In order to overcome the disadvantages of conventional sliding mode reaching law, such as large chattering and slow convergence rate, an improved quick law is proposed. The composed two terms which can, respectively, play leading role when system far away from or near surface. Thus, can arrive at surface with faster rate beginning end. Some other advantages converging in a finite time, second-order characteristic are proved. Furthermore, speed up equilibrium point along surface, global...

In modern radar detection systems, the particle filter technique has become one of core algorithms for real-time target and tracking due to its good nonlinear non-Gaussian system state estimation capability. However, when dealing with complex dynamic scenes, traditional algorithm exposes obvious deficiencies. The main expression is that sample degradation serious, which leads a decrease in accuracy. multi-target states, difficult effectively distinguish stably track each target, increases...

An adaptive ant colony algorithm is proposed to overcome the premature convergence problem in conventional algorithm. The composed of three groups ants: ordinary ants, abnormal ants and random ants. Each searches path with high concentration pheromone at probability, each low randomly regardless concentration. Three provide a good initial state trails together. As optimization calculation goes on, number decreases gradually. In late stage, all transform which can rapidly concentrate optimal...

Gait energy image (GEI) preserves the dynamic and static information of a gait sequence. The common includes appearance shape human body variation frequency phase. However, there is no consideration time that normalizes each silhouette within GEI. As regards this problem, paper proposed accumulated frame difference (AFDEI), which can reflect characteristics. fusion moment invariants extracted from GEI AFDEI was selected as feature. Then, recognition accomplished using nearest neighbor...

In order to improve the detection performance of radar constant false alarm detector in a multiple-target environment, Kaigh–Lachenbruch Quantile rate based on composite fuzzy fusion rules (CFKLQ-CFAR) is designed by combining and detector. Two sensors are used collect environmental information, membership function value calculated collected information. Furthermore, presence or absence target judged compositely four rules. CFKLQ-CFAR applied variability index CFAR (VI-CFAR) detector, an...

ABSTRACT To enhance the robustness of microgrid inverter system in islanded operation mode and speed up response system, a novel voltage control strategy based on improved sliding (SMC) composite nonlinear feedback (CNFC) is proposed. This specifically designed for disconnected from public grid. achieve fast response, an attenuation function to adjust strength between linear terms CNFC under different conditions. improve anti‐interference ability suppress chattering phenomenon, adaptive...

Fingerprint classification is an important indexing scheme to reduce fingerprint matching time for a large database efficient large-scale identification. The abilities of Curvelet transform capturing directional edges images make the suitable be classified higher accuracy. This paper presents algorithm combining (CT) and gray-level cooccurrence matrix (GLCM). Firstly, we use fast discrete warping (FDCT_WARPING) decompose original image into five scales coefficients construct filter by...

In target tracking, the complex background often has a negative effect on tracking quality. The feature extraction of moving can be extracted in order to suppress interference caused by tracking. An improved Camshift method is proposed this paper, gauss weight function chosen select area that called region interest and intercept from background, back projection map tracked independently, thus eliminating object. compared with traditional algorithm based multi fusion, experimental results...

Because the corn vein and noise influence contour extraction of maize leaf disease, we put forward a new recognition algorithm based on Curvelet Shape Context (SC). This method can improve speed accuracy disease recognition. Firstly, use Seeded Regional Growing (SRG) to segment image. Secondly, Modulus Correlation (CMC) is extract effective disease. Thirdly, combine CMC with SC obtain histogram features then these calculate similarities between template image target Finally, adopt n -fold...

In order to improve the control performance of global sliding mode method, a fast method is proposed accelerate response system by changing exponential decay function in surface as an bilateral which can make dynamic evolve into linear finite time. The reaching law used design law, and Lyapunov stability theory prove system. this article be applied uncertain nonlinear Simulation results show that has faster rate than conventional method. quad-rotor unmanned helicopter, its good...

The accuracy of tracking based on Camshift would decrease due to the similarity between target color and background or is obscured. For above problems, improved algorithm proposed in this paper. by using contour features target, search window updated according feature target. Thus, interference strong light weakened. Kalman filtering used predict motion state enhancing efficiency when Experiments show that combined with make more effectively under conditions background. And position

In order to improve the noise reduction performance and clarity of denoising images, a composite convolutional neural network composed autoencoder feature reconstruction is proposed. Multiple layers are added into extract image information performance, designed recover texture detail image. The cross-connected structure used fuse in network. Experimental results show that proposed method has better than existing methods for different intensity. More could be retained, clearer images obtained.

The border regression is a key technique of the regional convolution neural network (CNN) to locate target. However, it relies on label information large number sample data. Therefore, inefficient generate training set, and location target also inaccurate. For this, novel detection method based CNN particle search proposed. A small probe particles are generated roughly used extract image features, determine probability, recognize pattern searching placed near region where features detected...