Automated Vision-Based Beverage Bottle Quality And Level Inspection System

• Main Author: Abdul Rahman, Nor Nabilah Syazana
• Format: Thesis
• Language: English; English
• Published: 2018
• Subjects: T Technology (General); TA Engineering (General). Civil engineering (General)
• Online Access: http://eprints.utem.edu.my/id/eprint/23327/; http://plh.utem.edu.my/cgi-bin/koha/opac-detail.pl?biblionumber=112656

Automated vision inspection emerged as an important part of the product quality monitoring process.It is a requirement of International Organization for Standardization (ISO) 9001 to appease the customer satisfaction in terms of frequent improvement of the quality of products.It is totally impractical to rely on human inspector to handle a large scale quality control production because human has major drawback in their performance such as inconsistency and time consuming. Therefore,an automatic inspection is a promising approach to maintain product quality as well as to resolve the existing problems relate to delay outputs and cost burden. This research presents a computerized analysis to detect defects occur in beverages production in order to minimize the defective products.Image processing techniques are proposed to detect defects of beverages bottle.The defects are categorized into three classes which are bottle shape defect, color concentration defect and liquid level defect.For shape defect detection,three techniques are proposed namely local standard deviation (LSD),morphological operation and adaptive thresholding. Statistical histogram,gray level co-occurrence matrix (GLCM) and quadratic distance are applied for color concentration defect detection. The liquid level is detected using Hough transform and coordinate of point techniques. The classification process is analyzed using rule-based and decision tree classifiers.In developing automated beverage bottle quality and level inspection system, the performance is verified in terms of accuracy.The simulation result demonstrate LSD,statistical histogram and Hough transform are selected as the best technique by achieving 98% of shape,93% of color concentration and 91% of liquid level. For the system result,93% average accuracy has achieved for three defect detections. The system is ready for internet of things (IoT) platform which is using raspberry pi that gives benefit to user for wirelessly access and monitor the results.For the results validation,field testing is conducted,and the proposed system shows the capability to classify the bottle defect accurately.Thus,it has proven the proposed system is appropriate to be implemented in real-time application for beverage bottle quality inspection.