With the rapid development of China's economy， the mileage scale of roads and railways is becoming larger and larger. As an important node of the road network， bridges have also achieved very rapid development. At the same time， because the bridge structure will be eroded and damaged by strain， temperature and humidity， it often causes crack diseases in the bridge structure. Therefore， based on the above reasons， it is necessary to detect the crack disease of the bridge structure in order to reasonably arrange personnel to maintain and maintain the bridge. In this paper， the full convolution neural network model is designed， and the crack image data set is used to train and verify the model， so as to achieve the purpose of crack recognition， and has high accuracy and recall. The full convolution neural network model can be used to identify the forms of three bridge crack diseases， and has great detection advantages compared with other crack identification methods. The theoretical analysis of this paper and the identification and detection technology for bridge crack diseases have positive guiding significance and application value for relevant research.