed to realize the change and maintenance of 3D machining procedure model by combining machining procedure model’s information in the machining process route. In the end, some numerical examples are given to explain both rationality and feasibility of the proposed approaches.

Manifold learning and classifiers based on sparse representation are widely used in pattern recognition. Most of the conventional manifold learning methods are subjected to the choice of parameters. In this paper, we present a Regularized Locality Projection based on Sparsity Discriminant Analysis (RLPSD) method for Feature Extraction (FE) to understand the high-dimensional data such as face images. In RLPSD, firstly, we show the sparse representation of training samples by collaborative representation-based classification (CRC). Secondly, the idea of part optimization based on sparse representation is used to ensure the within-class compactness which combines with the labels of measurements and the weights of sparse presentation can be as small as possible. Finally, whole optimization can be directly obtained without the iteration of local optimization. Meanwhile, the separability information of between-class can be well discriminated by scatter matrix which is similar to Fisher linear discriminant analysis (LDA). The great recognition performance of the proposed method is verified by comparing with the popular algorithms on Yale, ORL, AR and Extended YaleB face databases and Oxford 102 flowers dataset.

In this paper, combined with physiological anatomy knowledge, the complete finite element model of L3–L4 lumbosacral segment of human lumbar was established by using the 3D graphic of human spine L3–L4 segment, which is obtained by image diagnosis technique (CT scan). This model includes the sections of the bones, the intervertebral disc and the anterior ligament, the posterior ligament, the ligamentum flavum, the fibrous ring and other major spine attached soft tissue. Then the finite element model meshing and the material properties of the corresponding part setting were done on the constructed model, and different loads and boundary conditions were imposed to simulate the displacement and stress and strain nephogram of the normal model and intervertebral disc herniation, senile degeneration and other models in different movement states. And the effectiveness of model data is verified by analyzing its biomechanical properties. The biomechanical properties of the spine obtained by the finite element method can be used to provide biomechanical basis for the diagnosis and treatment of intervertebral disc herniation and degeneration.

Convolutional neural networks (CNNs) have better performance in feature extraction and classification. Most of the applications are based on a traditional structure of CNNs. However, due to the fixed structure, it may not be effective for large dataset which will spend much time for training. So, we use a new algorithm to optimize CNNs, called directly connected convolutional neural networks (DCCNNs). In DCCNNs, the down-sampling layer can directly connect the output layer with three-dimensional matrix operation, without full connection (i.e., matrix vectorization). Thus, DCCNNs have less weights and neurons than CNNs. We conduct the comparison experiments on five image databases: MNIST, COIL-20, AR, Extended Yale B, and ORL. The experiments show that the model has better recognition accuracy and faster convergence than CNNs. Furthermore, two applications (i.e., water quality evaluation and image classification) following the proposed concepts further confirm the generality and capability of DCCNNs.