| Popis: |
Human fall detection is a problem to be solved in video processing, and it is applied to the intelligent transportation and the smart medical care. A new method that combines the yolov5 algorithm and the ResNet-50 network is proposed in this paper to realize real-time human fall detection in video images. First, the yolov5 algorithm is used to detect moving targets in video images, and then the detected images with moving human targets are preprocessed. Furthermore, the processed images are sent to the convolutional neural network-ResNet 50 to perform the classification of determining whether the human body falls. In this paper, ResNet-50 network uses UR Fall Datasets—an online open dataset—to process video frames into images, as sample data for training. Results show that the accuracy of the algorithm proposed in this paper can reach 93.9%, which has a good application effect in the field of fall detection. |
