Seishi Ninomiya

Combining regression trees and radial basis function networks.

We describe a method for non-parametric regression which combines regression trees with radial basis function networks. The method is similar to that of Kubat, who was first to suggest such a combination, but has some significant improvements. We demonstrate the features of the new method, compare its performance with other methods on DELVE data sets and apply it to a real world problem involving the classification of soybean plants from digital images.

Quantitative evaluation of common buckwheat (Fagopyrum esculentum Moench) kernel shape by elliptic Fourier descriptor

Kernel shape of 27 common buckwheat cultivars was evaluated quantitatively by image analysis using elliptic Fourier descriptors and their principal components. The relationships between the quantitative information on kernel shape and several agronomic characteristics were clarified. The closed contour of each kernel projection was extracted, and 80 elliptic Fourier coefficients were calculated for each contour. The Fourier coefficients were standardized so that they were invariant of the size, rotation, shift and chain-code starting point. Then, the principal components on the standardized Fourier coefficients were evaluated. The cumulative contribution at the fifth principal component was higher than 93%. It was found that the first, second, third and fifth principal components represented the aspect ratio of the kernel, the location of the kernel centroid, the sharpness of the two kernel tips and the roundness of the base in the kernel contour, respectively. Analysis of the variance indicated that genotypic differences in these principal components were significantly large. Because these variations of the principal components were continuous, the effect of polygenes on shape was suggested. The relationships between these principal components and agronomic traits, including flowering period, yield, 1000 seed weight and volume weight, were also investigated. It was clarified that the first principal component was closely correlated with agronomic traits such as flowerig period, 1000 seed weight and volume weight. In conclusion, it was clarified that principal component scores based on elliptic Fourier descriptors yield useful quantitative parameters for evaluating kernel shape for common buckwheat breeding.

Statistical Models for Prediction of Dry Weight and Nitrogen Accumulation Based on Visible and Near-Infrared Hyper-Spectral Reflectance of Rice Canopies

Summary Much information is obtainable from hyper-spectral data, which measure solar radiation consecutively at less than about 10-nm intervals. In constructing statistical prediction models, however, problems of overfitting may arise due to the excessive number of variables, and multicollinearity may occur between variables ; thus a few specific wavelengths must be chosen. Various multivariate regression models were examined with ten-fold cross-validation to develop efficient, accurate models to predict dry weight and nitrogen accumulation of rice crops from the maximum tiller number stage to the meiosis stage, using plant-canopy reflectance of hyper-spectra within the 400-1100 nm domain without any variable selection. The results showed that the principal component regression using hyperspectra gave better fits and predictability than that using specific wavelengths. On the other hand, partial least squares regression was the most useful among the models tested ; this method avoided overfitting andmulticollinearity by using all wavelength information without variable selection and by inclusion of both x and y variations in its latent variables.

Discrimination of soybean leaflet shape by neural networks with image input.

It is necessary to correctly evaluate intra- and inter-specific variations for the efficient collection and preservation of genetic resources, and leaf shape is one of the important characteristics to be evaluated. It has been thought that a more consistent and quantitative method should be introduced to aid in the processes of practical discrimination. Several researchers have suggested leaf shape evaluation methods using shape features, and these methods have shown good results. The shape features selected in these methods have differed from one method to another, and new shape features must be redefined when these methods are applied to new cases. The processes for defining and extracting shape features are ad hoc. We, therefore, have attempted to develop a generalized model that requires neither the definition nor extraction of any shape features; the method uses neural networks into which leaf shape images are input. In this study, we applied a Hopfield model and a simple perceptron to the varietal discrimination of individual leaflet shapes of 364 soybean leaflets of 38 varieties. In the examination of up to ten varieties, the discriminant error of the neural networks with image input was satisfactorily low even under cross validation. We, therefore, concluded that this model works quite well for quantitative varietal discrimination in the case of soybean leaflets. The advantage of requiring neither the definition nor extraction of any shape features makes us expect that this model will be widely applicable to other cases, and we will attempt to verify this applicability.

Quantitative evaluation of the petal shape variation in Primula sieboldii caused by breeding process in the last 300 years

Primula sieboldii (E. Morren) has been a popular garden plant at least since the Edo period, about 300 years ago. We compared petal form between cultivars and wild populations in order to characterise the changes that have occurred during domestication. The comparison was made using EF-PCA analysis, which describes overall petal shape mathematically by transforming petal contour coordinates into elliptic Fourier descriptors; it subsequently summarises these descriptors by principal component analysis (PCA). Rearing cultivars in a common-garden experiment identified the PCs with a substantial genetic element. A clear heritable component was detected for the PCs characterising symmetrical variation in flower shape, but not the asymmetrical variation. Wild populations of this species have become endangered owing to habitat destruction by human activity, and many lowland floodplain habitats have been lost. Variation within the remaining wild populations was significantly lower than in the cultivars for PC1 (aspect ratio), PC3 (curvature of proximal and distal parts) and petal area; but not for PC2 (depth of head notch) and PC4 (position of the centre of gravity). The shifts in petal form from the wild populations to the cultivars parallel those seen in other crop-types following domestication, including an increase in size and diversity of forms: cultivars have shallower head notches, more fan-shaped petals and larger petals than do wild P. sieboldii.

