Machine Learning for an Assembly Line Application

Online inspection of moving objects and count is an ongoing interest to industry. This paper has taken up a study of inspection and counting of processing products on a conveyor using MATLAB and their toolboxes. This model is aimed to ease inventory management, obtain manufacturing data on large-scale productions and automate the manual process of counting and inspection in an industrial environment to overcome time and cost constraints. The model works by initially pre-processing the video input file and then using image processing algorithms to identify the moving objects within the video frame. The identified moving objects are labelled with numbers and bounding boxes. After feature extraction the model is trained to inspect the objects with a sufficient dataset. Once training and testing are completed, the model can be used to inspect objects for faults and defects. This model can be used as a solution to automate manual processes in an assembly plant.