Cooperative perception represents an important technology to fulfil the higher automation levels of connected and automated mobility (CAM). In cooperative perception, the sensor data, either raw or processed, is shared among neighbour vehicles with the objective of enhancing or complementing the perception obtained by on-board sensors. The vehicle that requests this external perception data needs to have this data quickly. However, it first needs to discover the network address of the neighbour vehicle that wants to connect to. Specially in a dense urban area or in a congested radio channel, an inefficient method for neighbour vehicle discovery could prevent a timely start of the cooperative perception session. This paper describes a novel 5G multi-access edge computing (MEC) solution that that boosts the selection of interesting neighbour vehicles according to a geographical region of interest (ROI) after applying pertinent adjustments considering vehicles dynamics and network communication latencies. In contrast to broadcast-based methods, in the proposed method the vehicles are only sending their periodical position data to a MEC service, which centralises the vehicle discovery requests. The objective of this Vehicle Discovery Service (VDS) is to support the startup of Web Real-Time Communications (WebRTC)-based Extended Sensors CAM applications. The proposed VDS has been validated using a public vehicular traffic dataset evaluating geo-position accuracy. The WebRTC-based streaming pipeline has been validated testing its feasibility for a See-Through video streaming application.The growing amount of available car data brings new opportunities around the automotive industry. Hence, the exploitation of car data is strategic to achieve revenue generation, cost reduction and en-