Quantitative evaluation of flower colour pattern by image analysis and principal component analysis of Primula sieboldii E. Morren

In Primula sieboldii E. Morren, many cultivars have been bred with the purpose of obtaining various petal colour patterns. Colour pattern is an important breeding target in this species, and the availability of an objective and quantitative evaluation method is of vital importance for both genetic analysis and variety registration. Our objective was to establish a new quantitative evaluation method of P. sieboldii flower colour patterns in regions of interest (ROIs) by principal component analysis (PCA). We first set a ROI in each petal as a region that represented the petal colour pattern and defined the maximum square on each petal as the ROI. We then converted each ROI image to a 10 × 10 pixel mosaic image and defined a total of 300 variables (the colour values for red, green and blue components of the 100 pixels) per mosaic image. Finally, we summarized the information on the 300 variables by PCA, and redrew the mosaic images to correspond to some typical principal component scores to determine the effect of each principal component on colour pattern. By this method, we detected five different features of petal colour pattern, four of which were revealed to be mainly genetically controlled. Thus, we successfully established a procedure for evaluating petal colour patterns in P. sieboldii cultivars. This new procedure can be used as a basis for an objective and efficient variety registration system.

Field server: multi-functional wireless sensor network node for earth observation

1. Field Server Field Server is a small Web server, which can be installed globally for long term by using wireless LAN and the Internet. Field Server can monitor environmental conditions and ecosystems in real-time by high resolution camera, multiple sensors and LED lighting. Ad-hoc Wi-Fi mesh-network is created by Field Servers, and also Wi-Fi hotspots can be created around Field Servers for ubiquitous networking. Field Servers can be used also as a platform to use network devices and electroequipments in open fields. In Field Server’s case, all devices are connected by Ethernet or Wi-Fi, that is, inside of Field Server is also sensor-network, which can realize layout-free architecture, plug & play, robustness and various designs for Field Servers.

Genetic Combining Ability of Petal Shape in Garden Pansy ( Viola × wittrockiana Gams) based on Image Analysis

Flower appearance is an important target for improvement in garden pansy (Viola × wittrockiana Gams). Flowers of this species consist of three petal types: inferior, lateral, and superior. By means of principal-components (PC) analysis of elliptic Fourier descriptors (EFDs), we estimated the general and specific combining abilities (GCA and SCA) of the floral characteristics of F1 progeny derived from diallel crosses of four inbred lines. The greatest variation in petal shape was explained by the aspect ratio (the first PC) in each petal type. The second or third PC of each petal type was associated with the curvatures of the various parts of the petal. The highly significant GCA effects indicate the importance of additive genetic variance in the transmission of parental petal characteristics to the progeny. The fact that the SCA mean squares were not significant for aspect ratio and petal area indicates that these characteristics of a single-cross progeny can be sufficiently predicted on the basis of GCA. Significant SCA effects were observed in the curvatures of the distal and proximal parts of lateral and superior petals. Correlation analyses indicated several associations between the shape elements of lateral and superior petals, suggesting that genes for these shape elements may be associated, linked, or pleiotropic in the parents used in this study. We successfully demonstrated the use of EFD–PCA to evaluate the petal shape of garden pansy, and of analyzing combining ability and correlations based on the PC scores of EFDs.

Functional Regression in Crop Lodging Assessment With Digital Images

A method is proposed to predict a lodging score for a rice field based only on a digital overhead image of the field. This method converts the two-dimensional image data to a one-dimensional function by computing the average variance of transects across the image as a function of the transect angle. Then principles of functional data analysis are applied to estimate a regression function, and the predicted lodging score is an intercept term plus a measure of overall variability plus the inner product of the regression function and the periodogram of the average variance function.

Field Server Projects

A field server is a sensor node which can create a wireless sensor network and simultaneously serve as Wi-Fi hotspots in open fields. Since 2001, we have been working on several Research and Development projects to develop field server technology and its applications. So far, three generations of field servers have been developed and are in use in more than a dozen countries for collaborative research and educational applications such as a global sensor network for Earth observation, food safety/IT-agriculture, advanced sensor technology in fields and image monitoring for urban areas. The field servers have proved to be quite robust for field applications. More work is in progress to develop applications for the field server and resolve issues with power consumption and increasing the communication range for certain applications