A low-power cooperative parking license plate recognition system

Abstract

A low-power collaborative parking license plate recognition system includes a sensing device, a parking camera module, a back-end server, and a low-power processing unit. The sensing device generates a wake-up signal when it detects a vehicle entering a target parking space. The parking camera module is positioned in a parking pillar adjacent to the target parking space and captures a parking image containing the target parking space upon receiving the wake-up signal. The back-end server has an AI image recognition system to receive the parking image and perform parking recognition to output a corresponding license plate recognition result. Therefore, this system only activates the parking camera module when an actual parking event occurs, thereby reducing power consumption.

Claims

1. A low-power cooperative parking license plate recognition system for parking identification in a target parking space, comprising: A sensing device is configured in the target parking space. The sensing device has an identification code and generates a wake-up signal when it detects a vehicle entering the target parking space. A parking camera module is configured in a parking pillar adjacent to the target parking space and has a field of view covering the target parking space. When it receives the wake-up signal from the sensing device, the parking camera module switches from a sleep state to an active state to capture a parking image containing the target parking space. A back-end server has an AI image recognition system for receiving the parking image and performing parking recognition to output a corresponding license plate recognition result. A low-power processing unit connects the sensing device, the parking camera module, and the back-end server and processes the signal transmission and collaborative operation of the sensing device, the parking camera module, and the back-end server.

2. The low-power cooperative parking license plate recognition system as described in claim 1, wherein, The identification code of the sensing device can be copied and written to another sensing device.

3. The low-power cooperative parking license plate recognition system as described in claim 1, wherein, The sensing device includes at least one sensor, which is selected from the group consisting of a geomagnetic sensor, a millimeter-wave radar sensor, a light sensor, an infrared sensor, and a mobile communication signal strength sensor, and the wake-up signal is determined by the at least one sensor to generate it.

4. The low-power cooperative parking license plate recognition system as described in claim 3, wherein, The at least one sensor is preferably selected from a geomagnetic sensor, and when the parking camera module is in the sleep state, only the low-power processing unit and the geomagnetic sensor among the at least one sensor are enabled. When the geomagnetic sensor among the at least one sensor is triggered, the at least one sensor begins to decide whether to generate the wake-up signal.

5. The low-power cooperative parking license plate recognition system as described in claim 4, wherein, When the geomagnetic sensor in at least one of the sensors is not triggered, the low-power processing unit keeps the parking camera module in a sleep state.

6. The low-power cooperative parking license plate recognition system as described in claim 1, wherein, The sensing device transmits the wake-up signal to the parking camera module via Bluetooth, a mobile communication network, or a low-power wide area network.

7. The low-power cooperative parking license plate recognition system as described in claim 1, wherein, The AI ​​image recognition system includes a license plate localization model and a license plate sequence recognition model, and the license plate localization model and the license plate sequence recognition model are object detection models based on the Convolutional Neural Network (CNN) architecture.

8. The low-power cooperative parking license plate recognition system as described in claim 7, wherein, The license plate localization model generates at least one license plate candidate region for the parking image, selects a license plate region based on the recognition confidence value corresponding to the at least one license plate candidate region, and extracts a license plate image data based on the portion of the parking image corresponding to the license plate region. The license plate sequence recognition model is used to recognize the license plate image data.

9. The low-power cooperative parking license plate recognition system as described in claim 8, wherein, The AI ​​image recognition system sets a parking area range corresponding to the target parking space in the parking image, and determines whether the vehicle is parked in the target parking space based on whether the license plate area selected by the license plate positioning model is within the parking area range.

10. The low-power cooperative parking license plate recognition system as described in claim 9, wherein, The parking area is defined as either a rectangle or a polygon.

11. The low-power cooperative parking license plate recognition system as described in claim 9, wherein, The license plate area is determined by its center point to determine whether it is within the parking area.

12. The low-power cooperative parking license plate recognition system as described in claim 7, wherein the license plate sequence recognition model directly generates a license plate character sequence through a sequential convolutional neural network.

13. The low-power cooperative parking license plate recognition system as described in claim 1, wherein, When the parking camera module switches to this activated state, the duration of the activated state is between ten and twenty seconds.

14. The low-power cooperative parking license plate recognition system as described in claim 1, wherein, The parking camera module switches to the active state every preset time period and captures the parking image.

15. The low-power cooperative parking license plate recognition system as described in claim 1, wherein, The parking camera module transmits the parking image to the backend server via Bluetooth, a mobile communication network, or a low-power wide area network.

Images