Maximum Components Integration for Image Processing: An Application of Ultrasound for Detection of Small Objects in Containers

Abstract

Detection of small objects, those that lie on the bottom or stick to the wall of containers, constitutes a challenging issue for ultrasonic detection techniques. This is because echo signals from the object are fused with that of the inner surface of the container when subjected to ultrasound scanning. This study proposes a maximum component integration method based on the short‐time Fourier transform algorithm to detect these objects. Experiments were conducted using glass fragments of about 2 × 2 × 2 mm3 to test the proposed method. Compared with other signal‐processing methods (statistical calculations, backscattered amplitude integral and maximum frequency calculation), this method is able to make selective and full use of multiecho information, and hence demonstrated to have improved detection ability to the extent that it can detect small glass fragments contained inside glass containers. Principles are introduced for choosing the optimized WINDOW size and signal size to be processed when applying this method.

Detection of occluded small objects in glass bottles filled with beverages via ultrasound center frequency tracing

“…In the course of this project, the authors tried to use Hilbert transform method that was reported to find out the small channels in food plastic package sealing (Raum, Ozguler, Morris, & O'Brien, 1998). The detectability of Hilbert transform was not so effective as the proposed method in our earlier study (Zhao, Basir, & Mittal, 2007). We think this contrast result could be explained as follows.…”

Contact and non-contact ultrasonic measurement in the food industry: a review

“…A successful classification rate of 95% was achieved by the proposed method where seven neurons were generated for analysis. Furthermore, a research conducted by this team has successfully introduced a maximum component integration (MCI) method based on the short-time Fourier transform algorithm in order to detect small glass fragments having a size of around 8 mm 3 [88]. This research compared the MCI method with different signal processing methods such as maximum frequency pressure ratio, a backscattered amplitude integral (BAI), standard deviation, variance, skewness and kurtosis.…”