Method for Electric Vehicle Charging Port Recognition in Complicated Environment based on CNN

In the all-season indoor and outdoor background, facing the complicated environment formed by different lighting, partial blocked, pseudo-object interference, noise and other factors, the recognition and positioning of the charging port of an electric vehicle cannot be conventionally partitioned into a difficult problem. This paper studies the method for charging port recognition in a complex environment based on CNN, which not only ensures the accuracy and robustness of the recognition, but also provides a solution for accurately locating the charging port. The overall goal of the charging port recognition in this paper is to identify the category of the current image, and then identify the intensity of light for the image with the charging port. We built a sample set of charging port after the denoising of median filter, which is divided into four categories: complete, none, fake, and incomplete; In order to improve the generalization ability of the model, we add the number of LeNet-5 model and use the Relu activation function; Use the above two sample sets to train the models separately, save the models and parameters, and finally actually test. The experimental results show that the method uses the deep learning ability of convolutional neural network to automatically extract the features in the image, the recognition accuracy of the charging port is 99%, and the recognition accuracy of different light intensity is 100%. The integrity information and light intensity information are feedback to the automatic charging system, in order to accurately position the charging port subsequently, the camera position and the subtraction light strategy are adaptively adjusted to obtain a clearer image.