| Popis: |
Our aim is to develop a completely automated and reliable system to identify morphological landmarks in digital images. The performance of the system is aimed to replicate manual digitization with equivalent accuracy and reliability, based upon a small number of training examples. The analysis system is constructed from four stages; a feature based detection of fly wing structure, correspondence matching based upon the pairwise geometric histogram (PGH) representation, global location of the wing using a Probabilistic Hough Transform (PHT), and finally local correlation bas ed refinement of individual features. We evaluate this system and compare quantitative results to manually digitized data. 1 Background Morphological landmarks are points that can be located precisely and establish an unambiguous one-to-one correspondence among all the specimens and are widely used in shape analysis [1]. Points like the tip of the nose or the oute r corner of the left eye are possible landmark points of the human face. Analyses of shape investigate the arrangement of landmark points relative to each other. A substantial body of statistical methods is available for the analysis of configurations of landmark p oints [2]. This framework of shape analysis by landmarks is increasingly used in many biological and medical applications and widely applied in many other fields. The configuration of landmarks have helped identify the possible source of re-infesting specimens and encounter the epidemiologically challenging vectors of Chagas disease [3]. The potential of using geometrical morphometric techniques as an invaluable tool for recognizing taxonomic data is being explored [4]. Other scientific applicat ions include investigating the study of size and shape to examine the effects of experimental treatments, genotype or other factors directly in the anatomical aspect. The use of l andmarks has been adapted to specific biological contexts such as genetics [5, 6, 7]; geog raphic differentiation [8], and the study of morphological integration [9, 10]. The process of identifying the landmarks is an important and labour-intensive part of any such analysis. Presently, this is usually done manually. Plugins for the ImageJ software (for digitizing the standard sets of landmarks) on fly wi ngs and mouse mandibles), increases speed and reliability over a completely unaided process [11]. However, there is still a requirement for an observer to manually identify each landmark point and therefore this process can be time-consuming, and quite often, the research questions are dependant on the duration of obtaining these data. |
