Application of computer vision to egg detection on a production line in real time.

Abstract

In this paper we investigate the application of computer vision to the problem of egg detection on a production line in real-time. For this purpose a dedicated software using Python3 and specialized libraries was designed and implemented that exploited the advantages of neural networks or template matching approaches. For neural network 3 models were investigated: SSD-Mobilenetv2, FR-CNN and YOLOv3. The effectiveness and performance of the egg detection, tracking and counting process were evaluated for these models and the template matching method. The sensitivity of the effectiveness and performance to the number of objects in a video frame, to film resolution and to size differentiation among the eggs were within the scope of the evaluation. An improvement was proposed relying on processing only every n-th video frame and optimal value of n that raised the performance without loss of effectiveness was investigated. In case of template matching method, the impact of cascade templates on the effectiveness and performance were also verified. Finally, the recommendation was done regarding the methods and their configuration applicable to be utilized on a production line in a real industrial environment with strict real-time requirements.