Abstract
Most existing feature descriptors for video have limited representation ability. In order to improve the recognition accuracy of method for detecting the videos that include violent scenes and take advantage of the logical structure of video sequences, a novel feature constructing approach based on three dimensional histograms of gradient orientation (HOG3D), the Bag of Visual Words (BoVW) model, and feature pooling technology is proposed. This approach, combined with kernel extreme learning machine (KELM), can be used to detect violent scene. First, the HOG3D feature is extracted on the block level for video, and then the K-Means clustering algorithm is implemented to generate visual words. Then, the bag of visual words framework is used for the quantization of feature. And the feature pooling technology is operated to generate a feature vector for an entire video segment, and feature vectors of training data and testing data were used separately to train the model and evaluate the performance of the proposed approach. The experimental results showed that the proposed feature descriptor had good representation and generalization abilities. The proposed approach is efficient for violent scene detection, and the accuracy matches the best result on Hockey dataset, and it outperforms state-of-the-art on Movies.
Highlights
The rapid growth of the Internet has led to an increase in the number of user-generated videos (UGVs), and the need for filtering harmful content has augment significantly
The main contributions of this paper include: 1) A feature descriptor is constructed based on block scale HOG3D, bag of visual word and pooling technique; 2) The influences of different feature pooling techniques are compared through experiments on the well-known Hockey and Movies datasets; 3) The Kernel Extreme Learning Machine is used to solve the problem of detecting violent scene, and different types of kernel functions were studied; 4) Experiments using the proposed video descriptor and the different classifiers were conducted to check the validation of the descriptor and kernel extreme learning machine (KELM)
The basic procedure of testing is as follows: Step1: Step2: Step3: Step4: Extracting the features of HOG3D on the block level for the testing video clips; With the same method, the cluster centers are calculated by the K-means method, and the word frequency vectors are constructed for testing video sequences by using the Bag of Visual Words (BoVW) framework with the visual vocabulary obtained in training Steps 2 and 3; Feature pooling techniques are used to generate a fixed dimension feature for the test video clip; The trained KELM model obtained in the training procedure is used to detect the test violence video and to attain the accurate detection
Summary
The rapid growth of the Internet has led to an increase in the number of user-generated videos (UGVs), and the need for filtering harmful content has augment significantly. Their experimental results on Hockey indicated that the proposed deep neural network was effective for detecting violent scene problem. The main contributions of this paper include: 1) A feature descriptor is constructed based on block scale HOG3D, bag of visual word and pooling technique; 2) The influences of different feature pooling techniques are compared through experiments on the well-known Hockey and Movies datasets; 3) The Kernel Extreme Learning Machine is used to solve the problem of detecting violent scene, and different types of kernel functions were studied; 4) Experiments using the proposed video descriptor and the different classifiers were conducted to check the validation of the descriptor and KELM.
Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have