• Journal of the Korea Society of Computer and Information
• /
• v.21 no.11
• /
• pp.31-38
• /
• 2016

In this paper, we propose multi-scale infrared target detection algorithm with varied filter size using integral image. Filter based target detection is widely used for small target detection, but it doesn't suit for large target detection depending on the filter size. When there are multi-scale targets on the sky background, detection filter with small filter size can not detect the whole shape of the large targe. In contrast, detection filter with large filter size doesn't suit for small target detection, but also it requires a large amount of processing time. The proposed algorithm integrates the filtering results of varied filter size for the detection of small and large targets. The proposed algorithm has good performance for both small and large target detection. Furthermore, the proposed algorithm requires a less processing time, since it use the integral image to make the mean images with different filter sizes for subtraction between the original image and the respective mean image. In addition, we propose the implementation of real-time embedded system using FPGA.

• KSII Transactions on Internet and Information Systems (TIIS)
• /
• v.16 no.11
• /
• pp.3507-3522
• /
• 2022

Model pruning methods have attracted huge attention owing to the increasing demand of deploying models on low-resource devices recently. Most existing methods use the weight norm of filters to represent their importance, and discard the ones with small value directly to achieve the pruning target, which ignores the contribution of the small norm filters. This is not only results in filter contribution waste, but also gives comparable performance to training with the random initialized weights [1]. In this paper, we point out that the small norm filters can harm the performance of the pruned model greatly, if they are discarded directly. Therefore, we propose a novel filter contribution recycle (FCR) method for structured model pruning to resolve the fore-mentioned problem. FCR collects and reassembles contribution from the small norm filters to obtain a mixed contribution collector, and then assigns the reassembled contribution to other filters with higher probability to be preserved. To achieve the target FLOPs, FCR also adopts a weight decay strategy for the small norm filters. To explore the effectiveness of our approach, extensive experiments are conducted on ImageNet2012 and CIFAR-10 datasets, and superior results are reported when comparing with other methods under the same or even more FLOPs reduction. In addition, our method is flexible to be combined with other different pruning criterions.

• Journal of Korea Multimedia Society
• /
• v.16 no.6
• /
• pp.756-763
• /
• 2013

Bilateral filter (BF), functioning by two Gaussian filters, domain and range filter is a nonlinear filter for sharpness enhancement and noise removal. In infrared (IR) small target detection field, the BF is designed by background predictor for predicting background not including small target. For this, the standard deviations of the two Gaussian filters need to be changed adaptively in background and target region of an infrared image. In this paper, the proposed bilateral filter make the standard deviations changed adaptively, using variance feature of mean values of surrounding block neighboring local filter window. And, in case the variance of mean values for surrounding blocks is low for any processed pixel, the pixel is classified to flat background and target region for enhancing background prediction. On the other hand, any pixel with high variance for surrounding blocks is classified to edge region. Small target can be detected by subtracting predicted background from original image. In experimental results, we confirmed that the proposed bilateral filter has superior target detection rate, compared with existing methods.

• The Korean Journal of Applied Statistics
• /
• v.16 no.2
• /
• pp.239-246
• /
• 2003

In small area estimation, Best Linear Unbaised Predictor(BLUP) can be directly implicated ,specially, in use of the time series estimation. If there are correlations between observations and error terms over the time, Kalman Filter method can be used. Therefore, using kalman Filtering technique small area estimation of total of unemployments are estimated by BLUP. And for the example of this study, Economic Active Population Survey data were used.

• Journal of the Korea Institute of Military Science and Technology
• /
• v.14 no.4
• /
• pp.676-683
• /
• 2011

In this paper, we propose a robust method which can detect and track a high-speed small approaching target in a cluttered environment for Korean Active Protection System. The proposed method uses a temporal and spatial filter, tracking filter to detect and track a single target in consecutive order. And it is comprised of a candidate target detection step, a prior target selection step and a target tracking. Field tests on real infrared image sequences show that the proposed method could stably track a high speed and small target in complex background and target occlusion.

