In order to solve the problem of the difficulty and unsafety of the artificial inspection of oil pipelines, an aerial vehicle detection and identification system based on improved SSD was designed and implemented. First, denoise the video images collected by the UAV; then use the SSD (Single Shot MultiBox Detector) algorithm to detect small targets to detect construction vehicles, but the original network VGG-16 parameters of the SSD algorithm are huge This causes a large amount of calculation. In order to improve the detection efficiency of construction vehicles, a lightweight MobileNet network is used to replace the VGG-16 basic network to form a MobileNet-SSD model. Since the shallow receptive field of the MobileNet network is low, in order to expand the receptive field and improve the detection accuracy, an algorithm model based on dilated convolution is proposed. Experiments show that the detection of construction vehicles has improved in speed and accuracy.