Technical seminar

SREE KAVITHA ENGINEERING COLLEGE

UNDER THE SUPERVISION OF:

Mrs.D.SAILAJA M.Tech,(ph.D).;

Associate Professor.

Karepalli – 507122,Khammam Dt, TS

DEPT OFELECTRONICS AND COMMUNICATION

ENGINEERING

A LANE BOUNDARY DETECTION METHOD BASED ON HIGH DYNAMIC

RANGE IMAGE PRESENTED BY:K.LAVANYA(11C81A0457)

CONTENTS

• ABSTRACT

• INTRODUCTION

• HDR CAMERA

• LANE DETECTION OF HDR IMAGE

• MERGING OF IMAGES

• DIFFERENTIAL EXPOSED IMAGES

• MERG HIGH DYNAMIC RANGE IMAGE

• IMPROVED METHOD BASED ON HDR

• RESULTS

• ADVANTAGES AND DISADVANTAGES

• CONCLUSION AND FUTURE SCOPE

ABSTRACT

Every year many vehicle departure accidents

happen due to the driver's carelessness.

Lane Departure Warning System (LDWS) is a kind of

system which can relieve the stress of the drivers and reduce

traffic accidents.

But most traffic scences have greater dynamic range than

digital camera at present.

ADAS are systems which can help the drivers in their driving

process. They are designed with a safe Human-Machine Interface and it

should increase car safety and more generally road safety.

Nowadays the most common ADAS are navigation system, adaptive

cruise control (ACC) system, lane departure warning system (LDWS), lane

change assistance system, collision avoidance system, night vision, traffic

sign recognition and so on.

Dynamic range is the ratio between the largest light intensity and the

smallest possible light intensity values in a scene. The dynamic range in most

scenes in the natural environment is larger than 10^5

INTRODUCTION

DIFFICULT SCENES FOR LDWS

Affected by adverse light conditions such as reflective road

surfaces, road with the shadows of the trees

HDR CAMERA

LANE DETECTION ON HDR

IMAGE

TAKE DIFFERENTLY EXPOSED IMAGES

Tocci’s method is the most suitable for the real timesystem like the lane departure warning system.

Images taken by a camera continuous with changingexposure time. In the first row, the images exposure time are1/250 second, 1/1000 second, 1/60 second. In the secondrow, the images exposure time are 1/60 second,1/250second,1/15 second.

We can find out a lot of pixels are white in the properly-exposure image., and in the under-exposure image most pixels are almost black while in the over-exposure image

most pixels are almost white.

DIFFERENTLY EXPOSED IMAGES

DIFFERENTILY EXPOSED IMAGES

HISTOGRAM OF THE PROPERLY-EXPOSURE IMAGE HISTOGRAM OF UNDER-

EXPOSURE IMAGE

HISTOGRAM OF THE OVER-EXPOSURE

IMAGE

HISTOGRAM REPRESANTATION

LANE DETECTION OF HDR

IMAGE

RECOVER THE RESPONSE FUNCTION OF THE CAMERA:

We use the method Debevec proposed to calculate theresponse function .Let us call the response function f. It canbe assumed that Zij is the gray scale value of the pixel whichis numbered i and exposure time is Δtj. Ei is the realluminance of the point in the scene that pixel i take. Theymeet the equation as:

MERG OF HIGH DYNAMIC RANGE

IMAGES

We use the differently exposed images and the

camera response function to merge a 32 bit

image and then tone mapping it to a 8 bit

image.

CAMERA RESPONSE

AUTO EXPOSED IMAGE AND HDR IMAGE OF THE SAME SCENE

HISTOGRAM OF THE HDR

IMAGE

EDGE DETECTION

Edge detection of the image is to detect the

edge of the surrounding pixels have a gray-scale

step-like changes or changes in the roof of a

collection of pixels. In the images taken for lane

detection, the lane in the image is different from

the road. So we can detect the lane use EDGE

DETECTION ALGORITHM.

