Submitted to: Agricultural Engineering International Conference
Publication Type: Proceedings
Publication Acceptance Date: October 10, 2002
Publication Date: November 28, 2002
Citation: Lu, R. 2002. Near infrared imaging spectroscopy for detecting new and old bruises on apples. Agricultural Engineering International Conference Proceedings. Interpretive Summary: To be competitive in a global economy that the fruit industry must provide a better quality and more consistent fresh product for consumers. Fresh fruit must be free from surface blemishes and defects such as bruises. Bruises on apples can lower the quality grade of fruit, resulting in significant economic losses to the industry. Currently, machine vision is widely used in packinghouses to sort fruit for color and size or weight. However, sorting for defects, especially subtle ones such as bruises, still cannot be done effectively with machine vision. It is challenging to detect bruises because bruising occurs to the tissue beneath the fruit skin and the properties of bruised tissue change over time. This research explored the potential of using imaging spectroscopy for detecting both new and old bruises on apples in wavelengths longer than those of visible light. Imaging spectroscopy is a technique that combines the features of imaging and spectroscopy to allow us to acquire significantly more information from an object than either conventional imaging or spectroscopy alone. An imaging spectroscopy system was assembled and computer algorithms using complex mathematical methods were developed to acquire and process images from bruised apples. It was found that the optical properties of bruised tissue changed over time and this change was variety and fruit dependent. The wavelength range between 1,000 nm and 1,340 nm was appropriate for detecting bruises. The imaging spectroscopy system was able to detect bruises with the correct detection rates between 59 to 94%, depending on the time elapsed after bruising. The optimal number of spectral bands with different wavelengths for bruise detection was between 20 and 40. This research showed that the near-infrared imaging spectroscopy technique can be used to inspect fruit for defects such as bruises. The research provides the critical information needed to develop a useful detection system for quality inspection of fruit at packinghouses or export facilities.
Technical Abstract: The objective of this research was to investigate the potential of near-infrared (NIR) imaging spectroscopy for detecting new and old bruises on apple fruit in the spectral region between 900 nm and 1700 nm. Experiments were conducted to acquire hyperspectral image cubes from Red Delicious and Golden Delicious apples over a period of 47 days. Bruises changed with time from lower reflectance to higher reflectance and the rate of the change varied with fruit and variety. The spectral region between 1000 nm and 1340 nm was the most appropriate for detecting bruises. The correct bruise detection rate was affected by the time elapsed after bruising, ranging from 62% to 88% for Red Delicious and from 59% to 94% for Golden Delicious. The optimal number of spectral bands for bruise detection was between 20 and 40 bands with the corresponding spectral resolution between 17.3 and 8.6 nm. The NIR imaging spectroscopy technique is useful for detecting subtle defects such as bruises on apple fruit.