This paper proposes a two-column convolutional neural network algorithm for face point detection, which proves better performance than single-column. Because of the deep level structure, global texture features are ex...
详细信息
This paper proposes a two-column convolutional neural network algorithm for face point detection, which proves better performance than single-column. Because of the deep level structure, global texture features are extracted at higher layers of the network, and the results of keypoint location remains high accuracy. In the first column, we take the channel component of original image as input to train the neural network respectively and calculate the average location results. In the second one, we use Sobel operator to extract first-order derivative feature of original image and regard it as the input to train another convolutional network. The location results of two columns above are fused with appropriate proportion to ensure better results. Experimental result on LFW dataset shows the performance of the proposed parallel structure better than single convolutional network, and it stays robust due to occlusions, large pose variations and extreme lighting.
Face detection technology is a hot topic in the past recent years. It has been maturely applied to many practical areas. However, the driver face detection is still an open problem to solve. In this paper, we proposed...
详细信息
Face detection technology is a hot topic in the past recent years. It has been maturely applied to many practical areas. However, the driver face detection is still an open problem to solve. In this paper, we proposed an improved method to promote the face detection rate and apply it to the images from the monitoring videos. The first step is to detect the car from the images according to an off-line learning method. Then the method based on additional off-line learning method is the front level for skin color feature in order to correctly detect the driver face. The proposed systems are implemented on the various complicated road environment. The results show that the proposed method improves the efficiency of the driver face detection and is of strong robustness on having glasses, driver head rotation, and lighting change.
In order to discover forest fire as early as possible, forest fire detection should focus on the smoke in early fire. This paper focuses on three key issues: motion segmentation, feature extraction and classifier desi...
详细信息
ISBN:
(纸本)9781509040940
In order to discover forest fire as early as possible, forest fire detection should focus on the smoke in early fire. This paper focuses on three key issues: motion segmentation, feature extraction and classifier design. Background subtraction based on visual background extractor (Vibe) is chosen to divide suspected smoke area when taking into account the accuracy and time consumption. And then do some corresponding morphological processing. Later on, various static and dynamic characteristics of smoke were extracted and different tests were done based on different feature combinations in forest fire smoke detection system. Lots of smoke detecting system only think about static features which will result in a certain degree of misjudgment. Analyzing the false positive rate and recognition rate of these experiments' results, the combination of movement direction, high-frequency energy based on wavelet transformation and compactness is selected to constitute the final recognition vectors. In addition, the continuity which is not mentioned in other researches won't be ignored in this paper. The final experimental results showed that the accuracy rate of this method for smoke detection could reach 92.7%.
This paper investigates the distributed fixed-time consensus problem of networked agents under the connected interaction topology,featuring an upper bound of the convergence time and being independent of initial *** a...
详细信息
ISBN:
(纸本)9781509009107
This paper investigates the distributed fixed-time consensus problem of networked agents under the connected interaction topology,featuring an upper bound of the convergence time and being independent of initial *** achieve fixed-time consensus,a new class of distributed nonlinear consensus protocols are constructed for first-order dynamic agents with or without external *** studies indicate that under the proposed protocols,all states of the agents will converge to a common value in fixed ***,the effectiveness of the theoretical results is illustrated by simulation results.
Mono-crystalline silicon is widely applied in the photovoltaic and the integrated circuit *** power supply,as the key element of the mono-crystalline silicon growth furnace,plays an important role in the performance o...
详细信息
ISBN:
(纸本)9781509009107
Mono-crystalline silicon is widely applied in the photovoltaic and the integrated circuit *** power supply,as the key element of the mono-crystalline silicon growth furnace,plays an important role in the performance of the whole *** this paper,we propose a new kind of two-stage silicon growth furnace heating power supply topology *** first stage three phase rectifier employs the compound active clamp zero voltage soft switching(CACZVS) three-phase power factor correction(PFC) converter to achieve unity power factor and DC bus voltage regulation,and the second stage DC/DC converter employs phase shifted full bridge zero voltage and zero current soft switching converter(FB-ZVZCS) to regulate the output *** with the heating power supply of non-controlled rectifier type or three-phase power factor correction circuit using hard switching,the proposed two-stage converter structure can achieve high power factor,high efficiency and more flexibility of *** simulation results show the correctness of the analysis.
