An Efficient Approach to Feature Extraction for Crowd Density Estimation
Crowd feature extraction is important step for crowd density estimation. This paper proposes a simple and novel approach of feature extraction applicable for crowd density estimation. A 5 × 5 mask is proposed for extraction of density, which finds isolated components in the image. This helps in classification of the density in five levels using SVM and ANN classifiers. The method can be used for intelligent surveillance system in public places. It can easily be used for embedded applications.
KeywordsSupport vector machine (SVM) Artificial neural network (ANN) Convolution of mask Isolated components Statistical features Thresholding
- 2.Saleh, S.A.M., Suandi, S.A., Ibrahim, H.: Recent survey on crowd density estimation and counting for visual surveillance, pp. 103–114. Elsevier Ltd. (2015)Google Scholar
- 4.Nemade, N.A., Gohokar, V.V.: Feature representation for crowd counting by regression: a review. Int. J. Adv. Res. Comput. Commun. Eng. 5(1), 402–405 (2016)Google Scholar
- 5.Nemade, N.A., Gohokar, V.V.: A survey of video datasets for crowd density estimation. In: International Conference on Global Trends in Signal Processing, Information Computing and Communication, pp. 389–395. IEEE (2016)Google Scholar
- 6.Nemade, N.A., Gohokar, V.V.: Crowd Density as Dynamic Texture: Behavior Estimation and Classification, Information and Communication Technology, AISC, vol. 625, www.springer.com/us/book/9789811055072 (2016)