With the rapid development of technology networks, the security problem of wireless networks is becoming more serious. Among the existing related research works on wireless network intrusion detection, Artificial Neural Network (ANN) and Deep Neural Network (DNN) are the most widely used models which can better handle multidimensional data without considering the influence of feature correlation on the results. Therefore, this study proposes an approach that calculates the feature correlation to convert non-image data into images and combines feature selection and feature extraction to improve the accuracy of CNN model detection. This approach calculates the Pearson, Spearman, and Kendall correlation coefficients to sort the features into 1D or 2D images according to the highest to the lowest correlational order. The experiment adopts the Aegean Wi-Fi Intrusion Dataset (AWID). The result shows that the best accuracy of the CNN is 99.574%, and the average accuracy is much higher than that of the DNN for non-image data.