A new feature extraction model based on image gradient

Zhao, Minghua; Yuan, Yongqin; Li, Bing; Shi, Zhiping

doi:10.1109/cisp.2013.6744025

Cited by 3 publications

(2 citation statements)

References 12 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…16 The similarity between image pairs can be measured through a correlation method based on the gradient orientation difference at each pixel. 17 In seismic applications, the spatial gradient of the seismic wavefield is analyzed to investigate important wave characteristics that can be directly related to wave propagation in layered earth. 18 Since the discontinuity of material properties at the damaged area is known to induce changes in the spatial ultrasonic waveform, it may then be reasonably surmised that such damage may be reflected by the change of spatial gradient field as time progresses.…”

Section: Introductionmentioning

confidence: 99%

Laser Ultrasonic Imaging of Wavefield Spatial Gradients for Damage Detection

Chong¹,

Wu²,

Todd³

2021

Lecture Notes in Mechanical Engineering

View full text Add to dashboard Cite

This paper describes a new damage visualization method to investigate and analyze propagating guided Lamb waves using analyses of wavefield spatial gradients. A laser ultrasonic interrogation system (LUIS) was used to create fullfield ultrasonic data measurements for ultrasonic wavefield imaging (UWI). The laser scanning process was performed based on both a raster scan and a circle scan. From the high-resolution wavefield data, a spatial gradient-based image processing technique was developed using gradient vectors to extract features sensitive to defects. Local impedance changes at the damaged area would result in a local distortion of the waveform which was captured and quantified by the variation of the gradient vectors in the scanning area as time evolves. Such variation was accumulated over time with a statistical threshold filter to generate a gradient-orientation map for damage visualization. The proposed algorithm was capable of producing distinctive damage patterns when tested experimentally on a 3-mm aluminum plate with multiple simultaneous simulated defects. Compared to conventional techniques like local wavenumber estimation, the generation of the accumulated orientation map involves no filtering process in the frequency or wavenumber domain, at the expense of more accurate shaping of the defect. A spatial covariance analysis was adopted to locate damage from the results as well as to evaluate the correlation between different kinds of defects. Combining the proposed approach with conventional laser ultrasonic imaging techniques enables a fast and robust damage identification and characterization process which requires lower computational burden and practical operation.

show abstract

Section: Introductionmentioning

confidence: 99%

Laser Ultrasonic Imaging of Wavefield Spatial Gradients for Damage Detection

Chong¹,

Wu²,

Todd³

2021

Lecture Notes in Mechanical Engineering

View full text Add to dashboard Cite

show abstract

“…16 The similarity between image pairs can be measured through a correlation method based on the gradient-orientation difference at each pixel. 17 In seismic applications, the spatial gradient of the seismic wavefield is analyzed to investigate important wave characteristics that can be directly related to wave propagation in layered earth. 18 Since the discontinuity of material properties at the damaged area is known to induce changes in the spatial ultrasonic waveform, it may then be reasonably surmised that such damage may be reflected by the change in spatial gradient field as time progresses.…”

Section: Introductionmentioning

confidence: 99%

Laser ultrasonic imaging of wavefield spatial gradients for damage detection

Chong

Todd

2020

Structural Health Monitoring

View full text Add to dashboard Cite

This article describes a new damage visualization method to investigate and analyze propagating guided Lamb waves using analyses of wavefield spatial gradients. A laser ultrasonic interrogation system was used to create full-field ultrasonic data measurements for ultrasonic wavefield imaging. The laser scanning process was performed based on both a raster scan and a circle scan. From the high-resolution wavefield data, a spatial gradient–based image processing technique was developed using gradient vectors to extract features sensitive to defects. Local impedance changes at the damaged area would result in a local distortion of the waveform which was captured and quantified by the variation of the gradient vectors in the scanning area as time evolves. Such variation was accumulated over time with a statistical threshold filter to generate a gradient-orientation map for damage visualization. The proposed algorithm was capable of producing distinctive damage patterns when tested experimentally on a 3-mm aluminum plate with multiple simultaneous simulated defects. Compared to conventional techniques like local wavenumber estimation, the generation of the accumulated orientation map involves no filtering process in the frequency or wavenumber domain, at the expense of more accurate shaping of the defect. A spatial covariance analysis was adopted to locate damage from the results as well as to evaluate the correlation between different kinds of defects. Combining the proposed approach with conventional laser ultrasonic imaging techniques enables a fast and robust damage identification and characterization process which requires lower computational burden and practical operation.

show abstract