Floor Water Detection Technology Based on Improved Resnet

Abstract

With the rapid development of artificial intelligence and the arrival of an aging population, the need for intelligent service robots is increasing. Older adults and the blind are susceptible to slipping due to weak vision, especially when they cannot see water on indoor floors. We use computer vision technology to solve this problem. However, it is challenging to detect floor water using existing target detection algorithms due to uncertainty about its shape and size. This paper proposes a floor water detection technology based on improved Resnet, which can be deployed on the intelligent service robot to remind the elderly and the blind to be careful When the service robot detects water in the ground. Our proposed topics and methods can significantly reduce the probability of the elderly and blind people slipping. The method proposed in this paper is 3.6% higher than the original Resnet18 and 8.1% higher than Mobilenetv2; the number of parameters in our method is only 8.5 percent of VGG16_bn and yet achieves similar performance to VGG16_bn. This paper suggests a new trajectory for intelligent service robots by detecting water on the floor, and it has demonstrated promising results in accuracy and speed. It is hoped that this paper will arouse more scholars’ interest in the detection technology of floor water.