Outdoor non-fixed scene weather identification method based on deep learning

Abstract

The invention discloses an outdoor non-fixed scene weather identification method based on deep learning. The method comprises the following steps: constructing a basic structure of a light-weight convolutional neural network; collecting pictures of various weather and making the pictures into a data set in a specific format; training the lightweight convolutional neural network by using the data set; and transplanting the trained lightweight convolutional neural network into an embedded platform or a mobile device, taking a shot weather picture as an input of the lightweight convolutional neural network, and outputting probabilities corresponding to various weather conditions. On one hand, the defect that a traditional method can only recognize weather of a fixed scene is overcome, and onthe other hand, due to the fact that the calculated amount is very small, the method can be applied to an embedded platform or mobile equipment and is very high in practicability.

Description

technical field [0001] The invention belongs to the technical field of navigation, and in particular relates to an outdoor non-fixed scene weather recognition method. Background technique [0002] All-source navigation is the application of various types of sensors to realize the rapid integration and reconfiguration of various combination solutions according to different environments and task requirements, thus forming a navigation system that can accurately position, navigate and time service in various complex environments. GPS. The optical sensor will be limited by the weather environment, such as heavy rain, snowstorm, sandstorm, smog, etc., which have a greater impact on lidar, laser altimeter, infrared rangefinder, etc. Night vision equipment will fail on sunny days, and cameras Photos taken at night in the dark cannot provide useful navigation information. If these sensors are applied in a non-adapted environment, the accuracy of navigation and positioning will be ...

AI Technical Summary

Problems solved by technology

The optical sensor will be limited by the weather environment, such as heavy rain, snowstorm, sandstorm, smog, etc., which have a greater impact on lidar, laser altimeter, infrared rangefinder, etc. Night vision equipment will fail on sunny days, and cameras Photos taken at night in low light conditions do not provide valid navigation information

If these sensors are applied in a non-adapted environment, the accuracy of navigation and positioning will be greatly reduced

Images