The pavement surface condition assessment is a critical component for a proper pavement management system as well as for pavement rehabilitation design. A number of devices were developed to automatically record surface distresses in a continuous survey mode, but the software required for automatic distress identification remains a big challenge. In this study, a new method named PICture Unsupervised Classification with Human Analysis (PICUCHA) is proposed to circumvent many of the limitations of existing approaches, based on a combination of human and artificial intelligence. It was designed from scratch to be capable to identify sealed and unsealed cracks, potholes, patches, different types of pavements and others. The self-learning algorithms do not use any distresses predefinition and can process images taken by cameras with different brands, technologies and resolution. This study describes some key aspects of the new method and provides examples in which PICUCHA was tested in real conditions showing accuracy up to 96.9% in image pattern detection and classification.