I was trying to solve the cs188 ai project-1 search. In cornersHeuristic() function we need to return a heuristic, so that we use it when traversing the corners. And we want to use the shortest path so we use A-star search.
Summary of the code to understand it better: We have the locations of the corners in (x,y) format, and we have our characters current location again as (x, y) format.
The heuristic returned is maximum distance of the distances between the location to all the corners. But what is the logic here? Shouldn't we return the minimum distance to the nearest corner? Why do we return the max out of these, how does this help to find the shortest path?
The code is as below:
Some explanations for the code: The corner1, corner2, corner3, corner4, and location variables are all tuples and store x and y values. Such as (1, 1), (4, 5)
def cornersHeuristic(state, problem):
A heuristic for the CornersProblem that you defined.
state: The current search state
(a data structure you chose in your search problem)
problem: The CornersProblem instance for this layout.
This function should always return a number that is a lower bound on the
shortest path from the state to a goal of the problem; i.e. it should be
admissible (as well as consistent).
corners = problem.corners # These are the corner coordinates
walls = problem.walls # These are the walls of the maze, as a Grid (game.py)
location, corner1, corner2, corner3, corner4 = state
corners = [corner1, corner2, corner3, corner4]
heuristic = 0
for x in corners[1]:
heuristic = max(heuristic,(abs(location[0]-x[0]) + abs(location[1] - x[1])))
return max_r
The heuristic is a lower-bound on the actual distance. You want this lower-bound to be as large as possible, while still being guaranteed to be a lower-bound.
So you're correct that min would work. But max is better, because it's a larger lower-bound.
We can guarantee it will still be a lower-bound because the problem statement is that we want to visit all of the corners. The path-length required to do that will always be at least the cost of visiting the furthest corner. If the problem were to visit any one corner, then we'd need to use min
Related
I'm trying to find a match between a set of 2D boxes with coordinates (A) (from a template with known sizes and distances between boxes) to another set of 2D boxes with coordinates (B) (which may contain more boxes than A). They should match in terms of each box from A corresponds to a single Box in B. The boxes in A together form a "stamp" which is assymmetrical in atleast one dimension.
Illustration of problem
explanation: "Stanz" in the illustration is a box from set A.
One might even think of the Set A as only 2D points (the centerpoint of the box) to make it simpler.
The end result will be to know which A box corresponds to which B box.
I can only think of very specific ways of doing this, tailored to a specific layout of boxes, is there any known generic ways of dealing with this forms of matching/search problems and what are they called?
Edit: Possible solution
I have come up with one possible solution, looking for all the possible rotations at each possible B center position for a single box from set A. Here all of the points in A would be rotated and compared against the distance to B centers. Not sure if this is a good way.
Looking for the possible rotations at each B centerpoint- solution
In your example, the transformation between the template and its presence in B can be entirely defined (actually, over-defined) by two matching points.
So here's a simple approach which is kind of performant. First, put all the points in B into a kD-tree. Now, pick a canonical "first" point in A, and hypothesize matching it to each of the points in B. To check whether it matches a particular point in B, pick a canonical "second" point in A and measure its distance to the "first" point. Then, use a standard kD proximity-bounding query to find all the points in B which are roughly that distance from your hypothesized matched "first" point in B. For each of those, determine the transformation between A and B, and for each of the other points in A, determine whether there's a point in A at roughly the right place (again, using the kD-tree), early-outing with the first unmatched point.
The worst-case performance there can get quite bad with pathological cases (O(n^3 log n), I think) but in general I would expect roughly O(n log n) for well-behaved data with a low threshold. Note that the thresholding is a bit rough-and-ready, and the results can depend on your choice of "first" and "second" points.
This is more of an idea than an answer, but it's too long for a comment. I asked some additional questions in a comment above, but the answers may not be particular relevant, so I'll go ahead and offer some thoughts in the meantime.
