Improvement CNN Performance by Edge Detection Preprocessing for Vehicle Classification Problem

By B. Abdillah, Grafika Jati, Wisnu Jatmiko. 2018

5 citations recorded by the source corpus.

Vehicle classification is one of the most important task in Intelligent Transportation System (ITS). The large amount of vehicle data needs an approach that could classify correctly and effectively. A multiclass of vehicles in various conditions are an onerous challenge in classification. Furthermore, various shapes of vehicles that belongs to one class (divergence of the intra-class) makes vehicle classification task even more challenging. CNN based method is one of the most popular algorithms for recognizing image data. Conventional CNN and several vehicle classification algorithms still deal with shape variation problem. We proposed edge detection on CNN to strengthen the structural information of the vehicle. Canny edge detection is employed in the first layer of CNN. We also compared the performance of the proposed method with other preprocessing edge detection such as Gabor, Sobel, and Prewitt kernel. The proposed method achieved a better performance of 96% in accuracy compared to the state-of -the-art BIT Vehicle Dataset classification.

Subfields: Automotive Engineering, Computer Vision and Pattern Recognition, Media Technology

How researchers see this paper

The researchers whose trusted networks rank it highest, strongest advocate first. Every researcher gets a different order.

In Computer Vision and Pattern Recognition

In Media Technology

Citations in the corpus

ScaleNet: An Unsupervised Representation Learning Method for Limited Information (2023, background): “more on the edge information and specific semantic features, which efficiently improves image classification tasks [1,34].”