Bird nest shape quality assessment using machine vision system

Abstract

Swiftlets are birds contained within the four genera Aerodramus, Hydrochous, Schoutedenapus and Collocalia. They form the Collocaliini tribe within the swift family Apodidae. Swiftlet nest economy is currently envisaged to contribute significantly to foreign earnings of Malaysia. Many establishments are currently engaged in bird nest farming and trying to improve the quality and quantity of nest production. The raw bird’s nest (unprocessed) can achieve up to RM 4,000 per kilos. Processed and cleaned bird’s nest can reach up to RM 9,000 or more per kilo. To date, the bird nest grading is based on weight and shape. The inspection and grading for raw edible bird nest were performed visually by expert panels. This conventional method is relying more on human judgments. Unfortunately, it is a tedious process and often inconsistence from one person to another. Bird nest has an approximately two-dimensional nature, and, therefore they are most suitable for real-time machine processing. This experiment was performed on various camera angel and bird nest position. More than hundreds birds nest was used in this experiment obtained throughout west peninsular Malaysia. A Fourier-based shape separation (FD) method was developed from CCD image data to grade bird nest by its shape and size. FD was able to differentiate different shape such as round (oval) and 'v' shaped depending on the swiftlet species and geographical origin. Shape analysis was established using multivariate discriminant analysis. The Wilks' lambda analysis was invoked to transform and compress the data set comprising of large number of interconnected variables to a reduced set of variates. It can be further used to differentiate bird nest from different geographical origin. Overall, the vision system was able to correctly classify 100% of the V and Oval shaped and 81.3% for each grade in oval shape of the bird nest. The performances were compared with the expert panels and the results show that this technique achieved similar accuracy.