Systems and methods for detection, classification, and geolocation of traffic objects
Systems and methods of mapping traffic related objects. In one embodiment, a method includes: receiving image sensor data relating to an environment associated with a vehicle; receiving lidar sensor data relating to the environment associated with the vehicle; selecting, by a processor, at least one image region including a traffic related object from the image sensor data; selectively filtering, by a processor, the at least one image region based on the lidar sensor data; and generating, by a processor, a composite map including a representation of the traffic related object superimposed upon a defined map of the environment based on the selectively filtering.
1. A method of mapping traffic related objects, comprising:
receiving image sensor data relating to an environment associated with a vehicle;
receiving lidar sensor data relating to the environment associated with the vehicle;
identifying, by a processor, a traffic control object from the image sensor data based on an identified letter, an identified word, and a shape, wherein the traffic control object includes a traffic cone, a sign, and a barricade;
selecting, by a processor, at least one image region including the traffic control object from the image sensor data;
selecting, by a processor, a first subset of points of the lidar sensor data by projecting lidar points of the lidar sensor data onto the image sensor data and keeping a second subset of points from the first subset of points that fall within the image region;
selectively filtering, by a processor, the at least one image region based on the second subset of points of the lidar sensor data, wherein when a first number of the second subset of points fall within a height range and a second number of the second subset of points indicate a reflective value, the at least one image region is selected, wherein when none of the second subset of points fall within a height range or none of the second subset of points indicate a reflective value, the at least one image region is discarded; and
generating, by a processor, a composite map including a representation of the traffic control object superimposed upon a defined map of the environment based on the selectively filtering.
2. The method of claim 1 , further including localizing the traffic control object based on the lidar data.
3. The method of claim 2 , further comprising determining a centroid position of the subset of points and wherein the localizing is based on the centroid position.
4. The method of claim 1 , wherein selecting the at least one image region is based on at least one machine learning method.
5. The method of claim 4 , wherein the at least one machine learning method includes a convolutional neural network.
6. The method of claim 5 , further comprising training the convolutional neural network based on results of the selectively filtering.
7. The method of claim 1 , wherein the identifying the traffic control object and the selecting the image region is based on a region proposal network.
8. A system for controlling a vehicle, comprising:
an image sensor that generates image sensor data relating to an environment associated with a vehicle;
a lidar sensor data that generates lidar sensor data relating to the environment associated with the vehicle; and
a control module having a processor configured to identify a traffic control object from the image sensor data based on an identified letter, an identified word, and a shape within the image sensor data, wherein the traffic control object includes a traffic cone, a sign, and a barricade, select at least one image region including the traffic control object from the image sensor data, select a first subset of points of the lidar sensor data by projecting lidar points of the lidar sensor data onto the image sensor data and keeping a second subset of points from the first subset of points that fall within the image region, selectively filter the at least one image region based on the second subset of points of the lidar sensor data, wherein when a first number of the second subset of points fall within a height range and a second number of the second subset of points indicate a reflective value, the at least one image region is selected, wherein when none of the second subset of points fall within a height range or none of the second subset of points indicate a reflective value, the at least one image region is discarded, generate a composite map including a representation of the traffic control object superimposed upon a defined map of the environment based on the selectively filtering, and control the vehicle based on the composite map.
9. The system of claim 8 , wherein the control module is further configured to localize the traffic control object based on the lidar data.
10. The system of claim 8 , wherein the control module selects the at least one image region based on at least one machine learning method.
11. The system of claim 10 , wherein the control module determines a centroid position of the subset of points and wherein the localizing is based on the centroid position.
12. The system of claim 8 , wherein the at least one machine learning method includes a convolutional neural network.
13. The system of claim 12 , wherein the control module is further configured to train the convolutional neural network based on results of the selectively filtering.
14. The system of claim 8 , further comprising an interactive interface that receives, by a processor, user input for classifying the image region.
15. The system of claim 8 , wherein the control module is further configured to localize the traffic control object to real world coordinates based on a location and a heading of the vehicle.
16. The system of claim 8 , wherein the control module identifies the traffic control object and the selecting the image region is based on a region proposal network.