Object-Location Prediction based on CIE Color Difference for Deep Reinforcement Learning

In this work, we propose an Object-Location Prediction based on Color Difference from International Commission on Illumination (OLP-CIE) to predict the location of the object in the image. The predicted location based on OLP-CIE can improve the iterations of the training steps for the deep reinforcement learning (DRL). In OLP-CIE, the CIEDE 2000 formula is used to compare and calculate the color difference of the object color. Then, the calculation result of the color difference is converted into the feature map. Finally, the location of the object can be predicted by the analysis of the feature map. Compared to the related works, the accuracy of OLP-CIE can be maintained with simpler algorithm and fewer hardware operations.