At present, Automatic License Plate Recognition(ALPR) technology has been widely used in residential parking, high-speed intersection toll stations, roadside illegal parking, smart transportation and other fields. Although automatic license plate technology has been widely used in various fields, at present, whether it is commercial or academic methods, it is to explore the license plate recognition research of approximate frontal images in specific regions or specific countries (such as China, Brazil, and the United States). Aiming at real and complex scenarios, this paper builds a dataset for countries along the Belt and Road (such as Kenya, Nigeria, Togo, Ghana, etc.), called BR-ALPR dataset, designed to ALPR. We use yolov3 to complete the license plate detection. For license plate recognition, we use an improved Convolutional Recurrent Neural Network (CRNN) algorithm, which inserts the Spatial Transformation Network (STN) into the CRNN. In addition, we still use the method of template paste to complete the enhancement of the dataset. Experimental results show that our method is superior to advanced commercial methods for the detection and recognition of license plates in complex real Scenarios.