In order to improve the operation efficiency of fire image recognition neural network, the FPGA hardware acceleration of fire image recognition neural network is studied and implemented. Firstly, with the help of fire image database and TensonFlow machine learning platform, a fire image recognition neural network is trained with VGG19 as the neural network model. Then the FPGA hardware design of convolution layer, pooling layer, full connection layer and activation function is carried out for the trained neural network through Vivado. Secondly, the designed VGG19 fire image recognition convolution neural network accelerator is debugged on the ZYNQ7020 development board. Finally, the acceleration performance of fire identification convolutional neural network accelerator system is tested in three aspects: acceleration efficiency, resource utilization and power consumption. The experimental results show that the accelerator can reduce the clock cycle required by each convolution layer of fire image recognition neural network from one million to ten thousand, the resource utilization meets the chip requirements, and the chip power consumption is reduced to 2.067w. While improving the operation efficiency of neural network, it realizes low power consumption.
In multi-agent systems, agents coordinate their behavior and work together to achieve a shared goal through collaboration. However, in multi-agent systems, selecting qualified participants to form effective collaboration communities is challenging. In this paper, we propose a minimum circle covering algorithm, as a solution for on-demand participant selection for collaboration in multi-agent systems. Furthermore, a twenty-one point FBG sensors are divided into four sensing function agent in Structural Health Monitoring (SHM) system is experimented in an aircraft wing box. Correspondingly, there are four intelligent evaluation agents and one system collaborative agent in the multi-agent intelligent health monitoring system. For the damage loading position prediction on the aircraft wing box, the collaborative participation selection strategy based on the minimum circle coverage is verified experimentally. The research result indicates that the minimum circle covering algorithm can be used to select the participation in multi-agent intelligent health monitoring system, of all the participations in the collaboration, it enables them to identify and select a qualified participants.
In recent years, along with the computer operation speed unending enhancement, the computer is employed to carry on the dangerous cargos the examination and the recognition to obtain the more and more widespread applications. Aiming at the disadvantage of high false detection rate in target classification detection using existing feature training classifiers, the work proposes a detection algorithm for hazardous articles with convolutional neural network on the basis of deep learning. For the image to be checked, sliding windows of different scales are used to determine whether there is an object window. For object detection, a convolutional neural network is trained with a large number of positive and negative samples. In order to better adapt to object detection, the topology of the convolutional neural network is improved. The window of suspected hazardous article is input into the improved convolutional neural network for dangerous object detection, and the false detection rate is reduced while maintaining the original detection rate.
In this paper, a transient plasma is produced by focusing the 1064 nm radiation from a Q-switched Nd:YAG onto the one-yuan coin at room temperature in air at atmospheric pressure. Using the iterative Boltzmann algorithm, the plasma electron temperature of the one-yuan coin is calculated as 28144 K. Experiments show that the correlation coefficient increases from 0.197 to 0.997 as the number of iterations increases. Experimental results show that the laser induced one-yuan coin plasma meets the LTE model.
For the fast touching imaging of the capacitive touch screens (CTS) with a very large size, the conventional methods can only be achieved at the expense of the analog hardware complexity and power consumption. Firstly, the sparse feature of variation within the capacitance of CTS being touched was analyzed. Secondly, the principal for measuring the capacitance of CTS was introduced, and then the scheme of sparse touching imaging was given. Moreover, the hardware implementation for sparse sensing was presented. An algorithm of touching sensing for CTS and its implementation were proposed. The test result indicated the validity of sparse touching imaging for CTS. The proposed method can achieve touching imaging efficiently with lower hardware complexity and lower power consumption.
The conventional compressive sensing for videos based on the non-adaptive linear projections, and the measurement times is usually set empirically. As a result, the quality of videos reconstruction is always affected. Firstly, the block-based compressed sensing (BCS) with conventional selection for compressive measurements was described. Then an estimation method for the sparsity of multi-view videos was proposed based on the two dimensional discrete wavelet transform (2D DWT). With an energy threshold given beforehand, the DWT coefficients were processed with both energy normalization and sorting by descending order, and the sparsity of the multi-view video can be achieved by the proportion of dominant coefficients. And finally, the simulation result shows that, the method can estimate the sparsity of video frame effectively, and provides an active basis for the selection of compressive observation times. The result also shows that, since the selection of observation times is based on the sparsity estimated with the energy threshold provided, the proposed method can ensure the reconstruction quality of multi-view videos.
The conventional compressive sensing works based on the non-adaptive linear projections, and the parameter of its measurement times is usually set empirically. As a result, the quality of image reconstruction is always affected. Firstly, the block-based compressed sensing (BCS) with conventional selection for compressive measurements was given. Then an estimation method for the sparsity of image was proposed based on the two dimensional discrete cosine transform (2D DCT). With an energy threshold given beforehand, the DCT coefficients were processed with both energy normalization and sorting in descending order, and the sparsity of the image can be achieved by the proportion of dominant coefficients. And finally, the simulation result shows that, the method can estimate the sparsity of image effectively, and provides an active basis for the selection of compressive observation times. The result also shows that, since the selection of observation times is based on the sparse degree estimated with the energy threshold provided, the proposed method can ensure the quality of image reconstruction.
