You are on page 1of 5

ROAD DETECTION USING MASK AND EDGE DETECTION

INTRODUCTION Advanced driver-assistance systems aim to understand the environment of the vehicle contributing to trafc safety. In this context different types of cameras play a relevant role. Vision-based approaches are used to address functionalities such as lane marking detection, trafc sign recognition, pedestrian detection, etc. For structured road detection, there are several methods proposed for detecting the road lines using colour characterizations. Generally, for avoiding noises from whole image frame, some certain regions are always predefined for detection. The main idea of this method was to categorize the road image into two parts, road and non-road, based on the theory of machine learning using the texture and coordination as the features. In the road detection employs road model. In these methods, if the edge of the road was not obvious, the detection result could be influenced. Young detected the road lines by colour. They predefined the detected regions to avoid the influences from noises. The partial road lines were obtained from road face. And then, the road lane was detected and confirmed by angle length and width of partial road lines. STEPS INVOLVED IN DETECTING A ROAD LANE: LANE SEPARATION FROM BACKGROUND

Image Mask

Image Mask

In detection of lane in the road we use an image which is represented by Fig. This image is applied on the original image in a way such that if the pixel in the image is white it keeps the original pixel value of the road, otherwise it will make the particular pixel value of the road black. LINE POINTS DETECTION ON STRUCTURED ROAD In structured road detection, white and yellow lines on the road are the main targets to detect. Generally, the colour of road surface is charcoal gray or black. There are obvious colour differences between the road surface and white/yellow road lines. We employ vertical Sobel mask to obtain edge values in RGB components denoted in eqn. And we process Sobel operation every fixed 5 pixels vertical intervals in the middle trapezoid regions. Wxj ,yj=Srxj ,yj+Sgxj ,yj+Sbxj ,yj Where Srxj ,yj,Sgxj ,yj,Sbxj ,yj are values by vertical Sobel operation on the jth pixel. We keep the point which Wxj ,yj is greater than a predefined Threshold T as follows. Lxj ,yj=1,Wxj ,yj>T0,Otherwise After applying Sobel edge detection technique we use Fig. This image is used to eliminate the edges which are formed due to the application of the first mask Fig.The elimination is carried out by reducing the border edges of the of the new mask thus making it have a smaller area as compared to the original one.

Candida e Slope Selection Noise point Elimination

SLOPE FILTERING

After the detection of line points, we must eliminate the noise points. We firstly obtain two points Rc and Lc at bottoms of right and left middle trapezoid regions, respectively. Then, we find the nearest points to Rc and Lc and calculate the slope between these points, respectively. If slopes are in the range of table I, we will set the slopes in right and left regions as candidate slopes Rs and Ls. Otherwise we will remove the point and find the next nearest points. Fig shows the candidate slopes selection result. Once we have obtained that point we again recalculate the slope from scanned points and the point selected and compare it with table I as shown in Fig. And the left hand side is processed the same operation till all the points are done. This method can eliminate the noise points which are not the road lines in driving direction. RANGES OF SLOPES
The range of slopes Left -0.7 ~ -2.1445 Right 1.7 ~ 3.1445

ROAD STYLE DECISION The number of road line points is the determinant. If the number is less than a certain threshold value, it means that the road line points perhaps could not exist. The detected line points could be noises and the style of road could be unstructured.

ALGORITHM FOR CBIR USING KEKRE TRANSFORM METHOD STEP 1: STEP 2: STEP 3: STEP 4: STEP 5: STEP 6: STEP 7: STEP 8: READ THE IMAGE APPLY MASK SHOWN IN FIG ON IMAGE FIND THE EDGES BY USING SOBEL EDGE DETECTOR APPLY ASK SHOWN IN FIG TO ON IMAGE SELECT A RANDOM POINT AND CALCULATE THE SLOPE BETWEEN AN INITIALLY SELECTED POINT AND NEWLY SELECTED RANDOM POINT IF SLOPE IS WITHIN THE THRESHOLD GIVEN IN TABLE 4.1 MASK THE RANDOM POINT THE SELECTED POINT OR ELSE SELECT ANOTHER RANDOM POINT AND CALCULATE THE SLOPE PLOT A LINE USING ALL THE SELECTED POINTS DISPLAY FINAL IMAGE RESULTS OF PROGRAM FOR ROAD DETECTION

after applying on the figure

Road Mask

Road after performing Sobel Edge Detection

Road after applying Mask in the above figure

Detected

Final Road

You might also like