As you may know, point matching is its own problem domain, and if you search for 'point matching algorithm', you'll find various articles, papers, and other resources. It seems though that an ad hoc solution might be appropriate here (one that's simpler than more generic algorithms that are available).
I'll assume that the input point set can only be rotated, and not also flipped. If this idea were to work though, it should also work with flipping - you'd just have to run the algorithm separately for each flipped configuration.
In your example image, you've matched a point from set A with a point from set B so that they're coincident. Call this shared point the 'anchor' point. You'd need to do this for every combination of a point from set A and a point from set B until you found a match or exhausted the possibilities. The problem then is to determine if a match can be made given one of these matched point pairs.
It seems that for a given anchor point, a necessary but not sufficient condition for a match is that a point from set A and a point from set B can be found that are approximately the same distance from the anchor point. (What 'approximately' means would depend on the input, and would need to be tuned appropriately given that you're using integers.) This condition is met in your example image in that the center point of each point set is (approximately) the same distance from the anchor point. (Note that there could be multiple pairs of points that meet this condition, in which case you'd have to examine each such pair in turn.)
Once you have such a pair - the center points in your example - you can use some simple trigonometry and linear algebra to rotate set A so that the points in the pair coincide, after which the two point sets are locked together at two points and not just one. In your image that would involve rotating set A about 135 degrees clockwise. Then you check to see if every point in set B has a point in set A with which it's coincident, to within some threshold. If so, you have a match.
In your example, this fails of course, because the rotation is not actually a match. Eventually though, if there's a match, you'll find the anchor point pair for which the test succeeds.
I realize this would be easier to explain with some diagrams, but I'm afraid this written explanation will have to suffice for the moment. I'm not positive this would work - it's just an idea. And maybe a more generic algorithm would be preferable. But, if this did work, it might have the advantage of being fairly straightforward to implement.
[Edit: Perhaps I should add that this is similar to your solution, except for the additional step to allow for only testing a subset of the possible rotations.]
[Edit: I think a further refinement may be possible here. If, after choosing an anchor point, matching is possible via rotation, it should be the case that for every point p in B there's a point in A that's (approximately) the same distance from the anchor point as p is. Again, it's a necessary but not sufficient condition, but it allows you to quickly eliminate cases where a match isn't possible via rotation.]
Below follows a finished solution in python without kD-tree and without early outing candidates. A better way is to do the implementation yourself according to Sneftel but if you need anything quick and with a plot this might be useful.
Plot shows the different steps, starts off with just the template as a collection of connected lines. Then it is translated to a point in B where the distances between A and B points fits the best. Finally it is rotated.
In this example it was important to also match up which of the template positions was matched to which boundingbox position, so its an extra step in the end. There might be some deviations in the code compared to the outline above.
import numpy as np
import random
import math
import matplotlib.pyplot as plt
def to_polar(pos_array):
x = pos_array[:, 0]
y = pos_array[:, 1]
length = np.sqrt(x ** 2 + y ** 2)
t = np.arctan2(y, x)
zip_list = list(zip(length, t))
array_polar = np.array(zip_list)
return array_polar
def to_cartesian(pos):
# first element radius
# second is angle(theta)
# Converting polar to cartesian coordinates
radius = pos[0]
theta = pos[1]
x = radius * math.cos(theta)
y = radius * math.sin(theta)
return x,y
def calculate_distance_points(p1,p2):
return np.sqrt((p1[0]-p2[0])**2+(p1[1]-p2[1])**2)
def find_closest_point_inx(point, neighbour_set):
shortest_dist = None
closest_index = -1
# Find the point in the secondary array that is the closest
for index,curr_neighbour in enumerate(neighbour_set):
distance = calculate_distance_points(point, curr_neighbour)
if shortest_dist is None or distance < shortest_dist:
shortest_dist = distance
closest_index = index
return closest_index
# Find the sum of distances between each point in primary to the closest one in secondary
def calculate_agg_distance_arrs(primary,secondary):
total_distance = 0
for point in primary:
closest_inx = find_closest_point_inx(point, secondary)
dist = calculate_distance_points(point, secondary[closest_inx])
total_distance += dist
return total_distance
# returns a set of <primary_index,neighbour_index>
def pair_neighbours_by_distance(primary_set, neighbour_set, distance_limit):
pairs = {}
for num, point in enumerate(primary_set):
closest_inx = find_closest_point_inx(point, neighbour_set)
if calculate_distance_points(neighbour_set[closest_inx], point) > distance_limit:
closest_inx = None
pairs[num]=closest_inx
return pairs
def rotate_array(array, angle,rot_origin=None):
if rot_origin is not None:
array = np.subtract(array,rot_origin)
# clockwise rotation
theta = np.radians(angle)
c, s = np.cos(theta), np.sin(theta)
R = np.array(((c, -s), (s, c)))
rotated = np.matmul(array, R)
if rot_origin is not None:
rotated = np.add(rotated,rot_origin)
