# US 12,230,139 B2 · Predicting traffic violation hotspots using map features and sensors data

> Agents: this data is available as JSON from the patentgrep API and MCP server — see https://patentgrep.com/llms.txt

- Number: US 12,230,139 B2
- Document ID: US12230139B2
- Status: Granted
- Kind: B2
- Application number: 17/529,989
- Filed: 2021-11-18
- Published: 2023-05-18
- Granted: 2025-02-18
- Earliest claimed: 2021-11-18
- Assignees: HERE Global B.V.
- Applicants: HERE Global B.V.
- Inventors: Amarnath Nayak; Jerome Beaurepaire
- Web page: https://patentgrep.com/patent/US12230139B2

## CPC

- G08G 1/166: {for active traffic, e.g. moving vehicles, pedestrians, bikes}
- G06N 3/084: Backpropagation, e.g. using gradient descent
- G06N 20/00: Machine learning
- G08G 1/0112: {from the vehicle, e.g. floating car data [FCD]}
- G08G 1/0129: {for creating historical data or processing based on historical data}
- G08G 1/0133: {for classifying traffic situation}
- G08G 1/052: with provision for determining speed or overspeed {(speed measuring in general G01P)}
- G08G 1/096716: {where the received information does not generate an automatic action on the vehicle control}
- G08G 1/096888: {where input information is obtained using learning systems, e.g. history databases}
- G06N 3/0464: Convolutional networks [CNN, ConvNet]
- G06N 3/0475: Generative networks
- G08G: TRAFFIC CONTROL SYSTEMS (guiding railway traffic, ensuring the safety of railway traffic B61L; arrangement of road signs or traffic signals E01F9/00; radar or analogous systems, sonar systems, lidar systems specially adapted for traffic control G01S13/91, G01S15/88, G01S17/88; {radar or analogous systems, sonar systems, lidar systems specially adapted for anti-collision purposes G01S13/93, G01S15/93, G01S17/93})
- G06N: COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS

## Abstract

System and methods for determining traffic violation hotspots based on roadway feature and/or sensor data. Traffic violation hotspots include, for example areas where traffic violations are more likely to occur such as traffic light jumping, wrong way driving, over speeding, not wearing seatbelt, avoiding stop signs and red lights, distracted driving, passing other vehicles in a no-passing zone, among others. Traffic violation data from one region is used to train a model that is applied to a second region for which traffic violation data does not exist or is limited.

## Claims

### Claim 1

A method for predicting traffic violation hotspots, the method comprising:
acquiring historical traffic violation data for a plurality of locations in a first region;
training a model to identify traffic violation hotspots using the historical traffic violation data;
acquiring real-time data while traversing a location in a second region for which historical traffic violation data is limited;
generating, by the trained model using the real-time data, a probability score that represents how probable the location in the second region is to be a traffic violation hotspot; and
generating an update to a map displayed in a vehicle for the location based on the probability score, wherein the update to the displayed map includes one or more routing instructions for a different route associated with the second region.

### Claim 2 (depends on 1)

The method of claim 1, wherein the historical traffic violation data comprises map feature data and identified traffic violations, wherein the model is trained to learn which map features are relevant to traffic violations, wherein once those map features are identified and weighted in the trained model, the trained model is configured to generate the probability score based on the real-time data that comprises real-time acquired map feature data.

### Claim 3 (depends on 2)

The method of claim 2, wherein the map feature data comprises at least two of data related to a functional class of a road, category of a road, a number of lanes, a speed limit, a travel direction, or a presence of a physical divider.

### Claim 4 (depends on 1)

The method of claim 1, wherein a respective traffic violation hotspot comprises a location that exhibits a higher-than-average number of traffic violations than other locations in a defined region.

### Claim 5 (depends on 4)

The method of claim 4, wherein the traffic violations include at least one of traffic light jumping, wrong way driving, over speeding, not wearing seatbelt, avoiding stop signs and red lights, distracted driving, or passing other vehicles in a no-passing zone.

### Claim 6 (depends on 1)

The method of claim 1, wherein the historical traffic violation data comprises sensor data acquired by a vehicle and related traffic violations, wherein the model is trained to input real-time sensor data.

