Problem $5\mathrm{.}2$: Naive Matching $[8$ points$]$ this, namely, for each corner in image$1,$ find the best match from the detected corners in image$2($or$,$ if the NCC match score is too low, then return no match for that point$).$

Write a function and call it as below. Examine your results for $20$ detected corners in each image.

Note: Your matches may not be perfect.

$```$

def naive$\_$matching$($img$1,$ img$2,$ corners$1,$ corners$2,$ R$,$ NCCth$)$:

$"""$

Compute NCC given two windows.

Args:

img$1$: Image $1.$

img$2$: Image $2.$

corners$1$: Corners in image $1($nx$2)$

corners$2$: Corners in image $2($nx$2)$

R: NCC matching radius

NCCth: NCC matching score threshold

Returns:

matching: NCC matching returns a list of tuple $($c$1,$ c$2),$

c$1$ is the $1$x$2$ corner location in image $1,$

c$2$ is the $1$x$2$ corner location in image $2.$

$"""$

### YOUR CODE HERE

raise NotImplementedError$("$Delete this line once you implement the function"$)$

### END YOUR CODE

return matching

$``````$

# Detect corners on warrior and matrix sets

# You are free to modify the parameters

n$\_$corners $=20$

smooth$\_$std $=1$

window$\_$size $=17$

# Read images and detect corners on the images

imgs$\_$mat $=[]$

crns$\_$mat $=[]$

imgs$\_$war $=[]$

crns$\_$war $=[]$

for i in range$(2)$:

img$\_$mat $=$ io$.$imread$(\text{'}./$p$6/$matrix$/$matrix$\text{'}+$ str$($i$)+\text{'}.$png$\text{'})$

imgs$\_$mat.append$($rgb$2$gray$($img$\_$mat$))$

img$\_$war $=$ io$.$imread$(\text{'}./$p$6/$warrior$/$warrior$\text{'}+$ str$($i$)+\text{'}.$png$\text{'})$

imgs$\_$war.append$($rgb$2$gray$($img$\_$war$))$

# Match corners

crnsmatf$=$open$(\text{'}./$p$6/$matrix$/$crns$\_$mat.pkl$\text{'},\text{'}$rb$\text{'})$

crns$\_$mat$=$pickle.load$($crnsmatf$)$

crnswarf$=$open$(\text{'}./$p$6/$warrior$/$crns$\_$war.pkl$\text{'},\text{'}$rb$\text{'})$

crns$\_$war$=$pickle.load$($crnswarf$)$

R $=31$

NCCth $=2400$

matching$\_$mat $=$ naive$\_$matching$($imgs$\_$mat$[0]/255,$ imgs$\_$mat$[1]/255,$ crns$\_$mat$[0],$ crns$\_$mat$[1],$ R$,$ NCCth$)$

matching$\_$war $=$ naive$\_$matching$($imgs$\_$war$[0]/255,$ imgs$\_$war$[1]/255,$ crns$\_$war$[0],$ crns$\_$war$[1],$ R$,$ NCCth$)$

$]$:

# Plot matching result

def show$\_$matching$\_$result$($img$1,$ img$2,$ matching$)$:

fig $=$ plt$.$figure$($figsize$=(8,8))$

plt$.$imshow$($np$.$hstack$(($img$1,$ img$2)),$ cmap$=$'gray'$)$ # two dino images are of different sizes, resize one before use

for p$1,$ p$2$ in matching:

plt$.$scatter$($p$1[0],$ p$1[1],$ s$=35,$ edgecolors$=\text{'}$r$\text{'},$ facecolors$=$'none'$)$

plt$.$scatter$($p$2[0]+$ img$1.$shape$[1],$ p$2[1],$ s$=35,$ edgecolors$=\text{'}$r$\text{'},$ facecolors$=$'none'$)$

plt$.$plot$([$p$1[0],$ p$2[0]+$ img$1.$shape$[1]],[$p$1[1],$ p$2[1]])$

#plt$.$savefig$(\text{'}$matching$.$png$\text{'})$

plt$.$show$()$

print$("$Number of Corners:", n$\_$corners$)$

show$\_$matching$\_$result$($imgs$\_$mat$[0],$ imgs$\_$mat$[1],$ matching$\_$mat$)$

show$\_$matching$\_$result$($imgs$\_$war$[0],$ imgs$\_$war$[1],$ matching$\_$war$)$

$```$