When the unmanned aerial vehicle (UAV) performs infrared ship target reconnaissance, the detection algorithm has a great influence on the detection accuracy. In order to enhance the detection capability of the UAV′s infrared photoelectric load on the ship''s target, a support vector machine was used to detect the region proposal training algorithm to improve the accuracy of target detection. Region proposal classification data were obtained by training region features in advance. In the detection stage, the region classification data obtained by training was loaded, and the regions which were more likely to contain targets were selected, thereby improving the accuracy of the target detection. In the validation experiment, 368 long-wave infrared images captured by UAV were selected as data sets, and 139 infrared images were selected as test sets. The target detection experiments were carried out using the region proposal training method and non-region proposal training method. The detection results show that the mean accuracy is 14.6% higher when using the region proposal training method than when using the non-region proposal training method.