Fig 1.
Global description of extract real traffic information.
Fig 2.
a) The zx plane b) The xy plane in the W space, namely the S surface.
Fig 3.
Differences between IPM, Homography and MIPM a) Original image b) IPM method c) MIPM method d) Homography method.
Fig 4.
An example of video sequences with background subtraction result (Snowy weather with strong wind).
Fig 5.
Different frames in different locations used for testing purposes (Recorded in Madrid and Tehran) (a- k) a, b, c, d) Explanation in the text, e) Rainy weather with poor lighting, wet ground, high traffic and occlusion. urban location with camera placed in the center at 45 degrees, (f) City intersection, Camera is located on right side of the street at a 45 degrees angle, in day traffic, (g) Snowy weather, urban area with low traffic, Camera angle is 45 degrees, center of an uphill street, (h) Normal weather condition, highway with low Traffic, camera angle is 30 degrees, right side of the highway Urban area, (i) Sunny weather condition, highway with high Traffic, camera angle is 45 degrees and Center of the highway, (j) Sunny weather with low traffic, camera angle is 45 degrees. Center of high way, interurban area, (k) Sunny weather with low traffic, camera angle is 45 degrees. Center of highway.
Fig 6.
Example of sequences in KITTI Data set (a-b).
Fig 7.
Example of sequences in DETRAC dataset (a-e).
Fig 8.
a) Original image with perspective effect b) Remapped image that removed perspective with IPM c) Remapped image that removed perspective with MIPM d) Remapped image with Homography method.
Fig 9.
(a-g) a) Original image b) Removed perspective with MIPM c) Detected lines d) lane1 e) lane2 f) lane3 g) Original image.
Fig 10.
Detection of vehicle in lane 1 of the high way a) Remapped by MIPM b) Detected vehicle in lane 1 by GMM.
Fig 11.
Removal of shadows with the chromacity-based method in HSI color space a) Results of dot product between original mages and binary images b) Removal of shadows in the original images c) Shadows are detected d) Results of the difference between the shadow images and the binary ones.
Table 1.
Comparison of shadow detection methods.
Fig 12.
a) Binary images b) Geometric centers (red points) c) Aligning the geometric centers (white points) d) Path of vehicle (path tracking by frame-by-frame evaluation of blobs centers).
Fig 13.
(a) Car area variation along with its direction using the two methods: IPM (red) and MIPM (blue) (b) Comparison of actual vehicle’s area and obtained area from image considering position and direction of the vehicle. Results indicate that areas measured using the presented MIPM method are closer to the actual ones in the 90% of the tested cars (compared to real areas with R2>0.98).
Table 2.
Car area variation along with its direction using the 3 methods, IPM, Homography and MIPM.
Fig 14.
Some sequences in different weather conditions and locations and result of our detection and tracking using our MIPM method (a-f).
(a) Urban area, snowy weather with strong wind, frame n 56 (our dataset), b) Urban area, Snowy weather with strong wind, frame n 69 (our dataset), (c) Sunny weather with high traffic at high way frame n 411 (our dataset), (d) Urban area, high traffic, poor lighting in rainy weather, frame n 802 (our dataset), (e) High way, light traffic, poor lighting in normal weather, frame n 1494 (DETRAC dataset), (f) Intersection, sunny day, day traffic, Frame n 61 (KITTI dataset).
Table 3.
Comparison of detection rate using 3 Methods on KITTI and DETRAC datasets.
Table 4.
Analysis of detection rate of our method (MIPM), with 3 different data sets.
Table 5.
Selected sequences are in urban area with sunny weather.
Table 6.
Poor lighting conditions in highway.
Table 7.
In urban area with snowy weather.
Table 8.
Rainy weather in highway.
Table 9.
High traffic in an interurban (with occlusion).