Research And Implementation Of Road Disease Recognition System Based On Improved YOLOv4 Algorithm

In recent years,with the development of China’s economy and social progress,the rapid expansion of urban scale and the increasingly perfect traffic network complement each other.Faced with such a large-scale road network,road maintenance departments are facing increasing challenges in road maintenance.severe.The timely detection of road diseases and the timely completion of maintenance work can reduce capital investment and extend the service life of roads to the greatest extent.The traditional manual detection is restricted by factors such as safety and efficiency,and can no longer meet the demand.The field of target detection has benefited from the rapid development of convolutional neural networks and intelligent algorithms,and has achieved remarkable achievements in many application scenarios.The efficient and accurate detection of road disease types can enable road maintenance departments to effectively carry out daily road work.Maintenance work is particularly important for the development of China’s transportation industry.Based on the learning of the basic theory in the field of convolutional neural network and target detection,this paper proposes an improved YOLOv4 road disease recognition algorithm.The YOLOv4 algorithm has a deep network level,complex structure and huge amount of parameters,which puts forward certain requirements for the performance of hardware devices.In this paper,the depthwise separable convolution,linear bottleneck and inverted residual structure mentioned in the Mobile Netv2 network are introduced into the YOLOv4 algorithm,which greatly reduces the model volume and parameter amount while maintaining excellent performance;at the same time,the channel attention mechanism is improved,fuse it with the spatial attention mechanism and introduce it into the output part of the backbone feature network,so that the model can achieve multi-category road diseases including cracks,cracks,pits,and road marking wear.Targeted,accurate and efficient identification.The experimental results show that the number of parameters of the improved YOLOv4 algorithm is reduced to 1/4 of the original,and its average accuracy can reach 78.34%.Finally,the lightweight algorithm Mobile Netv2-YOLOv4 is applied in the transportation business to identify road diseases.The road disease identification system based on the improved YOLOv4 algorithm can meet the expected standards in terms of function and performance.The system contains the following three functional modules: information input function module,system management function module,road disease identification function module.The road disease identification module is the core module.After completing the identification mode selection,users can upload the data of the corresponding category,display and save the identification results,and provide the historical data viewing function to facilitate subsequent viewing.The road disease recognition system based on the improved YOLOv4 algorithm can effectively improve the recognition efficiency.