Design of an integrated machine vision system cabable of detecting hidden infestation in wheat grains

Project Report No. 262

Design of an integrated machine vision system capable of detecting hidden infestation in wheat grains


J Chambers1, C Ridgway1 and E R Davies2

1Central Science Laboratory, Sand Hutton, York YO41 1LZ
2Machine Vision Group, Department of Physics, Royal Holloway, University of London, Egham, Surrey TW20 0EX



Major advances have been made towards the World’s first system for the rapid, automated and non-destructive determination of internal infestation within kernels of post-harvest cereals. The system will be based on the analysis of images of the kernels recorded in the near infrared (NIR) region by machine vision using highly innovative computational algorithms. These algorithms are notably more effective at analysing the images than human inspection as well as being non-subjective and operating at constant efficiency. Laboratory tests suggest that the probability of finding a 0.5% infestation of grain weevils in a batch of cereal could be around 97%. The speed of processing the images has been markedly improved such that analysis of a 3 kg sample within 3 minutes should be possible. The discovery that the method works with images recorded in the very near infrared region, at 981 nm, means that it will be possible to use a cheaper camera which is less subject to drift than had originally been thought. The method works by detecting bright patches on infested kernels which are highly correlated with the present or past location of the insect and probably result from loss of starch due to insect feeding. Repeated scanning of infested kernels suggests that the NIR effect becomes detectable around 2-3 weeks after egg laying. This is only shortly after the infestation becomes detectable by the much more expensive X-ray method. The NIR method has been optimised using wheat of Mercia variety but preliminary tests with other varieties of different reflective appearance suggest that it has general applicability to wheat. The method could share the same sample presentation arrangement as the machine vision system being developed to detect contamination external to cereal kernels but the cameras and software analysis will need to remain separate. Now that these laboratory tests have proved the potential of the method, the next steps will be to confirm that it will work with cereals other than wheat and infestations other than the grain weevil, and to construct a prototype apparatus to test under practical conditions. The implications of this project are that, provided the advances described here are exploited, the UK cereal trade will be able to benefit, in particular for both long term storage and export, by being the first to have at its disposal a system which can confirm the high quality of UK grain and which will operate at an acceptable speed and provide efficient results at an affordable cost.

Related Publications

Document download

View a printer friendly version of this publication

Download this publication PDF