Neural network for the detection of misplaced and missing regions in images

This paper presents a neural-network-based approach for the detection of misplaced and missing regions in images. The main objective of this project is to develop an intelligent system that can identify a misplaced or missing region of a tested image. The system can be used to detect misplaced and missing components of printed circuit boards during the manufacturing process. Jigsaw puzzle pieces can be used to simulate the scenario of misplaced and missing components. The developed system consists of image processing and neural network phases. During image processing, the captured image is split into regions and the red-green-blue (RGB) value of the regions is obtained. The back-propagation neural network is used and is trained by the scaled conjugate gradient training algorithm. The neural network uses the RGB value obtained from image processing and analyzes regions to determine if a jigsaw puzzle piece is missing or misplaced. Two experiments on time performance have been conducted to analyze regions and the ability of the system to detect misplaced and missing jigsaw puzzle pieces. Results showed that the system requires approximately 89 s to process 20 pieces of the image. The system exhibits almost 100% accuracy in detecting missing or misplaced regions of a jigsaw puzzle image.