Enterprise Employee Work Behavior Recognition Method Based on Faster Region-Convolutional Neural Network

Abstract

The original Faster Region-Convolutional Neural Network (R-CNN) model based on Convolutional Neural Network (CNN) is not effective enough to solve the challenge of identifying employee work behavior data sets. To overcome this limitation, an innovative optimization strategy is proposed in this paper. First, we replace the traditional Visual Geometry Group (VGG) network with a Residual Network (ResNet) to ensure that the features extracted from the image are more comprehensive and detailed. Then, multi-scale feature fusion technology combined with Convolutional Block Attention Module (CBAM) is used to predict the multi-layer feature layers and further strengthen the fused feature maps. This method makes the feature map contain both high-level semantic information and low-level detail information, and provides a richer feature description for small size targets. In order to further improve the accuracy of target detection, Regional of Interest (ROI) align technology is selected to replace the traditional ROI pooling method. Through these improvements, an enhanced version of Faster R-CNN algorithm is successfully constructed. In comparison experiments, the performance of the improved Faster R-CNN algorithm is evaluated against Support Vector Machine (SVM), Extreme Learning Machine (ELM), Single Shot MultiBox Detector (SSD) and the original Faster R-CNN algorithm. The results show that under the condition of similar recognition speed, the improved Faster R-CNN shows significant advantages in recall rate, accuracy rate and accuracy rate. Specifically, the processing time of the algorithm is kept within 0.40 seconds, and the accuracy and accuracy of the algorithm have reached a high level of more than 90%.