return rotated
# Finds out a point in B_set and a rotation where the points in SetA have the best alignment towards SetB.
def find_stamp_rotation(A_set, B_set):
# Step 1
anchor_point_A = A_set[0]
# Step 2. Convert all points to polar coordinates with anchor as origin
A_anchor_origin = A_set - anchor_point_A
anchor_A_polar = to_polar(A_anchor_origin)
print(anchor_A_polar)
# Step 3 for each point in B
score_tuples = []
for num_anchor, B_anchor_point_try in enumerate(B_set):
# Step 3.1
B_origin_rel_point = B_set-B_anchor_point_try
B_polar_rp_origin = to_polar(B_origin_rel_point)
# Step 3.3 select arbitrary point q from Ap
point_Aq = anchor_A_polar[1]
# Step 3.4 test each rotation, where pointAq is rotated to each B-point (except the B anchor point)
for try_rot_point_B in [B_rot_point for num_rot, B_rot_point in enumerate(B_polar_rp_origin) if num_rot != num_anchor]:
# positive rotation is clockwise
# Step 4.1 Rotate Ap by the angle between q and n
angle_to_try = try_rot_point_B[1]-point_Aq[1]
rot_try_arr = np.copy(anchor_A_polar)
rot_try_arr[:,1]+=angle_to_try
cart_rot_try_arr = [to_cartesian(e) for e in rot_try_arr]
cart_B_rp_origin = [to_cartesian(e) for e in B_polar_rp_origin]
distance_score = calculate_agg_distance_arrs(cart_rot_try_arr, cart_B_rp_origin)
score_tuples.append((B_anchor_point_try,angle_to_try,distance_score))
# Step 4.3
lowest=None
for b_point,angle,distance in score_tuples:
print("point:{} angle(rad):{} distance(sum):{}".format(b_point,360*(angle/(2*math.pi)),distance))
if lowest is None or distance < lowest[2]:
lowest = b_point, 360*angle/(2*math.pi), distance
return lowest
def test_example():
ax = plt.subplot()
ax.grid(True)
plt.title('Fit Template to BBoxes by translation and rotation')
plt.xlim(-20, 20)
plt.ylim(-20, 20)
ax.set_xticks(range(-20,20), minor=True)
ax.set_yticks(range(-20,20), minor=True)
template = np.array([[-10,-10],[-10,10],[0,0],[10,-10],[10,10], [0,20]])
# Test Bboxes are Rotated 40 degree, translated 2,2
rotated = rotate_array(template,40)
rotated = np.subtract(rotated,[2,2])
# Adds some extra bounding boxes as noise
for i in range(8):
rotated = np.append(rotated,[[random.randrange(-20,20), random.randrange(-20,20)]],axis=0)
# Scramble entries in array and return the position change.
rnd_rotated = rotated.copy()
np.random.shuffle(rnd_rotated)
element_positions = []
# After shuffling, looks at which index the "A"-marks has ended up at. For later comparison to see that the algo found the correct answer.
