Training Computers to Inspect Food Banner

Training Computers to Inspect Food

A new multi-purpose computer vision system to identify sub-standard food products has been created by a team of computer scientists at the University of Lincoln.

Professor Tom Duckett’s research has developed a multi-purpose, user-trainable software technology which has a range of possible applications and overcomes the specificity of existing visual inspection systems. The research is achieving impact in several areas within the food industry, including quality analysis of fresh produce, food processing and food packaging.

The technique was initially developed by Professor Duckett using off-the-shelf hardware to enable affordable detection, identification and quantification of common defects affecting potatoes.

Academics from the Lincoln Centre for Autonomous Systems worked with the UK’s Potato Council to produce a low-cost system which can assist quality control staff and improve consistency, speed and accuracy of defect identification and quantification.

The British potato industry is worth around £3.5 billion a year and potatoes account for 40 per cent of carbohydrates consumed in the UK.

The main factor affecting consumer preference is physical appearance, with clear unblemished skin a significant selling point. Most potatoes are sorted into different grades by hand, often resulting in mistakes and losses.

"The project has been taken up by Ishida Europe Ltd, so we hope to see our vision systems being applied in food processing plants around the whole world and helping to improve the quality and safety of food products for everyone."

To combat this, the new system was trained to recognise different blemishes and analyse potatoes in near-real time. It uses a low-cost vision sensor and a standard desktop computer with a graphics processing unit, together with software algorithms, to automatically learn the appearance of different defects in potatoes.

Relying on initial input by an expert, the technology can then learn to autonomously identify blemishes, diseases and good specimens in batches of potatoes.

The prototype has been trialled for analysis of potatoes and other crops in storage at Sutton Bridge Crop Storage Research – the leading post-harvest applied research facility for agricultural storage in the UK. Further developments include extending the basic technology to incorporate 3D sensing.

A second arm of Professor Duckett’s research employs laser scatter imaging and polarised light stress images to detect faults in the heat seals of food packaging. Heat-sealed packs perform a number of functions in addition to containing food, such as preventing damage and atmosphere modification to keep contents fresh.

A study of 105 packaging facilities concluded that, in the UK alone, a potential 480,000 tonnes of food waste per year is generated through unsound seals in food packaging.

The results of a research project funded by the Department for Environment, Food and Rural Affairs (Defra) demonstrated the potential for the development of a commercially viable, non-contact seal inspection method. This method could be implemented in the majority of tray packing processes, improving the detection of unsound packaging seals and thereby reducing waste.

Professor Duckett is now working with industry leaders, Ishida Europe Ltd, the Potato Council and Branston Plc, to apply these findings to create new multi-purpose imaging technology to automate quality inspection tasks in food processing and packaging. The project is funded by an £823,000 grant from the Technology Strategy Board.