Object Recognition in Imagery

My company, CivilView, uses drones to generate high resolution orthomosaics. Several of our projects have involved collecting data over the same area multiple times for comparison purposes. I would like to develop a platform that would scan the image and note objects of import. An example would be the presence of cars for a parking study like in this example: [login to view URL] The imagery on its own is valuable, but developing a way quickly analyze parking usage would be extremely beneficially. The idea would be to scan each image, acknowledge the presence of cars in the parking spaces or not, and then generate some sort of report on the findings of the scan.