# This is to represent the actual case, where I will get a bunch of unordered bboxes.
rnd_map = {}
indexes_translation = [num2 for num,point in enumerate(rnd_rotated) for num2,point2 in enumerate(rotated) if point[0]==point2[0] and point[1]==point2[1]]
for num,inx in enumerate(indexes_translation):
rnd_map[num]=inx
# algo part 1/3
b_point,angle,_ = find_stamp_rotation(template,rnd_rotated)
# Plot for visualization
legend_list = np.empty((0,2))
leg_template = plt.plot(template[:,0],template[:,1],c='r')
legend_list = np.append(legend_list,[[leg_template[0],'1. template-pattern']],axis=0)
leg_bboxes = plt.scatter(rnd_rotated[:,0],rnd_rotated[:,1],c='b',label="scatter")
legend_list = np.append(legend_list,[[leg_bboxes,'2. bounding boxes']],axis=0)
leg_anchor = plt.scatter(b_point[0],b_point[1],c='y')
legend_list = np.append(legend_list,[[leg_anchor,'3. Discovered bbox anchor point']],axis=0)
# algo part 2/3
# Superimpose A onto B by A[0] to b_point
offset = b_point - template[0]
super_imposed_A = template + offset
# Plot superimposed, but not yet rotated
leg_s_imposed = plt.plot(super_imposed_A[:,0],super_imposed_A[:,1],c='k')
#plt.legend(rubberduckz, "superimposed template on anchor")
legend_list = np.append(legend_list,[[leg_s_imposed[0],'4. Templ superimposed on Bbox']],axis=0)
print("Superimposed A on B by A[0] to {}".format(b_point))
print(super_imposed_A)
# Rotate, now the template should match pattern of bboxes
# algo part 3/4
super_imposed_rotated_A = rotate_array(super_imposed_A,-angle,rot_origin=super_imposed_A[0])
# Show the beautiful match in a last plot
leg_s_imp_rot = plt.plot(super_imposed_rotated_A[:,0],super_imposed_rotated_A[:,1],c='g')
legend_list = np.append(legend_list,[[leg_s_imp_rot[0],'5. final fit']],axis=0)
plt.legend(legend_list[:,0], legend_list[:,1],loc="upper left")
plt.show()
# algo part 4/4
pairs = pair_neighbours_by_distance(super_imposed_rotated_A, rnd_rotated, 10)
print(pairs)
for inx in range(len(pairs)):
bbox_num = pairs[inx]
print("template id:{}".format(inx))
print("bbox#id:{}".format(bbox_num))
#print("original_bbox:{}".format(rnd_map[bbox_num]))
if __name__ == "__main__":
test_example()
Result on actual image with bounding boxes. Here it can be seen that the scaling is incorrect which makes the template a bit off but it will still be able to pair up and thats the desired end-result in my case.
I'm interested in a fast way to calculate the rotation-independent center of a simple, convex, (non-intersecting) 2D polygon.
The example below (on the left) shows the mean center (sum of all points divided by the total), and the desired result on the right.
Some options I've already considered.
bound-box center (depends on rotation, and ignores points based on their relation to the axis).
Straight skeleton - too slow to calculate.
I've found a way which works reasonably well, (weight the points by the edge-lengths) - but this means a square-root call for every edge - which I'd like to avoid.(Will post as an answer, even though I'm not entirely satisfied with it).
Note, I'm aware of this questions similarity with:What is the fastest way to find the "visual" center of an irregularly shaped polygon?
However having to handle convex polygons increases the complexity of the problem significantly.
The points of the polygon can be weighted by their edge length which compensates for un-even point distribution.
This works for convex polygons too but in that case the center point isn't guaranteed to be inside the polygon.
