Provided is a multitask fake plate vehicle vision detection system based on deep convolution nerve network, comprising a camera disposed on a city road, a traffic cloud server, and a fake plate vehicle reverse driving detection system. The camera is used for obtaining the snapshot image data at the city road and is disposed above a road. The traffic cloud server is used for receiving the road video data obtained from the camera, and submitting the road video data to the fake plate vehicle detection system for detection and identification. The fake plate vehicle detection system comprises a Faster R CNN-based vehicle positioning detection module, a vehicle type identification module, a fake plate positioning and identifying module, a license plate legality detection module, a logic consistency detection module, a vehicle inspection mark fine comparison module, and an alarm notification generating module. The invention also provides a multitask fake plate vehicle vision detection methodbased on deep convolution nerve network. The invention is advantageous in that vehicles with fake plates can be rapidly and accurately positioned, and the efficiency of criminal detection can be improved.