Method and system for detecting and classifying lanes
The present disclosure is related in general to field of machine learning and image processing, that provides a method and system for detecting and classifying lanes. A lane detection and classification system receives an input image from a data source and segments the input image into plurality of segments using a trained semantic segmentation model. Further, one or more lane markings are detected in the segmented image and, lane pattern and lane colour of each of the one or more lane markings, wherein each lane marking is associated with a priority. Subsequently, a binary image comprising lane markings of ego lanes of the host vehicle is generated and coefficient values of the ego lanes of the host vehicle are determined based on the priority value associated with the lane markings of the ego lanes and current position of the host vehicle, using a trained Convolutional Neural Network (CNN) model.
1. A lane detection and classification system for detecting and classifying lane patterns, the lane detection and classification system comprising:
an image sensor mounted to a host vehicle, to capture images of an area in front of the host vehicle;
a processor communicatively connected to the image sensor and configured to:
receive an input image from a data source, wherein the input image is an RGB image captured by the image sensor;
segment the input image into plurality of segments using a trained semantic segmentation model;
detect one or more lane markings in the segmented image and, lane pattern and lane colour of each of the one or more lane markings, wherein each of the one or more lane markings is associated with a priority value based on the corresponding lane pattern and lane colour;
generate a binary image comprising lane markings of ego lanes of the host vehicle, wherein the lane markings of ego lanes are extracted from the one or more lane markings detected in the segmented image; and
determine coefficient values of the ego lanes of the host vehicle based on the priority value associated with the lane markings of the ego lanes and current position of the host vehicle, using a trained Convolutional Neural Network (CNN) model.
2. The lane detection and classification system as claimed in claim 1 , wherein the coefficient values of the ego lanes indicate offset of the ego lanes from the current position of the host vehicle, slope of the ego lanes and curvature of the ego lanes.
3. The lane detection and classification system as claimed in claim 1 , wherein the ego lanes comprises a left ego lane and a right ego lane.
4. The lane detection and classification system as claimed in claim 1 , wherein the processor is further configured to convert the binary image from a two-dimensional form to a one-dimensional form, when determining the coefficient values of the ego lanes of the host vehicle.
5. The lane detection and classification system as claimed in claim 1 , wherein the processor is further configured to generate warning signals to the host vehicle based on the coefficient values of the ego lanes of the host vehicle and the lane pattern and the lane colour of each of the one or more lane markings.
6. The lane detection and classification system as claimed in claim 1 , further wherein the processor is further configured to perform upscaling and downscaling operations on the input image, during segmentation, using the trained semantic segmentation model.
7. The lane detection and classification system as claimed in claim 1 , wherein training the CNN model comprises determining error in determined coefficient values of ego lanes using predefined loss minimization techniques for optimizing the determination of coefficient values in subsequent computations, wherein the training is based on trial RGB images.
8. The lane detection and classification system as claimed in claim 1 , wherein the processor is further configured to overlay graphical lanes on the input image based on the coefficient values of the ego lanes and the lane pattern of the ego lanes, to indicate position of the ego lanes in the input image.
9. An Advanced Driver-Assistance System (ADAS) comprising:
a lane detection and classification system;
a warning generating system; and
a vehicle controlling system;
the lane detection and classification system comprising:
an image sensor mounted to a host vehicle, to capture images of an area in front of the host vehicle;
a processor communicatively connected to the image sensor and configured to:
receive an input image from a data source, wherein the input image is an RGB image captured by the image sensor;
segment the input image into plurality of segments using a trained semantic segmentation model;
detect one or more lane markings in the segmented image and, lane pattern and lane colour of each of the one or more lane markings, wherein each of the one or more lane markings is associated with a priority value based on the corresponding lane pattern and lane colour;
generate a binary image comprising lane markings of ego lanes of the host vehicle, wherein the lane markings of ego lanes are extracted from the one or more lane markings detected in the segmented image; and
determine coefficient values of the ego lanes of the host vehicle based on the priority value associated with the lane markings of the ego lanes and current position of the host vehicle, using a trained Convolutional Neural Network (CNN) model;
the warning generating system associated with the lane detection and classification system is configured to:
generate warning signals to the host vehicle based on the coefficient values of the ego lanes of the host vehicle and the lane pattern and the lane colour of each of the one or more lane markings; and
the vehicle controlling system associated with the warning generating system and the lane detection and classification system is configured to:
receive warning signals related to the host vehicle from the warning generating system; and
control navigation of the host vehicle, in real-time, based on the received warning signals.
10. The ADAS as claimed in claim 9 , wherein controlling the host vehicle comprises at least one of lane shifting, overtaking other vehicles, moving from one side to another side of the road within the ego lanes of the host vehicle and maintaining safe distance of the host vehicle from the ego lanes.
11. A method of detecting and classifying lanes, the method comprising:
receiving, by a processor communicatively connected to an image sensor mounted to a host vehicle, an input image from a data source, wherein the input image is an RGB image captured by the image sensor;
segmenting, using the processor, the input image into plurality of segments using a trained semantic segmentation model;
detecting, using the processor, one or more lane markings in the segmented image and, lane pattern and lane colour of each of the one or more lane markings, wherein each of the one or more lane markings is associated with a priority value based on the corresponding lane pattern and lane colour;
generating, using the processor, a binary image comprising lane markings of ego lanes of the host vehicle, wherein the lane markings of ego lanes are extracted from the one or more lane markings detected in the segmented image; and
determining, using the processor, coefficient values of the ego lanes of the host vehicle based on the priority value associated with the lane markings of the ego lanes and current position of the host vehicle, using a trained Convolutional Neural Network (CNN) model.
12. The method as claimed in claim 11 , wherein the coefficient values of the ego lanes indicate offset of the ego lanes from the current position of the host vehicle, slope of the ego lanes and curvature of the ego lanes.
13. The method as claimed in claim 11 , wherein the ego lanes comprises a left ego lane and a right ego lane.
14. The method as claimed in claim 11 further comprises converting the binary image from a two-dimensional form to a one-dimensional form, when determining the coefficient values of the ego lanes of the host vehicle.
15. The method as claimed in claim 11 further comprises generating warning signals to the host vehicle based on the coefficient values of the ego lanes of the host vehicle and the lane pattern and lane colour of each of the one or more lane markings.
16. The method as claimed in claim 11 further comprises performing upscaling and downscaling operations on the input image, during segmentation, using the trained semantic segmentation model.
17. The method as claimed in claim 11 , wherein training the CNN model comprises determining an error in determined coefficient values of ego lanes using predefined loss minimization techniques for optimizing the determination of coefficient values in subsequent computations, wherein the training is based on trial RGB images.
18. The method as claimed in claim 11 further comprises overlaying graphical lanes on the input image based on the coefficient values of the ego lanes and the lane pattern of the ego lanes, to indicate position of the ego lanes in the input image.