Image analysis for biological objects

Many processes in agriculture and horticulture require the visual assessment of the product. This is particularly true for grading fruit and vegetables, but also applies to automatic fruit harvesting, pruning, plant handling and many other applications where the workpieces are variable. Work at AFRC Engineering is investigating the possible applications of image processing and the techniques most likely to be useful. This paper uses two pieces of work as examples to express some of the basic considerations for using image processing in agriculture. Work on grading potatoes to select those suitable for baking needed an immediate solution, that was fast, robust and flexible, and at a reasonable price. Long-term work on neural networks is aimed at developing algorithms which can be trained to make decisions on ill-defined features, such as the shape of potatoes. These techniques are necessary for many of the grading decisions which currently have to be made manually