Detection of Disease Symptoms on Hyperspectral 3d Plant Models

Abstract

Commission VII, WG VII/4 KEY WORDS: Hyperspectral 3D plant models, close range, anomaly detection, sparse representation, topographic dictionaries ABSTRACT:We analyze the benefit of combining hyperspectral images information with 3D geometry information for the detection of Cercospora leaf spot disease symptoms on sugar beet plants. Besides commonly used one-class Support Vector Machines, we utilize an unsupervised sparse representation-based approach with group sparsity prior. Geometry information is incorporated by representing each sample of interest with an inclination-sorted dictionary, which can be seen as an 1D topographic dictionary. We compare this approach with a sparse representation based approach without geometry information and One-Class Support Vector Machines. One-Class Support Vector Machines are applied to hyperspectral data without geometry information as well as to hyperspectral images with additional pixelwise inclination information. Our results show a gain in accuracy when using geometry information beside spectral information regardless of the used approach. However, both methods have different demands on the data when applied to new test data sets. One-Class Support Vector Machines require full inclination information on test and training data whereas the topographic dictionary approach only need spectral information for reconstruction of test data once the dictionary is build by spectra with inclination.

Approaches, applications, and future directions for hyperspectral vegetation studies: An emphasis on yield‐limiting factors in wheat

“…These effects must be overcome so that the reflectance is characteristic of the underlying biochemical properties rather than these illumination effects. Different approaches have tried to account for the illumination effects (Al Makdessi et al., 2019; Vigneau et al., 2011; Behmann et al., 2016; Roscher et al., 2016), although there is still no consensus on a reliable and proven method for removing such properties.…”

Early Detection of Plant Viral Disease Using Hyperspectral Imaging and Deep Learning

“…Intensive research on plant pathology scouting has leveraged hyperspectral imaging for early identification of pathogens and diseases at varying spatial, spectral, and temporal scales. Examples include using leaf reflectance to differentiate the signal change caused by the foliar pathogens in sugar beet [ 11 , 12 , 13 ], wheat [ 14 , 15 ], apple [ 16 ], barley [ 17 , 18 ], and tomato [ 19 ]. At the field scale, hyperspectral images are useful for early detection of toxigenic fungi on maize [ 20 ], yellow rust on wheat [ 21 ], orange rust on sugar cane [ 22 ], and tobacco disease [ 23 ].…”

Proximal Hyperspectral Imaging Detects Diurnal and Drought-Induced Changes in Maize Physiology

“…Once 3D information of each hyperspectral pixel is available, the angle and the distance between the pixel and the light source can be calculated, and their effect can be removed by a correction factor for each pixel determined from the light field of illumination intensities ( Behmann et al, 2016 ). Distance and inclination effects can also be removed by using the 3D information to estimate the parameters of a linear reflectance model ( Vigneau et al, 2011 ; Asaari et al, 2018 ), by incorporating distance and angle information in trait prediction models ( Roscher et al, 2016 ) or by developing a bidirectional reflectance distribution function ( Behmann et al, 2016 ). Combining 3D information with hyperspectral data is a promising approach to reduce the illumination effects, however, this information is not always available, and the sheer volume of this extra data further challenges storage and processing capacities.…”

“…This additional non-biological reflectance variation can mask biological effects and complicate hyperspectral data interpretation. Several methods, such as 3-dimensional (3D) modeling and standard normal variate normalization, have been proposed to reduce this variation ( Vigneau et al, 2011 ; Behmann et al, 2016 ; Roscher et al, 2016 ; Asaari et al, 2018 ). These approaches require, however, additional 3D information or perform transformations, which may complicate data interpretation by limiting the use of indices ( Asaari et al, 2018 ).…”