Psudo-code:
def poly_center(poly):
sum_center = (0, 0)
sum_weight = 0.0
for point in poly:
weight = ((point - point.next).length +
(point - point.prev).length)
sum_center += point * weight
sum_weight += weight
return sum_center / sum_weight
Note, we can pre-calculate all edge lengths to halve the number of length calculations, or reuse the previous edge-length for half+1 length calculations. This is just written as an example to show the logic.
Including this answer for completeness since its the best method I've found so far.
There is no much better way than the accumulation of coordinates weighted by the edge length, which indeed takes N square roots.
If you accept an approximation, it is possible to skip some of the vertices by curve simplification, as follows:
decide of a deviation tolerance;
start from vertex 0 and jump to vertex M (say M=N/2);
check if the deviation along the polyline from 0 to M exceeds the tolerance (for this, compute the height of the triangle formed by the vertices 0, M/2, M);
if the deviation is exceeded, repeat recursively with 0, M/4, M/2 and M/2, 3M/4, M;
if the deviation is not exceeded, assume that the shape is straight between 0 and M.
continue until the end of the polygon.
Where the points are dense (like the left edge on your example), you should get some speedup.
I think its easiest to do something with the center of masses of the delaunay triangulation of the polygon points. i.e.
def _centroid_poly(poly):
T = spatial.Delaunay(poly).simplices
n = T.shape[0]
W = np.zeros(n)
C = 0
for m in range(n):
sp = poly[T[m,:],:]
W[m] = spatial.ConvexHull(sp).volume
C += W[m] +np.mean(sp, axis = 0)
return C / np.sum(W)
This works well for me!
I have a set of m non-rotated, integer (pixel) aligned rectangles, each of which may or may not overlap. The rectangles cover thousands of pixels. I need to find the minimum sized bounding box that covers all areas that are covered by n of the m rectangles.
A (dirty) way of doing this is to paint a canvas that covers the area of all the targets. This is O(mk) where m is the number of rectangles and k is the number of pixels per rectangle. However since k is much greater than m I think there is a better solution out there.
This feels like a dynamic programming problem...but I am having trouble figuring out the recursion.
Solution which is better but still not great:
Sort the start and end points of all the rectangles in the X direction O(mlogm), iterate and find the x positions that may have over n rectangles, O(m) loop. For each x position that may have over n rectangles, take the rectangles at that position and sort the starts and stops at that position (O(mlogm)). Find the region of overlap, keep track of the bounds that way. Overall, O(m^2logm).
Hello MadScienceDreams,
Just to clarify, the bounding box is also non-rotated, correct?
If this is the case, then just keep track of the four variables: minX, maxX, minY, maxY–representing left-most, right-most, top-most, and bottom-most pixels–that define the bounding box, loop through each of the rectangles updating the four variables, and defining the new bounding box given those four variables.
EDIT
It looks like you are asking about finding the bounds of some subset of rectangles, not the whole set.
So you have M rectangles, and you choose N rectangles from them, and find the bounds within that.
Even in this situation, looping through the N rectangles and keeping track of their bound would be at most O(m), which isn't bad at all.
I feel that I must be misunderstanding your question since this response isn't what you are probably looking for; is your question actually trying to ask how to precompute the bounds so that given any subset, know the total bounds in constant time?
Is this defines your question? For bounding box => #rect_label >= n
How about we starts with one box and find the next box that has nearest furthest corner from it. Now we have a region with two box. Recursively find the next region, until we have n boxes.
While we need to start on every box, we only need to actively work on the currently smallest regions. The effect is we start from the smallest cluster of boxes and expand out from there.
If n is closer to m than 0, we can reverse the search tree so that we start from the omni-all-enclosing box, chopping off each bordering box to create the next search level. Assuming we only actively work on the smallest remaining region, effect is we chop off the emptiest region first.
