A Convolutional Neural Network Combined with Local Binary Pattern and Self-Attention Mechanism based on MC4LDevicefor Indoor Positioning

In earlier research, we proposed a vision-based ranging algorithm based on a Monocular Camera and four Lasers (MC4L) device for indoor positioning in dark environment call as Logarithmic Regression Algorithm (LRA). The linear relationship between the irradiation area and the real distance is established based on a LRA to control the positioning error within 2.4 cm. However, limited by the ranging mode of MC4Ldevice, the indoor positioning algorithm cannot distinguish whether the measured object is a wall or an obstacle. Hence, its application in environments with obstacles is limited. In order to address this issue, we proposed a Convolutional Neural Networks (CNNs) combined with a Local Binary Pattern (LBP) and self-attention mechanism called as LBP-CNNs model. This LBP-CNNs model can achieve distance measurement and obstacle recognition by modifying activation function and loss function of output layer. Experimental results show that the LBP-CNNs model can reduce the indoor positioning error to 1.27 cm, and the obstacle recognition accuracy reaches 92.3%.