Machine parts recognition and defect detection in automated assembly systems using computer vision techniques

Abstract

Assembly line automation gets more significance to cope up with increasing needs of latest technology machines which are used in industry and society. This paper presents a computationally efficient 2D computer vision based approach to recognize the machine parts and detect damaged parts on the assembly line. The image acquisition system which is part of the assembly line setup acquires data from the moving machine parts in line. Captured machine part image data undergoes image preprocessing techniques like background subtraction, binarization, scaling, and noise and holes removal to transform the data suitable for further processing. Then a contour of the machine parts are extracted and normalized by equal part area method to describe the shape. It gives important clues for machine part shape recognition and defect identification. For experimental purpose a model shape for each machine part is developed, the shape recognition and defect detection are performed with only reference to the model shape. The defects in the machine part such as damage, cracks are identified by the similarity measure between model shape and the data extracted from machine part of the assembly line. The detection and identification of defects at the early stage will helps smooth production process, saves production cost and time.

Real-Time Object Classification on an Enamel Paint Coating Conveyor Line Using Mask R-CNN

“…In [19], computer vision techniques were used to identify defective parts on an assembly line. The captured images were preprocessed to remove background artifacts, reduce noise, correct orientation, and provide scaling for further processing.…”

Assembly Line Anomaly Detection and Root Cause Analysis Using Machine Learning

“…Therefore, RCA has been used in numerous areas and is usually concerned with finding the root causes of events with safety, health, environmental, quality, reliability, production, and performance impacts [10], [11]. The process of RCA involves sorting the unstructured data and uncovering input, output relationships to identify the root causes and generally consists of the following four major steps [11]- [13]:…”

“…The texture and color characteristics are used to identify surface defects. On the other hand, shape dimensions measures are used to identify part damages, and both tasks are to ensure high-quality shipping products by notifying the robot controller to throw anomalous parts in the tray [12], [13].…”