Data fusion process is based on the association, correlation, and combination of
data and information from single and multiple sources to achieve refined position,
identify estimates, and complete and timely assess situations, threats, and their
significance (White 1991). These techniques have been broadly employed on multisensory environments to extract different types or variants of the input data. Their
objective is to obtain lower detection error probability and higher reliability in
multisensory environments (Castandeo 2013).
This chapter includes two different kinds of data fusion examples and some basic
information about wavelets and deep learning. The first example, given in Sect. 7.3,
is a linear regression model application through the segmentation of input data
against the nonlinearity of the system. The second example is demonstrated in
Sect. 7.4 by using fully connected artificial neural networks for time domain
estimations. Wavelets and convolutional neural networks are increasingly used in
data fusion processes. Sections 7.5 and 7.6 contain basic information about wavelets
and convolutional neural networks.
7.2 Agricultural Information Systems
Real-time crop status monitoring, yield prediction, irrigation management, precision
farming, resource management, and policy management are some application fields
in agriculture where sensor and data fusion methods are utilized. Basin resource
management requires information about the existing crop patterns, yield efficiency,
and probable alternative crop patterns with respect to ecological appropriateness
under the sustainability restrictions.
Although crop yield prediction by using the remote sensing data together with
agro-meteorological observations has never been a strategic issue, accuracy limitation appears as the major restriction. An example of fusion model that improves crop
yield prediction performance by using the observation data from the “Agricultural
Monitoring and Information Systems Project” (TARBIL) monitoring network in
Turkey is demonstrated in Sect. 7.3.
Multitemporal satellite images are widely used in crop monitoring. Very-highresolution (VHR), multispectral, and hyperspectral images are important spatial data
for precision agriculture. Although those spatial data are acquired within some
programmed or periodic timing, quality parameters differ depending on cloudiness,
viewing angle, and atmospheric condition as well as surface snow coverage and
wetness. Phenological timing is an additional important consideration in agricultural
data acquisition since crop growth depends on phenological stages. A plant may
appear at different forms at the same chronological calendar time of the season in
different years due to the shift of the phenological stage depending on the varying
seasonal conditions. Observation systems usually provide data in three different
sampling types:
7 Data Fusion in Agricultural Information Systems
105
data and information from single and multiple sources to achieve refined position,
identify estimates, and complete and timely assess situations, threats, and their
significance (White 1991). These techniques have been broadly employed on multisensory environments to extract different types or variants of the input data. Their
objective is to obtain lower detection error probability and higher reliability in
multisensory environments (Castandeo 2013).
This chapter includes two different kinds of data fusion examples and some basic
information about wavelets and deep learning. The first example, given in Sect. 7.3,
is a linear regression model application through the segmentation of input data
against the nonlinearity of the system. The second example is demonstrated in
Sect. 7.4 by using fully connected artificial neural networks for time domain
estimations. Wavelets and convolutional neural networks are increasingly used in
data fusion processes. Sections 7.5 and 7.6 contain basic information about wavelets
and convolutional neural networks.
7.2 Agricultural Information Systems
Real-time crop status monitoring, yield prediction, irrigation management, precision
farming, resource management, and policy management are some application fields
in agriculture where sensor and data fusion methods are utilized. Basin resource
management requires information about the existing crop patterns, yield efficiency,
and probable alternative crop patterns with respect to ecological appropriateness
under the sustainability restrictions.
Although crop yield prediction by using the remote sensing data together with
agro-meteorological observations has never been a strategic issue, accuracy limitation appears as the major restriction. An example of fusion model that improves crop
yield prediction performance by using the observation data from the “Agricultural
Monitoring and Information Systems Project” (TARBIL) monitoring network in
Turkey is demonstrated in Sect. 7.3.
Multitemporal satellite images are widely used in crop monitoring. Very-highresolution (VHR), multispectral, and hyperspectral images are important spatial data
for precision agriculture. Although those spatial data are acquired within some
programmed or periodic timing, quality parameters differ depending on cloudiness,
viewing angle, and atmospheric condition as well as surface snow coverage and
wetness. Phenological timing is an additional important consideration in agricultural
data acquisition since crop growth depends on phenological stages. A plant may
appear at different forms at the same chronological calendar time of the season in
different years due to the shift of the phenological stage depending on the varying
seasonal conditions. Observation systems usually provide data in three different
sampling types:
7 Data Fusion in Agricultural Information Systems
105