We use Sobel operator to detect the lane

boundary. The Sobel operator use the weighted

difference of the gray scale value of the field

point of each point in the image.

Image binaryzation

Before detect the line, we should do the image thresholdsegmentation first. Image binaryzation is the procedure ofseparating the target and the background.

The algorithm assumes that the image to be thresholded

contains two classes of pixels or bi-modal histogram,

then calculates the optimum threshold separating thosetwo classes so that their combined spread.

Hough transform

The final step of the detection is to detect the boundary of

the lane. Most of the lane are straight line or clothoid curve,

we can use hough transform to detect the line.

Carried out in voting procedure.

This voting procedure is carried out in the parameter space,

in the parameter space the object candidates are obtained as

local maxima in an accumulator space that is constructed by

the algorithm for computing the Hough transform.

IMPROVED METHOD BASED ON HDR IMAGE

EXPOSURE FUSION

Exposure fusion is an alternative way to merge high dynamic range

images. Compared with the traditional HDR merging method, it

leaves out many computation procedures. By fusion the different

exposed images, the image has greater dynamic range than one

image. But without scientific calculations, the image merged by

exposure fusion is not as good as the image merged in the

traditional HDR merging method.

The improved method

As we all know, color image has red, green, blue channels, in the high

dynamic range merging procedure, we should calculate three times to merge a

HDR image. And the time of calculate one channel is very long. But the

computation on binary images is very quick.

So we propose a improve method.

The lane detection procedure of the method based on HDRimage is as

shown in and the lane detection procedure of the improved method

Exposure fusion in the improved method

The lane detection system is a real-time system. We should consider the

computation time first.

If we choose a complex exposure fusion method, it will not use in a real

lane departure warning system.

PROCEDURE OF THE METHOD BASED ON HDR IMAGE

FLOW CHART REPRESENTATION

PROCEDURE OF THE IMPROVED

METHOD

IMPROVED METHOD

Lane Departure Warning System (LDWS) is a kind of systemwhich can relieve the stress of the drivers and reduce trafficaccidents.

The high dynamic range image can improve the accuracy of the lane detection method.

We proposed an improved method based on exposure fusion to reduce the computational time of the system.

ADVANTAGES

DISADVANTAGES

Processing of merging HDR image is very time consuming. Itmakes HDR image can't be used in real-time LDWS.

We proposed an improved method based on exposure fusionto reduce the computational time of the system.

APPLICATIONS

1)LDWS

2)Lane keeping assistance

3)Blind spot recognition

4)Traffic sinal recognition

5) Pedistrain recognition

COMPUTATION TIME

After a series of experiments, we get the

computation time use each method. The average

computation time on a single image is about 40ms, the

average computation on HDR image is about 2 seconds, and

the computation time use the improved method is about

100ms. We can get the method based on high dynamic

range images cost too much time, it can’t be used in the

system. But the improved method based on exposure fusion

costs about 65ms, it can be used in the real-time system.

EXPERIMENTAL RESULTS

DETECTION RESULTS

Edge detection result of the auto exposed imagesand high dynamic range image

CONCLUSION

We use three images with different exposure to merge a high dynamic

range image and detect the lane in the HDR image.

The experimental results show that the high dynamic range image can

improve the accuracy of the lane detection method. However, the processing of

merging HDR image is very time consuming. It makes HDR image can't be

used in real-time LDWS.

We proposed an improved method based on exposure fusion to reduce the

computational time of the system.

FUTURE WORK

The average computation time of with the improved method based

on exposure fusion is about 100ms. It meets the need of the lane

departure warning system.

In order to detect the lane in the image, future work on develop a

high dynamic range camera system should be excavated and utilized.

By incorporating a departure warning system, the functionalities of

the lane marking system can be enhanced.

Hough transform can be implemented on FPGA board. FPGA

implementation consumes less power also it is very compact and fast.