### Claim 7 (depends on 6)

The method of claim 6, wherein the sensor data comprises image data acquired by the vehicle.

### Claim 8 (depends on 1)

The method of claim 1, wherein the historical traffic violation data comprises sensor data acquired by a vehicle, map feature data, and identified traffic violations.

### Claim 9

A system for calculating a probability of a traffic violation, the system comprising:
one or more navigation devices configured to acquire sensor data related a plurality of locations on a roadway;
a mapping system comprising:
a geographic database configured to store map feature data for the plurality of locations; and
a server configured to store a first machine learning model configured to generate a first probability score when input real-time sensor data for a location for which there is limited traffic violation data, the server further configured to store a second machine learning model configured to generate a second probability score when input real-time mapping data for the location;
wherein the server is configured to provide the first probability score, the second probability score, or the first probability score and the second probability score to the one or more navigation devices, wherein the server is configured to provide an update to a map displayed in a vehicle for the location based on the probability score, wherein the update to the displayed map includes one or more routing instructions for a different route associated with the location.

### Claim 10 (depends on 9)

The system of claim 9, wherein the one or more navigation devices are configured to store a copy of the first machine learning model and a copy of the second machine learning model and apply the first machine learning model and the second machine learning model when traversing a region comprising at least the location to determine a risk of a traffic violation at an upcoming location for a device of the one or more navigation devices.

### Claim 11 (depends on 9)

The system of claim 9, where the server is configured to generate a combined probability score to provide to the one or more navigation devices, the combined probability score calculated as a weighted average of the first probability score and the second probability score.

### Claim 12 (depends on 9)

The system of claim 9, wherein the map feature data comprises at least two of data related to a functional class of a road, category of a road, a number of lanes, a speed limit, a travel direction, or a presence of a physical divider.

### Claim 13 (depends on 9)

The system of claim 9, wherein the traffic violation comprises a traffic light jumping, wrong way driving, over speeding, not wearing seatbelt, avoiding stop signs and red lights, distracted driving, or passing other vehicles in a no-passing zone.

### Claim 14 (depends on 9)

The system of claim 9, wherein the one or more navigation devices are further configured to generate an alert based on the first probability score, the second probability score, or the first probability score and the second probability score.

### Claim 15

An apparatus comprising:
at least one processor; and
at least one memory including computer program code for one or more programs; the at least one memory configured to store the computer program code configured to, with the at least one processor, cause the at least one processor to:
acquire data related to a location of a roadway for a first region;
generate, using at least one model trained using machine learning, a probability score for how probable is the location is to be a traffic violation hotspot, the at least one model trained using historical traffic violation data from a second region;
determine that the probability score exceeds a threshold score; and
generate an update to a map displayed in a vehicle for the location, wherein the update to the displayed map includes one or more routing instructions for a different route associated with the second region.

### Claim 16 (depends on 15)

The apparatus of claim 15, wherein the data related to the location comprises map feature data from the first region, wherein the at least one model is trained to learn which map features are relevant based on map features and identified traffic violations in the second region, wherein once the map features are identified and weighted in the at least one model, the at least one model is configured to determine the generate the probability score based on the acquired map feature data for the first region.

### Claim 17 (depends on 16)

The apparatus of claim 16, wherein the map feature data comprises at least two of data related to a functional class of a road, category of a road, a number of lanes, a speed limit, a travel direction, or a presence of a physical divider.

### Claim 18 (depends on 15)

The apparatus of claim 15, wherein the data related to the location comprises sensor data acquired by a vehicle from the first region, wherein the at least one model is trained to input sensor data and output the probability score that the location is traffic violation hotspot.

### Claim 19 (depends on 15)

The apparatus of claim 15, wherein the data related to the location comprises map feature data and sensor data acquired by a vehicle in the first region, wherein the at least one model comprises a first model and a second model, wherein the first model is trained to learn which map features are relevant based on map features and traffic violations in the second region and output a first probability score, wherein the second model is trained to input sensor data and output a second probability score, wherein the probability score is calculated as a weighted average of the first probability score and the second probability score.