jason_rhl的博客

# Author: Mohamed Aly <malaa@caltech.edu> # Date: 10/7/2010 ============================================================================ REAL TIME LANE DETECTOR SOFTWARE ============================================================================ This package contains source code and dataset that implements the work in the paper [1]. ========= Contents ========= src/: contains the C/C++ source files |_ CameraInfo.conf: contains the camera calibration ifo |_ CameraInfoOpt.*: contain gengetopt files for parsing the camera info files |_ cmdline.*: contains gengetopt files for parsing command lines |_ InversePerspectiveMapping.*: code for obtainig the IPM of an image |_ LaneDetector.*: code for the bulk of the algorithm, including Hough Transforms, Spline fitting, Post processing, ... |_ Lanes.conf: the typical configuration file for lane detection |_ main.*: code for the main binary |_ Makefile: the Make file |_ mcv.*: contain utility functions |_ ranker.h: code for obtaining the median of a vector |_ run.sh: Shell script for running the detector on the four clips in Caltech Lanes Dataset matlab/: contains the Matlab source files |_ ccvCheckMergeSplines.m: checks if two splines are matching |_ ccvEvalBezSpline.m: returns points on a spline given its control points |_ ccvGetLaneDetectionStats.m: computes stats from detections and ground truth |_ ccvLabel.m: handles the ground truth labels |_ ccvReadLaneDetectionResultsFile.m: reads a detection file output from the binary file LaneDetector32/64 |_ Stats.m: computes stats for the detections on the Caltech Lanes Dataset and its ground truth labels ============== Prerequisites ============== 1. OpenCV 2.0 or higher http://sourceforge.net/projects/opencvlibrary/ 3. (Optional) Gengetopt http://www.gnu.org/software/gengetopt/ =========== Compiling =========== Unzip the archive somewhere, let's say ~/lane-detector: unzip lane-detector.zip -d ~/lane-detector cd ~/lane-detector/src make release This will generate LaneDetector32 or LaneDetector64 depending on your system. ====================== Caltech Lanes Dataset ====================== To view the lane detector in action, you can download the Caltech Lanes Dataset available at http://www.vision.caltech.edu/malaa/datasets/caltech-lanes =========== Running =========== To run the detector on the Caltech Lanes dataset, which might be in ~/caltech-lanes/ cd ~/lane-detector/ ln -s ~/caltech-lanes/ clips cd ~/lane-detector/src/ bash run.sh This will create the results files inside ~/caltech-lanes/*/list.txt_results.txt To view the statistics of the results, open Matlab and run the file: cd ~/lane-detector/matlab/ matlab& >>Stats ====================== Command line options ====================== LinePerceptor 1.0 Detects lanes in street images. Usage: LinePerceptor [OPTIONS]... [FILES]... -h, --help Print help and exit -V, --version Print version and exit Basic options: --lanes-conf=STRING Configuration file for lane detection (default=`Lanes.conf') --stoplines-conf=STRING Configuration file for stopline detection (default=`StopLines.conf') --no-stoplines Don't detect stop lines (default=on) --no-lanes Don't detect lanes (default=off) --camera-conf=STRING Configuration file for the camera paramters (default=`CameraInfo.conf') --list-file=STRING Text file containing a list of images one per line --list-path=STRING Path where the image files are located, this is just appended at the front of each line in --list-file &#40;default=`'&#41; --image-file=STRING The path to an image Debugging options: --wait=INT Number of milliseconds to show the detected lanes. Put 0 for infinite i.e. waits for keypress. (default=`0') --show Show the detected lines (default=off) --step Step through each image (needs a keypress) or fall through (waits for --wait msecs) (default=off) --show-lane-numbers Show the lane numbers on the output image (default=off) --output-suffix=STRING Suffix of images and results (default=`_results') --save-images Export all images with detected lanes to the by appending --output-suffix + '.png' to each input image (default=off) --save-lanes Export all detected lanes to a text file by appending --output-suffix + '.txt' to --list-file &#40;default=off&#41; --debug Show debugging information and images (default=off) =========== References =========== [1] Mohamed Aly, Real time Detection of Lane Markers in Urban Streets, IEEE Intelligent Vehicles Symposium, Eindhoven, The Netherlands, June 2008.