In order to solve the scheduling problem of distributed processing data stream in Internet of vehicles, two kinds of improved Heterogereous Earliest Finish Time (HEFT) scheduling algorithm for multi-dimensional QoS (Quality of Service) were proposed. Firstly, the weighted directed acyclic graph model of distributed processing data stream task and the seven-dimensional QoS attributes weighted undirected topology model of distributed computing resources of Internet of vehicles were established. Secondly, the method of building lists in the classic HEFT scheduling algorithm was improved to a new one based on the priority of highest-level, minimum-successor tasks. Meanwhile, the seven-dimensional QoS attributes of computing resources in Internet of vehicles were grouped and reduced into two-dimensional composite attributes: the computing priority and communicating priority. Based on this improvement, two different improved HEFT scheduling algorithms due to the preferences of users for multi-dimensional QoS were proposed. Finally, the analysis of an example shows that the overall performance of these two multidimensional QoS improved HEFT scheduling algorithms are better than the classical HEFT scheduling algorithm and the round-robin scheduling algorithm.