Computer vision developments for the automatic inspection of fresh and processed fruits

Abstract

The quality of a fresh or processed fruit or vegetable is defined by a series of characteristics which make it more or less attractive to the consumer, such as ripeness, size, weight, shape, colour, presence of blemishes and diseases, presence or absence of fruit stems, seeds, etc. In summary, these characteristics may cover all of the factors that exert an influence on the product’s appearance, on its nutritional and organoleptic qualities or on its suitability for preservation. Most of these factors have traditionally been assessed by visual inspection performed by trained operators. However, the application of machine vision in agriculture has increased considerably in recent years since it provides substantial information about the nature and attributes of the produces, reduces costs, guarantees the maintenance of quality standards and provides useful information in real time. Moreover, machine vision opens the possibility of exploring agricultural products in invisible regions of the electromagnetic spectrum, as in the ultraviolet or infrared regions. Instituto Valenciano de Investigaciones Agrarias (IVIA) has developed during the past 15 years computer vision systems for the automatic, on-line inspection of fresh and processed fruits and vegetables. This paper shows the most important outcomes in this matter achieved by the department called Centro de Agroingeniería. One of such systems is a machine for the automatic inspection of pomegranate arils for fresh consumption. This machine individualizes, inspects, classifies and separates the arils in four categories, removing those that do not fulfil the minimal specifications. Multivariate analysis models are used to classify the arils with an average success about 90%. Another application is a machine to classify mandarin segments for canning. The system distinguishes among sound, broken or double segments, and is able to detect the presence of seeds in the segments. The system analyses the shape of the each individual segment to estimate morphological features that are used to classify it into different commercial categories. The machine classifies correctly more than 75% of the analyzed segments. Both systems are currently patent pending. In the field of computer vision systems for the inspection of fresh, whole fruit, most research has been focused on citrus fruits. While most commercial systems only detect the blemishes on the skin of fruit, a multispectral system has been developed to identify them. The system is capable of identifying the 11 most common defects of citrus skin using near infrared, colour and ultraviolet. It also uses induced ultraviolet fluorescence. The success rate achieved with such system reached 87% when identifying about 800 defects in five species of oranges and mandarins. The use of hyperspectral sensors makes it possible to conduct a more sophisticated analysis of the scene by acquiring sets of images corresponding to particular wavelengths. Using this technology, we have conducted different works aimed at detecting damages in citrus fruits, including fungal infestation. The acquired multi-dimensional spectral signature characterising a pixel has been used to analyse scenes and to detect different types of defects such as decay, more easily than using standard colour imaging systems.