UCL Discovery
UCL home » Library Services » Electronic resources » UCL Discovery

The development and evaluation of computer vision algorithms for the control of an autonomous horticultural vehicle

Southall, J. B.; (2000) The development and evaluation of computer vision algorithms for the control of an autonomous horticultural vehicle. Doctoral thesis (Ph.D), UCL (University College London). Green open access

[thumbnail of The development and evaluation of computer vision algorithms for the control of an autonomous horticultural vehicle.pdf]
Preview
Text
The development and evaluation of computer vision algorithms for the control of an autonomous horticultural vehicle.pdf

Download (10MB) | Preview

Abstract

Economic and environmental pressures have led to a demand for reduced chemical use in crop production. In response to this, precision agriculture techniques have been developed that aim to increase the efficiency of farming operations by more targeted application of chemical treatment. The concept of plant scale husbandry (PSH) has emerged as the logical extreme of precision techniques, where crop and weed plants are treated on an individual basis. To investigate the feasibility of PSH, an autonomous horticultural vehicle has been developed at the Silsoe Research Institute. This thesis describes the development of computer vision algorithms for the experimental vehicle which aim to aid navigation in the field and also allow differential treatment of crop and weed. The algorithm, based upon an extended Kalman filter, exploits the semi-structured nature of the field environment in which the vehicle operates, namely the grid pattern formed by the crop planting. By tracking this grid pattern in the images captured by the vehicles camera as it traverses the field, it is possible to extract information to aid vehicle navigation, such as bearing and offset from the grid of plants. The grid structure can also act as a cue for crop/weed discrimination on the basis of plant position on the ground plane. In addition to tracking the grid pattern, the Kalman filter also estimates the mean distances between the rows of lines and plants in the grid, to cater for variations in the planting procedure. Experiments are described which test the localisation accuracy of the algorithms in offline trials with data captured from the vehicle's camera, and on-line in both a simplified testbed environment and the field. It is found that the algorithms allow safe navigation along the rows of crop. Further experiments demonstrate the crop/weed discrimination performance of the algorithm, both off-line and on-line in a crop treatment experiment performed in the field where all of the crop plants are correctly targeted and no weeds are mistakenly treated.

Type: Thesis (Doctoral)
Qualification: Ph.D
Title: The development and evaluation of computer vision algorithms for the control of an autonomous horticultural vehicle
Open access status: An open access version is available from UCL Discovery
Language: English
Additional information: Thesis digitised by ProQuest
Keywords: Applied sciences; Autonomous vehicles
URI: https://discovery.ucl.ac.uk/id/eprint/10100995
Downloads since deposit
86Downloads
Download activity - last month
Download activity - last 12 months
Downloads by country - last 12 months

Archive Staff Only

View Item View Item