Pothole detection system
Example implementations described herein are directed to depression detection on roadways (e.g., potholes, horizontal panel lines of a roadway, etc.) through using vision sensor to realize improved safety for advanced driver assistance systems (ADAS) and autonomous driving (AD). Example implementations described herein detect candidate depressions in the roadway in real time and adjust the control of the vehicle system according to the detected depressions.
1. A method, comprising:
determining a difference image for images of a roadway received from one or more cameras of a vehicle;
identifying one or more candidate depressions on the roadway from the difference image;
classifying the one or more candidate depressions to determine types of depressions for each of the one or more candidate depressions based on a comparison to a negative Gaussian distribution; and
controlling at least one of suspension, steering, or speed of the vehicle from the determined types of depressions for each of the one or more candidate depressions;
wherein the classifying the one or more candidate depressions to determine types of depressions for each of the one or more candidate depressions based on the comparison to the negative Gaussian distribution comprises:
identifying a blind area of the one or more candidate depressions based on applying the negative Gaussian distribution to an estimated slope of visible pixels of the one or more candidate depressions;
estimate a depth of the one or more candidate depressions based on the application of the negative Gaussian distribution; and
determining the types of depressions for the each of the one or more candidate depressions based on the estimated depth and an estimated shape of the one or more candidate depressions.
2. The method of claim 1 , wherein the one or more cameras is a single camera, and wherein the difference image is generated from the images of the roadway received from the camera through a machine learning process.
3. The method of claim 1 , wherein the one or more cameras comprises a plurality of stereo cameras, wherein the difference image is assembled from the images received from the plurality of stereo cameras.
4. The method of claim 1 , wherein the identifying one or more candidate depressions on the roadway from the difference image comprises:
comparing edge image pixels of the difference image with a threshold disparity map;
associating each of the edge image pixels with a weight based on the comparison; and
identifying the one or more candidate depressions based on the weight associated with the each of the edge image pixels.
5. The method of claim 1 , wherein the identifying one or more candidate depressions on the roadway from the difference image comprises:
for a failure of identifying the one or more candidate depressions on the roadway, obtaining, from a cloud system communicating with the vehicle, locations of depressions on the roadway and the types of depressions for each of the obtained depressions.
6. A non-transitory computer readable medium, storing instructions for executing a process, the instructions comprising:
determining a difference image for images of a roadway received from one or more cameras of a vehicle;
identifying one or more candidate depressions on the roadway from the difference image;
classifying the one or more candidate depressions to determine types of depressions for each of the one or more candidate depressions based on a comparison to a negative Gaussian distribution; and
controlling at least one of suspension, steering, or speed of the vehicle from the determined types of depressions for each of the one or more candidate depressions;
wherein the classifying the one or more candidate depressions to determine types of depressions for each of the one or more candidate depressions based on the comparison to the negative Gaussian distribution comprises:
identifying a blind area of the one or more candidate depressions based on applying the negative Gaussian distribution to an estimated slope of visible pixels of the one or more candidate depressions;
estimate a depth of the one or more candidate depressions based on the application of the negative Gaussian distribution; and
determining the types of depressions for the each of the one or more candidate depressions based on the estimated depth and an estimated shape of the one or more candidate depressions.
7. The non-transitory computer readable medium of claim 6 , wherein the one or more cameras is a single camera, and wherein the difference image is generated from the images of the roadway received from the camera through a machine learning process.
8. The non-transitory computer readable medium of claim 6 , wherein the one or more cameras comprises a plurality of stereo cameras, wherein the difference image is assembled from the images received from the plurality of stereo cameras.
9. The non-transitory computer readable medium of claim 6 , wherein the identifying one or more candidate depressions on the roadway from the difference image comprises:
comparing edge image pixels of the difference image with a threshold disparity map;
associating each of the edge image pixels with a weight based on the comparison; and
identifying the one or more candidate depressions based on the weight associated with the each of the edge image pixels.
10. The non-transitory computer readable medium of claim 6 , wherein the identifying one or more candidate depressions on the roadway from the difference image comprises:
for a failure of identifying the one or more candidate depressions on the roadway, obtaining, from a cloud system communicating with the vehicle, locations of depressions on the roadway and the types of depressions for each of the obtained depressions.
11. A vehicle system, comprising:
one or more cameras; and
a processor, configured to:
determine a difference image for images of a roadway received from one or more cameras of a vehicle;
identify one or more candidate depressions on the roadway from the difference image;
classify the one or more candidate depressions to determine types of depressions for each of the one or more candidate depressions based on a comparison to a negative Gaussian distribution; and
instruct an Electronic Controller Unit (ECU) to control at least one of suspension, steering, or speed of the vehicle system from the determined types of depressions for each of the one or more candidate depressions;
wherein the processor is configured to classify the one or more candidate depressions to determine types of depressions for each of the one or more candidate depressions based on the comparison to the negative Gaussian distribution by:
identifying a blind area of the one or more candidate depressions based on applying the negative Gaussian distribution to an estimated slope of visible pixels of the one or more candidate depressions;
estimate a depth of the one or more candidate depressions based on the application of the negative Gaussian distribution; and
determining the types of depressions for the each of the one or more candidate depressions based on the estimated depth and an estimated shape of the one or more candidate depressions.
12. The vehicle system of claim 11 , wherein the one or more cameras is a single camera, and wherein the difference image is generated from the images of the roadway received from the camera through a machine learning process.
13. The vehicle system of claim 11 , wherein the one or more cameras comprises a plurality of stereo cameras, wherein the difference image is assembled from the images received from the plurality of stereo cameras.
14. The vehicle system of claim 11 , wherein the processor is configured to identify one or more candidate depressions on the roadway from the difference image by:
comparing edge image pixels of the difference image with a threshold disparity map;
associating each of the edge image pixels with a weight based on the comparison; and
identifying the one or more candidate depressions based on the weight associated with the each of the edge image pixels.
15. The vehicle system of claim 11 , wherein the processor is configured to identify one or more candidate depressions on the roadway from the difference image by:
for a failure of identifying the one or more candidate depressions on the roadway, obtaining, from a cloud system communicating with the vehicle, locations of depressions on the roadway and the types of depressions for each of the obtained depressions.