Is it too complicated? Sorry I can't remember the name of this search. I'm not good at maths, so I'll skip the O notation. >_<
I propose the following algorithm :
prepareData();
if (findBorder('left')) {
foreach (direction in ['top', 'right', 'bottom']) {
findBorder(direction)
}
} else noIntersectionExists
prepareData (O(mlogm)):
Order vertical bounds and horizontal bounds
Save the result as:
- two arrays that point to the rectangle (arrX and arrY)
- save the index as a property of the rectangle (rectangle.leftIndex, rectangle.topIndex, etc.
findBorder(left): // the other direction are similar
best case O(n), worst case O(2m-n)
arrIntersections = new Array;
//an intersection has a depth (number of rectangles intersected), a top and bottom border and list of rectangles
for(i=0; i < 2*m-n-1; i++){ // or (i = 2*m-1; i > n; i--)
if(isLeftBorder(arrX[i])){
addToIntersections(arrX[i].rectangle, direction);
if(max(intersections.depth) = n) break;
} else {
removeFromIntersections(arrX[i].rectangle, direction);
}
}
addToIntersections(rectangle, direction): // explanations for direction=left
Best case: O(n), worst case: O(m)
hasIntersected = false;
foreach(intersection in intersection){
if(intersect(intersection, rectangle)){
hasIntersected = true
intersections[] = {
depth: intersection.depth,
bottom: min(intersection.bottom, rectangle.bottom),
top: max(...)}
intersection.depth++
intersection.bottom = max(intersection.bottom, rectangle.bottom)
intersection.top = max(...)
}
}
if(!hasIntersected)
intersections[]={depth:1, bottom:rectangle.bottom, top:rectangle.top}
This gives an overall order between O(n^2) and O(m*(m-n/2))
I hope my pseudo code is clear enough
I have the x, y co-ordinates of a point on a rotated image by certain angle. I want to find the co-ordinates of the same point in the original, non-rotated image.
Please check the first image which is simpler:
UPDATED image, SIMPLIFIED:
OLD image:
Let's say the first point is A, the second is B and the last is C. I assume you have the rotation matrice R (see Wikipedia Rotation Matrix if not) et the translation vector t, so that B = R*A and C = B+t.
It comes C = R*A + t, and so A = R^1*(C-t).
Edit: If you only need the non rotated new point, simply do D = R^-1*C.
First thing to do is defining the reference system (how "where the points lies with respect to each image" will be translated into numbers). I guess that you want to rely on a basic 2D reference system, given by a single point (a couple of X/Y values). For example: left/lower corner (min. X and min. Y).
The algorithm is pretty straightforward:
Getting the new defining reference point associated with the
rotated shape (min. X and min. Y), that is, determining RefX_new and
RefY_new.
Applying a basic conversion between reference systems:
X_old = X_new + (RefX_new - RefX_old)
Y_old = Y_new + (RefY_new -
RefY_old)
----------------- UPDATE TO RELATE FORMULAE TO NEW CAR PIC
RefX_old = min X value of the CarFrame before being rotated.
RefY_old = max Y value of the CarFrame before being rotated.
RefX_new = min X value of the CarFrame after being rotated.
RefY_new = max Y value of the CarFrame after being rotated.
X_new = X of the point with respect to the CarFrame after being rotated. For example: if RefX_new = 5 with respect to absolute frame (0,0) and X of the point with respect to this absolute frame is 8, X_new would be 3.
Y_new = Y of the point with respect to CarFrame after being rotated (equivalently to point above)
X_old_C = X_new_C(respect to CarFrame) + (RefX_new(CarFrame_C) - RefX_old(CarFrame_A))
Y_old_C = Y_new_C(respect to CarFrame) + (RefY_new(CarFrame_C) - RefY_old(CarFrame_A))
These coordinates are respect to the CarFrame and thus you might have to update them with respect to the absolute frame (0,0, I guess), as explained above, that is:
X_old_D_absolute_frame = X_old_C + (RefX_new(CarFrame_C) + RefX_global(i.e., 0))
Y_old_D_absolute_frame = Y_old_C + (RefY_new(CarFrame_C) + RefY_global(i.e., 0))
(Although you should do that once the CarFrame is in its "definitive position" with respect to the global frame, that is, on picture D (the point has the same coordinates with respect to the CarFrame in both picture C and D, but different ones with respect to the global frame).)
