• P.Ilakkiya
• Dr.B.Indrani

Road crack detection is one of the safeties measured for any country and this more important thing for complex road system. Most of the Indian roads are well connected with cities and urban locations. The urban roads are getting damaged frequently because of many factors. We have focused on Indian urban road damage detection system using Deep Convolutional Neural Network (DCNN). In this paper, we developed a model for road crack detection system for Indian urban roads. We have collected more than 700 road damaged images from various places of urban location from Tamil Nadu for testing the proposed model. In this paper, we have adapted DCNN algorithm for road crack detection and DCNN is an efficient algorithm for crack detection. The proposed method uses Histogram Equalization (HE) and Contrast Limited Adaptive Histogram Equalization (CLAHE) algorithm for improving the contrast level of road side images. This is first initiative taken for developing road damage detection system for urban road ways in India. We have performed a comparative analysis for the pre-processing phase by applying Histogram Equalization (HE) and Contrast Limited Adaptive Histogram Equalization (CLAHE). In this paper, we have train the model with 5000 road side images which comprises of both crack and non-cracking images. During the training phase, we have performed an image enhancement process by applying HE and CLAHE techniques. The enhanced images are trained by using Convolutional Neural Network (DCNN). In the testing phase, we have used 700 road side images of both cracking and non-cracking images. The testing phase follows the same steps for testing samples and the accuracy measured based on the correctly identified road cracking images.Theresult analysis shows that the proposed model works well for crack detection in Indian urban roads for the CLAHE based image enhanced testing samples. We have compared the performance for the proposed model with existing models, ResNet, VGG16, and VGG19with same dataset used for the proposed model. The performance for the proposed model has been measure based on accuracy, precision, recall and F1-scores. We have achieved 98.6%, 98.5%, 99.6% and 99% with respect to accuracy, precision, Recall, and F1 scores.

road damage detection, urban roads, Convolutional Neural Network, Deep Learning

"A Road Damage Detection based on Contrast Limited Deep Convolution Neural Networks", International Journal of Emerging Technologies and Innovative Research (www.jetir.org), ISSN:2349-5162, Vol.6, Issue 2, page no.450-470, February-2019, Available :http://www.jetir.org/papers/JETIR1902G61.pdf

"A Road Damage Detection based on Contrast Limited Deep Convolution Neural Networks", International Journal of Emerging Technologies and Innovative Research (www.jetir.org | UGC and issn Approved), ISSN:2349-5162, Vol.6, Issue 2, page no. pp450-470, February-2019, Available at : http://www.jetir.org/papers/JETIR1902G61.pdf