• The Transactions of The Korean Institute of Electrical Engineers
• /
• v.59 no.5
• /
• pp.1005-1009
• /
• 2010

This paper presents a digitally-controlled filter calibration technique for biquad RC filter using self oscillation. The biquad RC filter is converted to a fully-differential ring oscillator by changing its resistor connections, where the oscillation frequency reflects the cut-off frequency. The proposed calibration circuit measures the oscillation frequency by counting with a fixed higher-frequency clock and then tunes it to a desired frequency with a digital frequency-locked loop including a PI controller. Because the proposed circuit directly measures the cut-off frequency of the filter itself and calibrates it with the small area digital circuits, the area and the power consumption are much small compared with conventional works. When it is implemented in a 65nm CMOS process, the calibration circuit except the filter consumes the area of 80um X 50um and power consumption is 443uA at 1.2 V supply voltage.

• The Journal of Korean Institute of Communications and Information Sciences
• /
• v.34 no.9C
• /
• pp.863-870
• /
• 2009

Bilateral filter (BF) is a nonlinear filter for sharpness enhancement and noise removal. The BF performs the function by the two Gaussian filters, the domain filter and the range filter. To apply the BF to infrared (IR) small target detection, the standard deviation of the two Gaussian filters need to be changed adaptively between the background region and the target region. This paper presents a new BF with the adaptive standard deviation based on the analysis of the edge component of the local window, also having the variable filter size. This enables the BF to perform better and become more suitable in the field of small target detection Experimental results demonstrate that the proposed method is robust and efficient than the conventional methods.

• Journal of electromagnetic engineering and science
• /
• v.3 no.2
• /
• pp.104-110
• /
• 2003

A ring filter is proposed as a wide-banded filter. It consists of a ring and two short stubs, which are connected at 90$^{\circ}$ and 270$^{\circ}$ points of the ring. Since the termination impedance at 90$^{\circ}$ and 270$^{\circ}$ points of the ring and the characteristic impedance of the short stub have an effect on designing of it, the relation between them and bandwidths has been studied. Based on the study, two types of small-sized wideband CVT(constant VSWR-type impedance transformer)- and CCT(constant conductance-type impedance transformer)-ring filters are introduced, designed, simulated and one of two, a CCT -ring filter, is tested. The circumference of the ring can be reduced theoretically up to 60$^{\circ}$ and two of many cases having about 300$^{\circ}$ circumferences are simulated. The simulated results show more than 100 % fractional bandwidth, which can be obtained with more than 5 stages in conventional filter-design techniques. To test the designed CCT-ring filter, it has been fabricated in microstrip technology and the measured results show good agreement with the simulated ones, having more than 100 % fractional bandwidth.

• Journal of electromagnetic engineering and science
• /
• v.11 no.3
• /
• pp.186-191
• /
• 2011

This paper presents a simple and effective method for improving stopband rejection characteristics of previously studied dual-band filters. Small electric couplings were applied to the symmetrically positioned shunt resonators, which divided each transmission zero into two transmission zeros without any effect on passbands. We were able to achieve improved stopband rejection characteristics by these additional transmission zeros. For the filter application, we designed a dual-band filter with improved stopband characteristics using microstrip quasi-lumped elements. The electric couplings that control the location of transmission zeros are controlled by the distance between symmetric open stubs of the filter. The filter was fabricated with a relative dielectric constant of 3.5 and a thickness of 0.76 mm. The fabricated filter has a small size ($14.6{\times}13.2{\times}0.76$ mm) and a low insertion loss when compared with conventional filters.

• Transactions on Electrical and Electronic Materials
• /
• v.14 no.4
• /
• pp.187-192
• /
• 2013

Because of the importance of small target detection in infrared images, many studies have been carried out in this area. Using a Kalman filter and mean shift algorithm, this study proposes an algorithm to track multiple small moving targets even in cases of target disappearance and appearance in serial infrared images in an environment with many noises. Difference images, which highlight the background images estimated with a background estimation filter from the original images, have a relatively very bright value, which becomes a candidate target area. Multiple target tracking consists of a Kalman filter section (target position prediction) and candidate target classification section (target selection). The system removes error detection from the detection results of candidate targets in still images and associates targets in serial images. The final target detection locations were revised with the mean shift algorithm to have comparatively low tracking location errors and allow for continuous tracking with standard model updating. In the experiment with actual marine infrared serial images, the proposed system was compared with the Kalman filter method and mean shift algorithm. As a result, the proposed system recorded the lowest tracking location errors and ensured stable tracking with no tracking location diffusion.