A wavelet packet based adaptive filter-bank construction combined with Deep Belief Network(DBN) feature learning method is proposed for speech signal processing in this paper. On this basis, a set of acoustic features...
详细信息
The problem of assembly job shop scheduling(AJSS) is studied with the objective of minimizing the weighted sum of earliness and tardiness penalties. An insertion search algorithm is proposed to solve the problem. The ...
详细信息
ISBN:
(纸本)9781510830073
The problem of assembly job shop scheduling(AJSS) is studied with the objective of minimizing the weighted sum of earliness and tardiness penalties. An insertion search algorithm is proposed to solve the problem. The operations of job with tardiness or earliness are shifted left or right to optimize the sequence. Simulation results validate the effectiveness of the proposed model and algorithm.
The main benefit of 3D display over 2D display is the obvious ability to create a more lifelike character with high depth sense. However, the limitation of human eye's visual mechanism, unartful 3D scene structure...
详细信息
The main benefit of 3D display over 2D display is the obvious ability to create a more lifelike character with high depth sense. However, the limitation of human eye's visual mechanism, unartful 3D scene structure design, or bad viewing condition always emerges poor depth perception experience or even physiological discomfort during the watching time, which is often sub-optimal for mass high-quality 3D display productions. To solve this problem, we propose a novel 3D display parallel system for depth sense optimization and it empirically guides how the light field should be re-rendered. Structurally, the parallel system consists of an artificial perception measurement system, a display evaluation model and a light field display rendering system, which includes the display calibration, scene capture, light field data processing and display. Particularly, the system can systematically analyze and model various factors affecting the depth sense which learned through the measurement system, like scene structure, objects’ speeds in 3D video and so on. And those sense factors can be personally modified or increased according to the viewer's demands or technical improvement. Moreover, the light field could be real-time re-rendered, based on some image processing technology, optical flow analysis and object segmentation (or tracking) (especially the one-shot video segmentation). Theory and algorithms are developed and experimental validation results show a superior performance.
This paper investigates the pole placement problem of discrete-time systems. To retain the advantages of the parametric discrete-time Lyapunov equation, matrix-partitioning idea is used to derive a new pole shift lemm...
详细信息
This paper investigates the pole placement problem of discrete-time systems. To retain the advantages of the parametric discrete-time Lyapunov equation, matrix-partitioning idea is used to derive a new pole shift lemma. Starting from system matrix transformations, a recursive method that shifts every eigenvalue of a discrete-time system separately without mode decomposition in each step is proposed. Compared with existing methods, this one not only leads to analytical solutions to feedback gains but also may need low computation cost on flexible pole placement. As an application, the semiglobal stabilization with guaranteed flexible pole placement for saturated discrete-time systems is achieved using this method.
Face detection technology is a hot topic in the past recent years. It has been maturely applied to many practical areas. However, the driver face detection is still an open problem to solve. In this paper, we proposed...
详细信息
ISBN:
(纸本)9781509034857
Face detection technology is a hot topic in the past recent years. It has been maturely applied to many practical areas. However, the driver face detection is still an open problem to solve. In this paper, we proposed an improved method to promote the face detection rate and apply it to the images from the monitoring videos. The first step is to detect the car from the images according to an off-line learning method. Then the method based on additional off-line learning method is the front level for skin color feature in order to correctly detect the driver face. The proposed systems are implemented on the various complicated road environment. The results show that the proposed method improves the efficiency of the driver face detection and is of strong robustness on having glasses, driver head rotation, and lighting change.
暂无评论