Convolutional neural network-based road adhesion coefficient identification method

Abstract

The invention discloses a convolutional neural network-based road adhesion coefficient identification method in the field of image identification. In most of the existing technologies, dynamic modeling is adopted to indirectly estimate the road adhesion coefficients, and according to the method disclosed by the invention, an image processing method is adopted to directly identify the road adhesion coefficients, so that defects in the prior art can be overcome. The method comprises the following steps of: acquiring different road condition pictures; calibrating a road adhesion coefficient; establishing a road condition picture database; judging a position calibrated by a feature area of each road condition picture in the database by utilizing an image segmentation algorithm; extracting sensitive areas with different pixel sizes; inputting the sensitive areas which serve as training samples into a convolutional neural network to carry out training; and finally identifying road conditions by utilizing the trained convolutional neural network so as to obtain the road adhesion coefficient. Compared with the prior art, the method has the advantage that road adhesion coefficients (the road adhesion coefficients are assumed to be known in most conditions during trajectory tracking) are identified by adoption of images, so that the safety and comfort of drivers in the driving process are strengthened.

Description

technical field [0001] The invention belongs to the field of image recognition of road surfaces, and in particular relates to a method for recognition of road surface adhesion coefficient based on a convolutional neural network. Background technique [0002] The road surface adhesion coefficient not only affects the acceleration performance and braking performance of the vehicle, but also affects the stability of the vehicle when driving. Real-time identification of the road surface adhesion coefficient can greatly improve the safety and comfort of the vehicle during driving. The identification of road surface adhesion coefficient is one of the key technologies necessary for electronic stability control system (ESC). Due to the influence of road surface characteristics and other aspects, road surface adhesion coefficient has become one of the most difficult to accurately estimate the key parameters of vehicles[1,2 ]. [0003] At present, there are mainly two methods for est...

AI Technical Summary

Problems solved by technology

However, literature [5,6,7] only considered the longitudinal dynamics of the vehicle, and did not consider the influence of lateral dynamics on the coupling of the road adhesion coefficient, so there is a certain estimation error

This method adopts an adaptive mechanism method proposed in literature [8] to estimate the road surface adhesion coefficient in real time. This method can estimate the road surface adhesion coefficient in real time, but because the tire cornering stiffness will only change when it slips, so This method is not suitable for the estimation of road surface adhesion coefficient during normal driving

Images