Parking problem becomes one of the major issues in the city development. However, accuracy parking occupancy detection is one of the fundamental issues. This paper addresses this issue from the perspective of detecting the parking vehicles using magnetic sensors. Our research focuses on feature analysis of magnetic distortions induced by parking vehicles in different conditions. Besides, an accuracy parking detection solution based on signal feature and information collaborative mechanism has been proposed. By using signal feature of parking vehicle, more information of vehicle has been exploited to improve the performance of conventional state machine method. In order to improve the accuracy of detection, an information collaborative mechanism of adjacent magnetic sensors has been introduced to tackle the interferences from adjacent vehicles. To test our solution particularly in the challenging situations, real experiments were carried out in both under-ground and ground parking lot. The experiments reveal a high accuracy of about 99.77% in the un-ground and 99.54% in the ground parking plot. Our solution has a significant increase in detection accuracy.