Research On Image Recognition Technology Of Key Signs In Assisted Driving In Specific Scenarios

The vehicle's auxiliary and automatic driving system is a basic research content of vehicle intelligence.The road identification lines and road identification are the key parts.In real life,there are a large number of cargo demand in areas such as ports.Due to the denseness of vehicles in this area,it may limit the visibility and cause traffic accidents.Therefore,in this paper,combined with the needs of actual engineering,the identification of key identifications in assisted driving is studied by special application scenarios such as trucks,trolley buses,and intelligent logistics vehicles.In this paper,the camera is used to capture the road ahead,and the road center line is detected and extracted by image preprocessing and image processing.At the same time,the road sign is detected and located,finally the Convolutional Neural Network is used on the road to determine the signs.The main research contents of this paper are as follows:Firstly,the research status of road markings is introduced.Then the current research method theory is analyzed,and the model of road key identification in specific scenarios is designed.Finally,the overall design scheme of this paper is proposed.Then the camera model is built and calibrated.Using the Inverse Perspective Mapping to convert the captured image into a bird's-eye view,and finally ordinary image preprocessing techniques are performed to eliminate external influences.After the image preprocessing is completed,the straight line detection algorithm is analyzed and compared,and Progressive Probabilistic Hough Transform is used to detect the road center line.Finally,the center line is fitted.By comparing and analyzing the existing circular detection algorithm,it is finally decided to use the contour and minimum enclosing circle algorithm to detect the shape features of the road marking,and locate the location of the marking.This paper introduces the Convolutional Neural Network and Back Propagation algorithm,introduces and analyzes the VGG model,ResNet model and Densenet model,adjusts the network parameters to achieve the desired classification effect,and finally compares the performance and accuracy of road identification by experiment.