It might seem a bit complex put in this way; but it is really simple. You have just to think carefully about one case and create the algorithm performing all the actions. The idea is extremely simple: if I am on 8 inside something which starts in 5; I am on 3 with respect to the container.
------------ UPDATE IN THE METHODOLOGY
As said in the comment, these last pictures prove that the originally-proposed calculation of reference (max. Y/min. X) is not right: it shouldn't be the max./min. values of the carFrame but the minimum distances to the closer sides (= perpendicular line from the left/bottom side to the point).
------------ TRIGONOMETRIC CALCS FOR THE SPECIFIC EXAMPLE
The algorithm proposed is the one you should apply in any situation. Although in this specific case, the most difficult part is not moving from one reference system to the other, but defining the reference point in the rotated system. Once this is done, the application to the non-rotated case is immediate.
Here you have some calcs to perform this action (I have done it pretty quickly, thus better take it as an orientation and do it by your own); also I have only considered the case in the pictures, that is, rotation over the left/bottom point:
X_rotated = dx * Cos(alpha)
where dx = X_orig - (max_Y_CarFrame - Y_Orig) * Tan(alpha)
Y_rotated = dy * Cos(alpha)
where dy = Y_orig - X_orig * Tan(alpha)
NOTE: (max_Y_CarFrame - Y_Orig) in dx and X_orig in dy expect that the basic reference system is 0,0 (min. X and min. Y). If this is not the case, you would have to change this variables.
The X_rotated and Y_rotated give the perpendicular distance from the point to the closest side of the carFrame (respectively, left and bottom side). By applying these formulae (I insist: analyse them carefully), you get the X_old_D_absolute_frame/Y_old_D_absolute_frame that is, you have just to add the lef/bottom values from the carFrame (if it is located in 0,0, these would be the final values).
Given a "shape" drawn by the user, I would like to "normalize" it so they all have similar size and orientation. What we have is a set of points. I can approximate the size using bounding box or circle, but the orientation is a bit more tricky.
The right way to do it, I think, is to calculate the majoraxis of its bounding ellipse. To do that you need to calculate the eigenvector of the covariance matrix. Doing so likely will be way too complicated for my need, since I am looking for some good-enough estimate. Picking min, max, and 20 random points could be some starter. Is there an easy way to approximate this?
Edit:
I found Power method to iteratively approximate eigenvector. Wikipedia article.
So far I am liking David's answer.
You'd be calculating the eigenvectors of a 2x2 matrix, which can be done with a few simple formulas, so it's not that complicated. In pseudocode:
// sums are over all points
b = -(sum(x * x) - sum(y * y)) / (2 * sum(x * y))
evec1_x = b + sqrt(b ** 2 + 1)
evec1_y = 1
evec2_x = b - sqrt(b ** 2 + 1)
evec2_y = 1
You could even do this by summing over only some of the points to get an estimate, if you expect that your chosen subset of points would be representative of the full set.
Edit: I think x and y must be translated to zero-mean, i.e. subtract mean from all x, y first (eed3si9n).
Here's a thought... What if you performed a linear regression on the points and used the slope of the resulting line? If not all of the points, at least a sample of them.
The r^2 value would also give you information about the general shape. The closer to 0, the more circular/uniform the shape is (circle/square). The closer to 1, the more stretched out the shape is (oval/rectangle).
The ultimate solution to this problem is running PCA
I wish I could find a nice little implementation for you to refer to...
Here you go! (assuming x is a nx2 vector)
def majAxis(x):
e,v = np.linalg.eig(np.cov(x.T)); return v[:,np.argmax(e)]