Autonomous Guidance System for Weeding Robot in Wet Rice Paddy

Transcription

1 Closing Ceremony 214 Autonomous Guidance System for Weeding Robot in Wet Rice Paddy Choi, Keun Ha (14R5558) Department of Mechanical Engineering Korea Advanced Institute of Science and Technology Advisor In TITECH : Associate Professor Edwardo F. FUKUSHIMA/ FUKUSHIMA Roboti Lab.

2 I. Introduction Need to the robot in agriculture Aging in agriculture population Decrease the productivity and shortage of manpower Weeding Robots Need to the robot, Problem of the hand weeder, riding weeder Mechanization rate for weeding control 6.6% Environment-friendly organic agricultural Autonomous [Hand weeder] [Riding weeder] [Weeding robot] <2>

3 II. About Research Autonomous Guidance System based-on IR Vision Sensor for Weeding Robot in Wet Rice Paddy - Robust to illumination condition - Improve the accuracy of rice row and guidance line + Low cost and Simple system Research progress Crop row detection Process Crop-soil(water) Discrimination Crop row detection Guidance line Motion Control Research results in KAIST Research results in TITECH [Single image / Static thresholding ] [ SDI ] <3>

4 II. About Research System Architecture Improving the algorithm Research results in TITECH Extract the segmented crop line Make the extended virtual line NIR Image acquisition Sequential difference Image (SDI) Clustering rice pixels (K-means clustering) Extract guidance lines Gray images Image subtract Median filter Thresholdering (OTSU method) Crop segmentation Intersection distribution from lines Line fitting (Robust regression) No Evaluate guidance lines Yes Robot motion control Robot guidance information <4>

5 III. Concept of Algorithm Concept : Rice Morphology Characteristic Convergence of rice stem : leaf originate from a central stem symmetrically Find the algorithm which is represented the convergence of central rice stem Extended virtual line Segmented crop line Central stem Step 1. Extract the segmented crop line from crop image Step 2. Extend each segmented crop lines Make the extended virtual lines Step 3. Compute the intersection from lines Intersection distribution Intersection distribution Step 4. Extract the guidance lines from intersection distribution Step 5. Evaluate the guidance lines Convergence line = Segmented crop line + Extended virtual line <5>

6 IV. Process of Algorithm Step 1. Extract the segmented crop line from crop row image Preprocess of crop row image SDI image (binary image) Remove small blob noise Skeletonization (Zhang-Suen thining) p(, i j) pixel value is 1 (white pixel). P p all pixels of P area ( mn, ) area P, If P(, i j) Bolb noise elimination P 1, otherwise area (, i j) area k, where P number of pixels in the region which is connected each p( i, j). Clustering using K-means algorithm X x p( i, j) 1 n C c c 1,..., k where k = centroid locations of initial cluster k = the number of local peaks on sum of white pixels reference to y X x d( x, c ) d( x, c ), j 1,..., k i i n n i n j C c( X ), i = 1,..., k i Image coordinate (y) Image coordinate (x) 1 8 D N n1 d( x, c ) i( n) Dpre Dcur D 1 D n pre 4 Sum of wihte pixels Image coordinate (x) <6>

7 IV. Process of Algorithm Segmented crop line Hough transform(paul Hough, 1962) - The classical Hough transform was concerned with the identification of lines in the image r a) Original image, (b) Skeletonized image, (c) Segmented lines (start point: circle, end point: cross, length: green line), (d) Extended virtual lines (start point: rectangle, end point: cross, length: blue line) <7>

8 IV. Process of Algorithm Step 2 : Extend each segmented crop lines Extended virtual line, Segmented crop line,, Extended virtual line, le l where r lh, if 9, /cos( ) l, else s l e, l = length of segmented line r s pts pce, pte pce ( le cos( ), le sin( )) if y yce and / 2 pts pce, pte pce ( le cos( ), le sin( )) if y yce and / 2< pts p, pte p ( le cos( ), le sin( )) if y yce and / 2 Lm ( ) = pts p, pte p ( le cos( ), le sin( )) if y yce and / 2< pts pce, pte pce ( lh,) if y yce and / 2 pts p, pte p ( lh,) if y yce and / 2 [Segmented line] [Extended virtual line] <8>

9 IV. Process of Algorithm Step 3 : Compute the intersection from lines, Pixels = Moneybox Pixel which pass through the extended Line Put money( 1 ) in a moneybox Intersection point = {money > 1} Step 4 : Extract the guidance lines Intersection distribution Robust regression (James O., 1988) Least squares estimates for regression models are highly sensitive to (not robust against) outliers. t N t yi xi xw[( ˆ)/ ˆ] i 1 i yi x i i Cluster 1 Cluster 2 Cluster ( r ) e i for r wi where for r > t r ( y x ˆ )/ ˆ i i i i i t ˆ Med y x ˆ /.6745 i n/max( n) i i i <9>

10 V. Experiments Results [Intersection point] Cluster 1 Cluster 2 Cluster 3 [Guidance line] (red: LSE, green: robust regression) <1>

11 VI. Campus life in TITECH Friends Advisor Healing! Travel <11>