Video Analytics Based Guidance System for Efficient Production Lines

Abstract

Video Analytics involves the use of a camera to detect various components. Video Analytics can also be used as a guidance system in any hand assembly process of a factory. Every year, a certain amount of loss is always contributed by human errors in any semi-automated factory. For example in a small hand assembly process, a small component which is not placed in the chassis may not sound too bad. But when the same product moves on to the next stage, the component missed may form the base for that stage and it may also damage the next set of machines processing it. This causes a huge loss for the Organization. Also if somehow the same product goes undetected to the client, it might malfunction. In order to reduce these errors, we have come up with a guidance system which checks each action of the worker with an instruction feed running along it to guide the operator. The System consists of a camera placed above the chassis. Camera setup is connected to a Next Unit of Computing (NUC) where the object detection algorithm runs. The NUC in turn is connected to a Monitor. The monitor will display a User Interface which will have instructions running in it. This User interface forms the actual Guidance System. Here, once a particular component is detected, the UI moves onto the next instruction and the operator has to place the next component for detection. The detection of each component triggers the next step of the process via instructions. The transition to the next step occurs only when the system detects a particular component. The component/object detection is done using a Deep Learning Model which is trained with a pre-defined dataset [1]. The Deep learning used for the project work is Single Shot Detector [2]. The UI is designed using PYGAME, a library for GUI creation. The proposed system integrates Object detection with a user interface