Most existing works on compressive imaging require electronic devices to perform the spatial optical modulation (SOM) or sparsifying transform (ST), which increase the cost and power consumption in field application. For the implementation of compressive imaging with a cheaper cost, the sensing strategy with blocking random pulse sampling (BRPS) is proposed. Instead of using the SOM and ST required by conventional technologies, the BRPS sampling can be achieved by the random unit-pulse in spatial domains. For actual application, the BRPS can be realized by address controlling for CMOS image sensors with a low resolution. For BRPS sampling, the image can be reconstructed by TVAL3. Experimental results show that, the BRPS achieves better reconstruction than conventional compressive sensing with Gaussian random matrix. Therefore, the BRPS contributes to the implementation of compressive sensing with low cost, low power consumption, less memory requirement, and better reconstruction.
Plasma produced by the radiation of a 1064 nm Nd:YAG laser focused onto a standard aluminum alloy E311 was studied spectroscopically. The electron density was inferred by measuring the Stark broadened line profile of Cu I 324.75 nm at a distance of 1.5 mm from the target surface with the laser irradiance of 3.27 GW/cm2. The electron temperature was determined using the Boltzmann plot method with eight neutral iron lines. At the same time, the validity of the assumption of local thermodynamic equilibrium was discussed in light of the results obtained.
A new strategy for images fusion is developed on the basis of block compressed sensing (BCS) and multiwavelet transform (MWT). Since the BCS with structured random matrix requires small memory space and enables fast computation, firstly, the images with large amounts of data can be compressively sampled into block images for fusion. Secondly, taking full advantages of multiwavelet such as symmetry, orthogonality, short support, and a higher number of vanishing moments, the compressive sampling of block images can be better described by MWT transform. Then the compressive measurements are fused with a linear weighting strategy based on MWT decomposition. And finally, the fused compressive samplings are reconstructed by the smoothed projection Landweber algorithm, with consideration of blocking artifacts. Experiment result shows that the validity of proposed method. Simultaneously, field test indicates that the compressive fusion can give similar resolution with traditional MWT fusion.
A compressive fusion of remote sensing images is presented based on the block compressed sensing (BCS) and non-subsampled contourlet transform (NSCT). Since the BCS requires small memory space and enables fast computation, firstly, the images with large amounts of data can be compressively sampled into block images with structured random matrix. Further, the compressive measurements are decomposed with NSCT and their coefficients are fused by a rule of linear weighting. And finally, the fused image is reconstructed by the gradient projection sparse reconstruction algorithm, together with consideration of blocking artifacts. The field test of remote sensing images fusion shows the validity of the proposed method.
In this paper, we present a new and simple method to produce larger height (with millimeter-sized) helical surface
compared with the other methods. During the process, a convenient method for the fabrication of the helical substrate
made of B270 glass by using a smart oven is presented. A Global 7107 Coordinate Measuring Machines (CMM) is used
for the detection of the glass helicoid. The experimental results proved to be in agreement with the theoretical prediction
within the uncertainty of the error and can satisfied our requirement. Moreover, this method seems easy and simple to
produce larger height helical surface compared with other ways mentioned in the literature.
A novel strategy for remote sensing images fusion is presented based on the block compressed sensing (BCS). Firstly, the multiwavelet transform (MWT) are employed for better sparse representation of remote sensing images. The sparse representations of block images are then compressive sampling by the BCS with an identical scrambled block hadamard operator. Further, the measurements are fused by a linear weighting rule in the compressive domain. And finally, the fused image is reconstructed by the gradient projection sparse reconstruction (GPSR) algorithm. Experiments result analyzes the selection of block dimension and sampling rating, as well as the convergence performance of the proposed method. The field test of remote sensing images fusion shows the validity of the proposed method.
A novel strategy for images fusion is presented based on the block compressed sensing (BCS) and multiwavelet
transform (MWT). Since the BCS requires small memory requirement and enables fast computation, the images with
large amounts of data can be compressively sampled by the BCS. Secondly, taking full advantages of multiwavelet such
as symmetry, orthogonality, short support, and a higher number of vanishing moments, the compressive measurements
of images can be better represented by the MWT. Moreover, the compressive measurements are fused based on the
coherence of MWT decomposition coefficients. And finally, the fused image is reconstructed by the minimization of
total variance method, and an overlapped blocking technique is proposed to eliminate the block effects. Experiments
result shows the validity of the proposed method. Simultaneously, results also indicate that the compressive fusion can
produce better results than conventional fusion techniques such as the principle component analysis method, Laplacian
pyramid-based method, and wavelet transform method.
Access to the requested content is limited to institutions that have purchased or subscribe to SPIE eBooks.
You are receiving this notice because your organization may not have SPIE eBooks access.*
*Shibboleth/Open Athens users─please
sign in
to access your institution's subscriptions.
To obtain this item, you may purchase the complete book in print or electronic format on
SPIE.org.
INSTITUTIONAL Select your institution to access the SPIE Digital Library.
PERSONAL Sign in with your SPIE account to access your personal subscriptions or to use specific features such as save to my library, sign up for alerts, save searches, etc.