Road Target Recognition Based On MobileNet-SSD Model Technical Research And Its Android APP Development

In intelligent assisted driving and driverless systems,the detection and identification of road targets is very important.With the rapid development of artificial intelligence and deep learning,convolutional neural networks have achieved excellent results in the field of target detection and identification.The detection and identification of road targets based on convolutional neural networks has important research value and application value.The image in the driving recorder video contains rich road information,and it has important research significance to detect and identify road targets in the video.In this thesis,the deep learning method is used to detect and identify the road targets in the driving recorder video,and design it into Android application software.In order to implement such a system,The main work of this thesis is:1)The detection and recognition of the road target is completed on the computer client.The SSD target detection framework uses the VGG16 network as the backbone network to realize the detection and recognition of the target.In this thesis,the improved SSD detection framework is used,and the VGG16 network is replaced by the MobileNet network.The MobileNet-SSD network model is trained to realize the detection and recognition of the target.Three kinds of road targets are detected and identified: traffic signs,cyclists and pedestrians.There are 105 categories of these traffic signs.10 common categories for identification are selected.2)Because the image in the video of the driving recorder is relatively fuzzy,the definition is not high,which affects the recognition effect of the road target.In this thesis,the original image is deblurred by using Scale-recurrent network,and the image after deblurring is used for network training.The new MobileNet-SSD network model is used to detect and identify road targets.3)Implement Android application development,configure opencv environment in Android Studio software,use opencv dnn module,transplant the trained MobileNet-SSD network model to mobile phone,and realize detection and recognition of road target on Android application software.The analysis of the experimental results shows that the system can effectively detect and identify road targets.