Utility of hyperspectral image analysis for early detection of leaf yellowing of Phalaenopsis
| Autor: | Chung-Yu Wu, 吳宗祐 |
|---|---|
| Rok vydání: | 2017 |
| Druh dokumentu: | 學位論文 ; thesis |
| Popis: | 105 Taiwan is a source of Phalaenopsis crops in the world. Because of the ideal weather conditions, Taiwan has the most abundant variety of Phalaenopsis. That is the reason why “Phalaenopsis kingdom” is one of the nicknames of Taiwan. In recent years, the symptoms, leaf yellowing, leaf falling and sheath root rotting, have often been discovered in transportation and cultivation of Phalaenopsis. This phenomenon is called the “black head” by farmers in Taiwan. In serious case, infected plants will dead, which reduces the worth of Phalaenopsis seriously. The export of Phalaenopsis almost utilizes ocean transportation in Taiwan. The cost of ocean transportation time is longer than the other transportation. In addition, the leaf yellowing disease of Phalaenopsis has longer incubation period, which cause the difficult detection by human’s eyes before pathogenesis. However, the growth time of the pathogens is provided by long-time transportation, and symptoms are discovered after arrival. It reduces the export worth of Phalaenopsis seriously. In order to solve this problem, this thesis proposes a method to detect the Phalaenopsis with leaf yellowing disease at early time based on infrared hyperspectral image analysis. The spectral bands of traditional digital image only cover a few of bands of visible spectrum normally. Therefore, the targets detection of traditional digital image processing always bases on spatial resolution. If the targets image has low spatial resolution, the targets detection based on traditional digital image processing is difficult to achieve. However, hyperspectral image is the high dimension image, which combines the spectrum and spatial resolution. The bands of spectrum contains the ultraviolet bands, visible bands and infrared bands. Based on this property, the targets which cannot identified by human’s eyes can detected by spectral signature. That is the reason why hyperspectral image is very suitable for early detection of leaf yellowing of Phalaenopsis. This thesis proposed a hyperspectral image analysis based method which detected the Phalaenopsis with leaf yellowing disease before the symptoms appearance by infrared spectral range (wavelength at 900nm – 1700nm). In order to collect the spectral signatures of leaf yellowing disease, this method generated the infected Phalaenopsis plants by artificial inoculation method, and acquired the hyperspectral image data of infected plant samples by infrared hyperspectral imager. Then, the hyperspectral image analysis based preprocessing method is proposed which utilized to segment the leaf area. This method contains orthogonal subspace projection approach, maximum connected component approach and morphological image processing. Further, the band selection and normalization method were utilized to inhibit the noise and find the band which were influenced by leaf yellowing disease. Finally, the infected target detectors were designed by Neyman-Pearson detector that the infected targets could be detected at early time, and the infected abundance estimator was also designed which was utilized to support the results of infected targets detection and tried to improve the performance of detectors. |
| Databáze: | Networked Digital Library of Theses & Dissertations |
| Externí odkaz: |
|