The study of change using conditional morphological operators

This paper presents a case study on the dynamics of Ambrosia artemisiifolia populations in the surroundings of Trois-Rivières (Canada) that will enable us to demonstrate the validity of our arguments on a new procedure for change detection. Ambrosia artemisiifolia (common ragweed) is a very common and abundant annual plant in the open spaces of Northeastern America. The pollens emited by this plant are the principal cause of hay fever Eastern Canada. Ambrosia artemisiifolia is also causing, as a bad plant, decreasing yields in many agricultural industries through invasion of prairies and cultivated land. It is interesting to study the evolution of Ambrosia artemisiifolia populations under the urban dynamics angle since these populations are almost always linked to changes in vegetal covers and land use caused by human activities. INTRODUCTION The direct remote sensing of Ambrosia artemisiifolia is very difficult even when using hyperspectral data. In a previous study [1] [2] using an experimental agricultural prairie covering 7 hectares (18 acres) located in northwestern Montreal (Canada), 153 evenly spaced stations were surveyed. For each station covering 0.16 square meters a photograph and a measure of reflectance between 400 and 1100 nanometers were taken. A reading of present species was collected and in laboratory, slides were showed to 3 experts for defenitive identification of species and estimation of relative presence on each station. Three tables of contingency were then structured in order to cross the tags of each station and the proportion of each species recorded, the proportion of Ambrosia artemisiifolia for each station and the relative reflectance between 400 and 1100 nanometers, finally to cross stations completely covered by Ambrosia artemisiifolia or plants known to be difficult to distinguish from Ambrosia artemisiifolia and their relative reflectance for selected intervals. Our data analysis results showed that it was very difficult to seperate the spectra of Ambrosia artemisiifolia from companion plants like artemisia and asclepias. Furthermore, the situation was worse for mixed covers that are common to open spaces. The conclusions of this study [1] [2] were that the direct remote sensing of Ambrosia artemisiifolia populations through regular means of caption could be attempted at a coarser scale and using knowlegde on the ecolgy of the plant. Together with the very limited knowledge practionners have on the ecology of the plant, a thing is sure : the populations of Ambrosia artemisiifolia are closely linked to changes occuring on the urban fabric. Unfortunately, the study of change in remote sensing images bare problems [3] that must be overcome in order to obtain satisfying results. Among these problems the present paper will treat of comparison noise suppression, change detection and spatial queries through the use of conditional morphological operators.