Verified integrated vehicle identification system and method thereof

A method for automatically reconciling disparities identified in events captured by a tolling system, the method comprising:

• providing, to each on-site server of a plurality of different on-site servers, a first neural network executable by each of the plurality of different on-site servers, wherein each of the plurality of different on-site servers corresponds to a different tolling station comprising one or more cameras, and wherein the first neural network is configured to perform digital conversions of images of license plates to numeric strings;
• receiving, by a central server, from a first on-site server of a first tolling system, and after providing the first neural network:
• an image, captured via a first camera of the first tolling system at a first time, of a license plate of a vehicle,
• first vehicle classification data, and
• a first distributed numerical array, wherein the first distributed numerical array was generated, by the first on-site server, by processing, using the first neural network, the image of the license plate wherein the first distributed numerical array comprises an array of numbers representing a region of interest of the image of the license plate;
• generating, by the central server, a second distributed numerical array by processing, using a second neural network, the image of the license plate;
• determining, by the central server and using a third neural network:
• a first confidence level corresponding to a match between the first distributed numerical array and a first unique vehicle record of a plurality of unique vehicle records;
• a second confidence level corresponding to a second match between the second distributed numerical array and a second unique vehicle record of the plurality of unique vehicle records;
• identifying a matching unique vehicle record of the plurality of unique vehicle records by selecting, based on the first confidence level and the second confidence level, one of the first distributed numerical array or the second distributed numerical array;
• receiving, by the central server, via a transponder reader of the first tolling system, and at a second time after the first time, transponder data and second vehicle classification data corresponding to the matching unique vehicle record;
• identifying, by the central server, a vehicle classification disparity by:
• identifying, by the central server, at least one difference between the first vehicle classification data, the second vehicle classification data, or third vehicle classification data corresponding to the matching unique vehicle record;
• iteratively improving the second neural network by modifying, based on the vehicle classification disparity, one or more configurations of the third neural network to tune future correlations between numerical arrays and the plurality of unique vehicle records;
• and
• modifying, based on the at least one difference, the matching unique vehicle record.