Automatic human face recognition system of image processing based on BP neural network paradigm

Observation video analysis is useful in recognising human faces in crowded and coinciding scenarios. Overlapping images result in false recognition due to non-semantic textural features. The boundary analysis varies for this process, generating segments exceeding masks of the original image. Backpropagation learning (BPL) based textural-edge detection and recognition model (TED-RM) is designed to resolve this issue. The proposed model exploits the masked and un-masked textural features for identifying the semantics of the input. After this identification process, appropriate features are analysed for semantics and correlation with the inward and overlapping video image input edges. The masked and un-masked regions' semantic features are recurrently correlated with the previous datasets for independent human faces. The mapping feature points are identified and correlated with the actual edge of the training input. The non-semantic edge points are classified for further training and validation to detect errors in further input analysis. The proposed TED-RM improves 10.84% high accuracy, 11.5% less processing time, 10.2% high true positives, 5.55% less error, and 10.6% high recall compared to existing methods.