Treffer: Identification and classification of wild animals from video sequences using hybrid deep residual convolutional neural network.

In recent decades, wild animal classification from the video sequence is considered the trending research domain. Existing techniques utilize image processing for wild animal classification; however, the video-based classification is not much concentrated on any other previous research. So, in this research, we have concentrated on the wild animal classification using many videos that may contain accurate details about the wild animals. For such classification, Deep Residual Convolutional Neural Network (DRCNN) is integrated with (TSO) algorithm to perform the video processing, accurately classifying the different classes of wild animals using the Serengeti dataset. The video sequences are initially converted into video frames to initiate the wild animal classification. Then, unwanted noise from each video frame is removed using the Fast Average Peer Group (FAPG) filter in the pre-processing stage. Before the filtering process, all the images are resized into the same size. Next to the pre-processing, the threshold-based segmentation process is performed to subtract the background portion from the video frame. Then, features like colour, texture, etc., are extracted from the segmented image to perform the classification process. Finally, the extracted features are given input to the hybrid DRCNN-TSO algorithm for class label prediction. The TSO algorithm achieves the hyperparameter tuning of DRCNN. The proposed method has been executed in the Python platform. Finally, the performance of the proposed methodology is evaluated using the performance metrics (i.e. accuracy, false alarm rate, sensitivity, precision, F1 score, and false discovery rate), which are calculated using true positive (TP), false positive (FP), true negative (TN), and false-negative (FN) values. The obtained results are compared with existing techniques. Further, the Specificity, F1 score, false alarm rate and false discovery rate are compared with filtering and without filtering o show the efficiency of the proposed methodology. [ABSTRACT FROM AUTHOR]

Copyright of Multimedia Tools & Applications is the property of Springer Nature and its content may not be copied or emailed to multiple sites without the copyright holder's express written permission. Additionally, content may not be used with any artificial intelligence tools or machine learning technologies. However, users may print, download, or email articles for individual use. This abstract may be abridged. No warranty is given about the accuracy of the copy. Users should refer to the original published version of the material for the full abstract. (Copyright applies to all Abstracts.)