Traffic Sign Recognition Algorithm Based On Local Feature Fusion

In recent years,with the rapid development of China in the field of artificial intelligence,the construction of smart cities has begun to take shape,and intelligent transportation systems play a pivotal role in smart cities,for urban passenger flow and logistics.The smooth flow plays an important role.In the field of intelligent transportation,the intelligent detection and identification of traffic signs has attracted the attention and research of many engineers and academics,and related intelligent products have been initially applied in the field of unmanned driving.However,due to the many obstacles such as damage,foreign object occlusion,fading,and motion blur in the actual traffic scene,many engineers and scholars have conducted in-depth exploration and research on related issues.Results,but the technical results are not yet mature.Due to traffic congestion and accidents,the situation in the country with a large population such as China is becoming more and more serious.These problems need to be solved urgently.Therefore,the research on intelligent detection and identification of traffic signs has important theoretical guidance and practical significance for the planning and construction of current and future smart cities.The thesis conducts in-depth research on the important theories and key technologies involved in smart transportation.Focus on the following aspects and related research results:(1)Aiming at the fact that the image acquired by the camera will produce a certain motion blur during the real vehicle travel,a blind restoration method based on the sparse representation and the Fichner theorem is proposed.Firstly,the algorithm uses the filter to predict the edge of the blurred image significantly.Then,the blind restoration model generated by the preprocessing is sparsely regularized,and the Faithner theorem is used to blindly recover the collected samples.The results show that the method can effectively reduce the motion blur and visual artifacts of the image itself,and can obtain good restoration results.(2)Since there is a large difference in the number of different types of traffic signs in the normal sample set,the detection performance of the classifier is weakened.To solve this problem,thesis proposes an MLBP-HOG traffic sign detection algorithm based on regional feature fusion.Firstly,the MLBP features and HOG features in the sample are extracted by using the characteristics of local features.Then,the acquired two types of features are merged.Finally,the extracted fusion features are input and the SVM classifier is pre-trained to accurately target the region of interest.Ground location and detection.In this paper,the German public data set GTSRB is used for the experimental test.The experimental results show that the method has the advantages of less time-consuming and high regional detection accuracy.(3)Quickly and accurately identify the classification is the ultimate goal of traffic sign recognition.Therefore,thesis proposes an extreme learning machine recognition algorithm based on MLBP-HOG feature fusion.Firstly,the HOG features,LBP features,MLBP features and MLBP-HOG fusion features in the sample set are extracted sequentially,and then the extracted features are input into the SVM(Support Vector Machine)classifier,the CNN classifier and the ELM classifier for training..Through experimental analysis,it is proved that ELM classifier based on MLBP-HOG fusion feature can obtain excellent classification results in complex scenes.