Noise Robustness of a Texture Classification Protocol for Natural Leaf Roughness Characterisation

Thomas Decourselle, Jean-Claude Simon, Ludovic Journaux, Frédéric Cointault, and Johel Miteran


texture classification, dimensionality reduction, information losses, robustness


In the context of leaf roughness study for precision spraying applications, this article deals with its characterisation by computer vision techniques. Texture analysis is a primordial step for applications based on image analysis such as medical or agronomical imaging. The aim is to classify textures after extraction of discriminating features. However, this problem remains complex in the case of natural leaves because of changes in lighting, scaling or orientation. There we consider a family of invariants from the frequency domain called Generalized Fourier Descriptors whose dimensionality is proportional to the spatial resolution of the images. These features used with a Support Vector Machines classifier lead to good results in terms of classification error rate when the dimensionality is small but it gives more errors when the dimensionality increases; we use there different kinds of dimensionality reduction techniques (linear or non-linear) whose aim is to keep most information in a vector of small dimensionality. It implies losses of information even if small. This is not the only source of losses, another one is related to the noise present in the images due to acquisition conditions and sensor sensitivity. Based on well-known methods methods but really adapted to natural context, we propose here to demonstrate the robustness of our method of classification despite these losses of information.

Important Links:

Go Back