Roy Makumborenga

Masters student

Links:

Masters Project:

Enhancement of the Fynbos Leaf Optical 
Recognition Application (FLORA-E)

Fynbos is the term used to refer to the distinct vegetation of South Africa`s mountainous Cape region. It is part of the Cape Floral Kingdom, constituting approximately 80% of the total species that make up the kingdom, with over half of these species endemic to the Cape region [1,2]. The aim of the FLORA project is to build a platform capable of accurately identifying the various fynbos species that make up the Cape Floristic Region (CPR).

The fourth industrial revolution has seen the advent of new innovations in technological fields such as nanotechnology, image processing and artificial intelligence. The FLORA project is not only significant because of its alignment with global trends arising from the fourth industrial revolution, but it is also a powerful educational and conservationist tool. The Fynbos biome is very delicately balanced with the invasion of alien plant species and disruptive human activities constantly threatening the continued survival of many fynbos species across the Cape. The FLORA platform can be used to educate children and adults about the diversity of the fynbos biome, giving them knowledge that will help them have a greater appreciation of the environment. To the conservationist and botanist, FLORA is a powerful tool documenting the unique vegetation of the CPR. A detailed database of fynbos species continuously updated is invaluable to conservation efforts and initiatives in the region [1,2].

FLORA-E Research Objectives

The FLORA project is concerned with the building of an artificially intelligent platform that will be used for the identification of different species of fynbos across the fynbos biome. The key concepts that this version of FLORA will seek to address are:

  1. Expansion of FLORA`s recognition capabilities to encompass Ericaceae fynbos, characterised by leafy stems and bulb shaped flowers, as well as the grassy Restionaceae fynbos family, on top of the broad leafed Proteaceae fynbos varieties it`s already trained on.
  2. Point i) will invariably lead to a re-evaluation of the image processing and classification algorithms used by FLORA. To identify a wider variety of species, the applications current process flow will need redesign.
  3. Earlier attempts to accelerate FLORA via general purpose computing on a Graphics Processing Unit (GPU) were unsuccessful. FLORA-E will seek to find an alternative acceleration model that will allow utilisation of the computational power of a GPU [6].
  4. Integration of FLORA mobile applications. The first to be used as an end user front-end, allowing for on-the-go fynbos classification on a user’s mobile phone connected to the FLORA server core via TCP/IP for example. The second to be used as a data collection tool which will allow for easier uploading and expansion of the FLORA dataset.
  5. Design of the FLORA web front-end to give it a modern responsive look and feel catering to a broad audience ranging from children to botany experts.

References

  1. B.W. van Wilgen “The evolution of fire and invasive alien plant management practices in fynbos,”  South African Journal of Science  vol.105 n.9-10, Pretoria Sep/Oct. 2009 [Online]. Available:. [Accessed Feb 1, 2019]

  2. R.M. Cowlinga, R.L. Presseyb, M. Rougetc , A.T. Lombarda, “A conservation plan for a         global biodiversity hotspot— the Cape Floristic Region, South Africa,”  Biological Conservation n 112 (2003) 191–216. 2002 [Online]. Available: http://www.bio-nica.info/biblioteca/Cowling2003ConservationPlanCape.pdf . [Accessed Feb 1, 2019]

  3. K. Naidoo, “GPU Acceleration of the Fynbos Leaf-based Online Recognition Application,”  MSc dissertation, Department of Electrical and Computer Engineering. University of Cape Town, pp. 55-56. 2015. Available From: http://www.vula.uct.ac.za