No abstract

Abstract-We present a system to recognize underwater plankton images from the shadow image particle profiling evaluation recorder (SIPPER). The challenge of the SIPPER image set is that many images do not have clear contours. To address that, shape features that do not heavily depend on contour information were developed. A soft margin support vector machine (SVM) was used as the classifier. We developed a way to assign probability after multiclass SVM classification. Our approach achieved approximately 90% accuracy on a collection of plankton images. On another larger image set containing manually unidentifiable particles, it also provided 75.6% overall accuracy. The proposed approach was statistically significantly more accurate on the two data sets than a C4.5 decision tree and a cascade correlation neural network. The single SVM significantly outperformed ensembles of decision trees created by bagging and random forests on the smaller data set and was slightly better on the other data set. The 15-feature subset produced by our feature selection approach provided slightly better accuracy than using all 29 features. Our probability model gave us a reasonable rejection curve on the larger data set.Index Terms-Feature selection, learning, plankton recognition, probabilistic output, support vector machine (SVM).

No abstract

When insect population density varies within the same cotton field, estimation of abundance is difficult. Multiple population densities of the same species occur because cotton fields (due to edaphic and environmental effects) are apportioned into various habitats that are colonized at different rates. These various habitats differ temporally in their spatial distributions, exhibiting varying patterns of interspersion, shape and size. Therefore, when sampling multiple population densities without considering the influence of habitat structure, the estimated population mean represents a summary of diverse population distributions having different means and variances. This single estimate of mean abundance can lead to pest management decisions that are incorrect because it may over-or under-estimate pest density in different areas of the field. Delineation of habitat classes is essential in order to make local control decisions. Within large commercial cotton fields, it is too laborious for observers on the ground to map habitat boundaries, but remote sensing can efficiently create geo-referenced, stratified maps of cotton field habitats. By employing these maps, a simple random sampling design and larger sample unit sizes, it is possible to estimate pest abundance in each habitat without large numbers of samples. Estimates of pest abundance by habitat, when supplemented with ecological precepts and consultant/producer experience, provide the basis for spatial approaches to pest control. Using small sample sizes, the integrated sampling methodology maps the spatial abundance of a cotton insect pest across several large cotton fields.