Đang chuẩn bị liên kết để tải về tài liệu:
Radar technolog Part 6

Tham khảo tài liệu 'radar technolog part 6', kỹ thuật - công nghệ, cơ khí - chế tạo máy phục vụ nhu cầu học tập, nghiên cứu và làm việc hiệu quả | Adaptive Ground Penetrating Radar Systems to Visualize Antipersonnel Plastic Landmines Based on Local Texture in Scattering Reflection Data in Space and Frequency Domains 135 Fig. 5. a Conventional and b proposed textural feature extraction methods based on correlation in real-space and frequency domains Nakano Hirose 2009 . We prepare a local observation window of L X L pixels to extract the local textural feature in the window by calculating correlations between pixel values. In 6 - 10 M is the average of pixel values in the window at a base frequency fb which we select among the 10- frequency observation points in advance. The vector Ks in 7 is the local correlations in the L X L real-space window at fb while Kf is the correlations between pixel values at fb and other frequencies fn at identical positions. The effectiveness of Kf fn in 10 as a feature vector was suggested by the following frequency-dependent interference. A plastic landmine usually has parallelism among its ceiling bottom and air gap inside if any which causes interference whose spectral profile is periodic in the frequency domain. That is to say we will observe a resonance at integral multiple of a certain frequency periodically in the frequency domain resulting in a specific peak at certain fn in Kf fn . We intended to capture this phenomenon in 10 . However we found in our series of experiments that we normally observe only a single peak in the 812GHz band. If we expect multiple frequency peaks we have to expand the observation bandwidth. However very high-frequency electromagnetic wave cannot penetrate ground so deep. Then we have changed our direction. Note that in the spatial texture case described above we paid attention to local correlation caused by the Markovian property. In the same way also in the frequency domain we decided to calculate the local correlation to observe the frequency space texture in a simple manner. Figure 5 b illustrates our proposal namely the frequency-domain .