Schrei- 
3-186. 
r, 1997. 
Eastern 
“lorence, 
"Multi- 
je Swiss 
| III/IV 
lling of 
5-18. 
snow co- 
for ope- 
Disserta- 
itute of 
Computer-assisted recognition of Olive trees in digital imagery 
Simon Kay, Olivier Léo, Steve Peedell', 
Space Applications Institute, JRC of the European Commission, Ispra, Italy 
Giovanna Giardino^, RSDE SrL, Milan, Italy 
Commission VII, Working group 4 
KEY WORDS: olive tree, photointerpretation, Europe, computer assisted, algorithm 
ABSTRACT 
In response to a direct requirement of the European Commission (EC) and Member States (MS), who 
need to manage some 500 million olive trees in Mediterranean countries, the Agricultural Information 
Systems Unit (of the Space Applications Institute, JRC) has developed an algorithm and software 
application for the computer-assisted recognition and counting of trees, designed to' be used on the 
main type of image data available (scanned panchromatic aerial photography, 8-bit, 1m pixel). 
The algorithm operates on four main parameters: 
- grey value threshold (minimum, maximum) 
- tree diameter (maximum, minimum) 
- crown density (maximum, minimum) 
— aspect ratio (range) 
The approach used significantly increases identification performance over existing methods, due to an 
iterative processing of the image, which thereby helps resolve problems associated with varying 
background brightness and tree crown densities. Early results have yielded very promising results and 
the algorithm (coded in C++) was demonstrated to the Commission and Member States in a custom 
built GIS (ArcView™) application. 
In September 1997, the EC (DGVI, Agriculture) launched the “OLISTAT” project, the goal being to 
estimate the number of olive trees in France, Italy, Spain, Portugal and Greece: The JRC algorithm will 
be used by the JRC as a checking tool for results obtained in each MS. Additionally, two MS 
administrations have tested the algorithm in the context of the Olive Tree Register. Together, these 
three activities - which include a significant level of field checking - represent an unprecedented 
dataset, covering a wide range of conditions, for the scientific validation of the algorithm. 
This paper, and the associated conference presentation, covers: (i) an overview of the algorithm; (ii) 
examples of how it functions; (iii) a description of its application in the quality checking of the OLISTAT 
project; (iv) an assessment of its effectiveness; (v) and, in conclusion, proposals for modifications. 
  
even 1000's of trees. Olive trees can be 
cultivated in pure stands or associated with 
arable crops, and sometimes even with other 
1. Introduction 
1.1 Why count trees? 
Long-standing ^ regulations (EEC 1975) 
concerning the management of the olive oil 
market sector, concentrated in five member 
states? of the European Union (EU), currently 
oblige an onerous survey of producers' trees in 
order to justify payments - a task that in Italy 
alone has involved the count of more than 160 
million trees. The main purpose of the survey 
was to establish a register, thus providing a 
basis for the payment and checking of aid to 
farmers (Kay et al., 1997). 
Olive tree stands in the Mediterranean can be 
varied in size, ranging from 10's to 100's and 
Intemational Archives of Photogrammetry and Remote Sensing. Vol. XXXII, Part 7, Budapest, 1998 
tree species (almond, apricot, oaks). 
Aerial photography is the only available data 
suitable for this kind of survey, and computer- 
assisted photo-interpretation and counting 
techniques become essential if the work is to 
be completed in a reasonable time-frame. 
1.2 The Olicount project 
In order to assist the planned register creation 
in Portugal and Greece in the near future, a 
demand was identified by the Commission to 
develop a tree counting tool, to assist Member 
State administrations in the task of 
implementing or maintaining their olive tree 
357.