Contextual knowledge to enhance workplace hazard recognition and interpretation in a cognitive vision platform

The combination of vision and sensor data together with the resulting necessity for formal representations builds a central component of an autonomous Cyber Physical System for detection and tracking of laborers in workplaces environments. This system must be adaptable and perceive the environment as automatically as possible, performing in a variety of plants and scenes without the necessity of recoding the application for each specific use. But each recognition system has its own inherent limits, especially those which task is to work in unidentified environments and deal with unknown scenarios and specifications. The platform described in this paper takes this into account by connecting the probabilistic area of event detection with the logical area of formal reasoning in a Cognitive Vision Platform for Hazard Control (CVP-HC). In order to support formal reasoning, additional relational scene information is supplied to the recognition system. In this platform, the contextual knowledge is used to improve the recognition and interpretation of detected events. This relational data together with all collected information is represented explicitly as a Set of Experience Knowledge Structure (SOEKS), categorized and stored as a Decisional DNA (DDNA), a decisional safety fingerprint of a company. By these means, the systems assesses and addresses critical unsafe behaviors whilst gives support to an explicit long term culture change process. By the use of context the CVP-HC is capable adjust accordingly without the need of rewriting the application’s code every time conditions or specifications changes.