Visualization enhancement of autonomous controlling vehicles system by thermal image processing technique

The visual camera in the autonomous vehicles is inadequate to capture a prominent image of objects, especially in the night. The headlights of the car also have limited perceptibility up to few meters. So, it is utterly essential to implement such a technique so that visibility in the night would also be cleared as crystal. In this paper, a thermal imaging camera is proposed to be felicitated in autonomous vehicles for better identification of objects especially in the night when visibility is very less. This technique is framed to use as a data acquisition tool to recognize the objects. Due to the huge variation in grayscales and pseudo coloring values in the thermal image, a fuzzy-based CNN model is proposed to be applied to identify the boundaries of the objects. In this technique, the correlation between the thermal images of the moving object and its types is proposed to be trained with the novel FCNN model. The acquired data from different scenarios would also be compared with driving safety experts. By inter-connecting these practices, a novel and unique real-time thermographic image processing would be framed for the autonomous vehicle system. The proposed technique is implemented in a grayscaled thermal image to verify the process and its analysis indicates the robustness of the proposed method.