findEssentialMat with USAC returns None since version 4.8.0 for valid input #24482
Description
System Information
OpenCV python version: 4.8.1 build py310hff52083_3 from conda-forge
Operating System / Platform: Ubuntu 18.04
Python version: 3.10.13
Detailed description
Hi, starting with opencv version 4.8.1 from conda-forge cv2.findEssentialMat() returns None for many cases that returned a valid solution with opencv version 4.5.5 and 4.7.0 when using USAC. both versions return a sane output for essential matrix when using the classic RANSAC solver. I think this is related to #23078
Steps to reproduce
import cv2
import numpy as np
points = np.array([[ 71.35362447, -73.48993604, 71.35376603, -73.59021804],
[ -88.90364312, -73.61561921, -88.90364312, -73.61561921],
[-167.59229029, -152.62241114, -167.59229029, -152.62241114],
[ 85.94304076, -162.11201907, 85.94304076, -162.11201907],
[ 172.05289178, -162.39757996, 172.05289178, -162.39757996],
[ -82.58191992, -74.50782902, -82.58191992, -74.50782902],
[-272.6442017 , -271.23413327, -272.6442017 , -271.23413327],
[ -85.77380211, -273.03948033, -85.77380211, -273.03948033],
[ 103.14325005, -274.71871417, 103.24549557, -274.8227237 ],
[ 294.19549172, -276.49581391, 294.19549172, -276.49581391],
[-167.81015725, -168.7491632 , -167.70862579, -168.74839359],
[ -79.41421943, -169.26774438, -79.31358168, -169.2673819 ],
[ 98.36577158, -170.26210817, 98.36623108, -170.36341088],
[ -74.05887631, -77.30305365, -74.05887631, -77.30305365],
[ 10.260825 , -77.54947512, 10.16076431, -77.64968961],
[ 64.70994691, -159.00848207, 64.8104607 , -159.00876493],
[ 48.0005312 , -71.95875096, 48.00062479, -72.05898542],
[ 48.02296035, -233.95576851, 48.02296035, -233.95576851],
[-113.99758728, -153.02210264, -113.99758728, -153.02210264],
[ 128.95027293, -153.00278013, 128.95027293, -153.00278013],
[-531.82226296, 193.00149743, -531.80771136, 192.4747974 ],
[-186.99847614, -123.50523002, -187.10073587, -123.60693988],
[ 825.53441926, -521.32613133, 825.79691647, -521.35039178],
[ 463.62302476, 145.67892849, 463.62110312, 145.57512415],
[ -28.49414672, -420.41603093, -28.391098 , -420.20116664],
[ 88.09145441, 107.10489058, 88.09092315, 106.90369874],
[ 292.93675706, -383.43338 , 293.04229644, -383.4365222 ],
[ 129.05236356, -477.10686816, 129.05063377, -476.99727052],
[ -90.58267673, 207.89241318, -91.58575871, 206.6746328 ],
[-228.89530246, 173.56976397, -228.27218077, 172.84930809],
[1065.99119855, 132.21384591, 1066.13767853, 132.21790114],
[ 318.41220056, -421.45654236, 318.30185375, -421.34411053],
[-294.17708945, 24.11502293, -294.17772815, 24.41862693],
[ 502.3126993 , 98.53781371, 502.09285013, 98.63878911],
[-128.0704807 , -435.95125478, -128.06893292, -435.84322097],
[ 129.84221632, -273.09259683, 129.9458877 , -273.30012846],
[-476.91443292, -386.09419314, -476.69686157, -386.19301364],
[ 575.52220458, -501.78447164, 575.53034267, -501.89910723],
[-690.58899158, 47.85699357, -690.34568101, 47.429626 ],
[-522.11192937, -463.88901488, -522.11192937, -463.88901488],
[-708.64648858, 70.8426063 , -708.75428266, 69.88293808],
[-958.54350679, 153.7748682 , -958.40575988, 153.77065468]])
focal = 1914.731689453125
threshold = 3
E, _ = cv2.findEssentialMat(points[:, :2], points[:, 2:], focal=focal, pp=(0, 0), method=cv2.USAC_MAGSAC,
threshold=threshold)
assert E is not None, 'unfortunately E is None'
Issue submission checklist
- I report the issue, it's not a question
- I checked the problem with documentation, FAQ, open issues, forum.opencv.org, Stack Overflow, etc and have not found any solution
- I updated to the latest OpenCV version and the issue is still there
- There is reproducer code and related data files (videos, images, onnx, etc)