Impurities Detection in Intensity Inhomogeneous Edible Bird’s Nest (EBN) Using a U-Net Deep Learning Model

Abstract

As an important export, cleanliness control on edible bird’s nest (EBN) is paramount. Automatic impurities detection is in urgent need to replace manual practices. However, effective impurities detection algorithm is yet to be developed due to the unresolved inhomogeneous optical properties of EBN. The objective of this work is to develop a novel U-net based algorithm for accurate impurities detection. The algorithm leveraged the convolution mechanisms of U-net for precise and localized features extraction. Output probability tensors were then generated from the deconvolution layers for impurities detection and positioning. The U-net based algorithm outperformed previous image processing-based methods with a higher impurities detection rate of 96.69% and a lower misclassification rate of 10.08%. The applicability of the algorithm was further confirmed with a reasonably high dice coefficient of more than 0.8. In conclusion, the developed U-net based algorithm successfully mitigated intensity inhomogeneity in EBN and improved the impurities detection rate.

Development of a hybrid autoencoder model for automated edible bird’s nest impurities inspection

“…Therefore, the performance of the model, in general, would be slightly lower than its counterpart with supervised learning, recording a 96.69% detection rate. 7 However, it should be noted that the sizes of detectable impurities were not reported. Nonetheless, this outcome was acknowledged as the reconstruction of EBN nonimpurity regions was not perfect.…”

“…Recently, the impurities detection performance has also been improved significantly up to 96.69% with a reduction in the misclassification rate to 10.08% using a deep learning U-net based segmentation algorithm. 7 Although a deep learning model with high detection accuracy has been attempted, 7 the natural characteristics of randomness and variation, the complex and inhomogeneous optical properties, and various types of impurities have made the impurities detection process extremely challenging and unsolved, for a human inspector or any existing machine vision system. Therefore, there is yet any proclaimed automatic inspection system for EBNs.…”

“…To date, several attempts have been made toward the automation of impurities detection, but the technology is still in its infancy 4 – 7 The lack of effective methodologies and algorithms for impurities detection under a complex and often inhomogeneous optical property associated with the EBN remains a challenge for industrial players. A natural EBN is often ivory-white and translucent with yellow/gold, orange, red and pink hues.…”

“…Table 1 summarizes previous works 4 – 7 and the methods proposed chronologically since 2015, including their methodological highlights, performance measures, and critical remarks, especially on the sizes of detectable impurities. Notably, a paradigm shift in the detection algorithms has evolved from image processing-based techniques to the integration of deep learning algorithms.…”

“…Notably, a paradigm shift in the detection algorithms has evolved from image processing-based techniques to the integration of deep learning algorithms. Recently, the impurities detection performance has also been improved significantly up to 96.69% with a reduction in the misclassification rate to 10.08% using a deep learning U-net based segmentation algorithm 7 …”