orisys-tech/orisys-sdk
3
0
Fork 0
Code Issues Pull Requests Actions Packages Projects Releases Wiki Activity

stereo first commit

Browse Source
This commit is contained in:
unknown
2026-01-03 11:13:44 +08:00
parent a3cd52a301
commit d5c73670fe
7 changed files with 214 additions and 14 deletions
Download Patch File Download Diff File Expand all files Collapse all files
config/compound_eye.png
BIN
View File
Binary file not shown.
Side by Side Swipe Overlay

Before

Width:  |  Height:  |  Size: 2.0 MiB

After

Width:  |  Height:  |  Size: 1.6 MiB

config/stereo.json
+35
View File
@@ -0,0 +1,35 @@
{
"name": "prototype",
"version": "1.0",
"camera": {
"resolution_raw": [1920, 1080],
"fps": 60,
"units": 15,
"units_used": 9,
"resolution_stitch": [400, 400]
},
"stitcher": {
"min_area_frac": 0.01,
"max_area_frac": 0.05,
"max_aspect_ratio": 100,
"index": [1, 2, 3, 6, 7, 8, 11, 12, 13],
"margin": 10,
"nominal_ov_pos": 100,
"n": [3, 3]
},
"tracker": {
"adaptive": true,
"lower_threshold": 12.0,
"upper_threshold": 18.0,
"loop_closure_interval": 1,
"dis_preset": "medium"
},
"contact": {
"min_contour_area": 3000,
"flow_threshold": 1,
"div_threshold": 0.02,
"expansion_threshold": 0.01,
"kernel_size": [5, 5],
"denoise_sigma": 15
}
}
src/orisys/Stitcher.py
+1 -1
View File
@@ -111,7 +111,7 @@ class Stitcher:
img_corr = self.__correct_edge_bloom(img_src)
self.__detect_nonblack_regions(img_corr,show = show)
return len(self.boxes)
return len(self.boxes), self.boxes
# return self.boxes
src/orisys/sensor.py
+15 -13
View File
@@ -131,7 +131,7 @@ class Sensor:
# Initialize components
self.__stitcher = Stitcher(isstitch=isstitch, config_stitcher=self.__config_stitcher, verbose=self.verbose)
self.stitcher = Stitcher(isstitch=isstitch, config_stitcher=self.__config_stitcher, verbose=self.verbose)
self.__tracker = Tracker(adaptive=True, cuda=cuda, verbose=self.verbose, dis_preset=self._config_dis_preset)
self.__contact3d = Contact3D(config_contact=self.__config_contact, verbose=self.verbose)
self.__hhd = None
@@ -168,17 +168,18 @@ class Sensor:
# Convert to grayscale
if log_timing: t0 = time.perf_counter()
gray = cv2.cvtColor(frame, cv2.COLOR_BGR2GRAY)
# gray = cv2.cvtColor(frame, cv2.COLOR_BGR2GRAY)
gray = frame
if log_timing: timings['cvt_color'] = (time.perf_counter() - t0) * 1000
# Extract and stitch tiles
if log_timing: t0 = time.perf_counter()
self.__stitcher.get_stitch_unit(gray)
self.stitcher.get_stitch_unit(gray)
self._extract_tiles(gray)
if log_timing: timings['get_tiles'] = (time.perf_counter() - t0) * 1000
if log_timing: t0 = time.perf_counter()
img = np.ascontiguousarray(self.__stitcher.stitch_rt_optimized())
img = np.ascontiguousarray(self.stitcher.stitch_rt_optimized())
if log_timing: timings['stitch'] = (time.perf_counter() - t0) * 1000
self.__img_size_raw = img.shape
@@ -379,12 +380,12 @@ class Sensor:
if self.verbose:
print("Loading calibration from", self.cal_path)
self.stitch_calibration = np.load(self.cal_path, allow_pickle=True)
self.__stitcher.stitch_calibration = self.stitch_calibration
return self.__stitcher.stitch_rt_optimized()
self.stitcher.stitch_calibration = self.stitch_calibration
return self.stitcher.stitch_rt_optimized()
else:
if self.verbose:
print("Creating new calibration...")
img, self.stitch_calibration = self.__stitcher.stitch_cal()
img, self.stitch_calibration = self.stitcher.stitch_cal()
os.makedirs("./config", exist_ok=True)
np.save(self.cal_path, self.stitch_calibration)
if self.verbose:
@@ -448,8 +449,8 @@ class Sensor:
if override:
if box is not None:
self.__stitcher.isstitch = self.isstitch = False
self.__stitcher.boxes = box
self.stitcher.isstitch = self.isstitch = False
self.stitcher.boxes = box
else:
print("Override=True but no box provided.")
else:
@@ -467,15 +468,16 @@ class Sensor:
v_mean = np.sum(cv2.mean(frame))
if abs(v_mean - v_mean0) < 5:
n_boxes = self.__stitcher.find_stitch_boxes(frame, show=show)
n_boxes, boxes = self.stitcher.find_stitch_boxes(frame, show=show)
if (cv2.getTickCount() - start_time) / cv2.getTickFrequency() > timeout:
print("Timeout finding vision units.")
self.__stitcher.isstitch = self.isstitch = False
self.stitcher.isstitch = self.isstitch = False
break
v_mean0 = v_mean
self.__n_boxes = n_boxes
self.boxes = boxes
self.stitch_valid = self.isstitch and n_boxes == self._config_units
status = "valid for stitching" if self.stitch_valid else "non-stitch mode"
@@ -484,8 +486,8 @@ class Sensor:
def _extract_tiles(self, img_src):
"""Extract tiles from source image."""
self.__stitcher.get_stitch_unit(img_src)
self.__stitcher.extract_tiles()
self.stitcher.get_stitch_unit(img_src)
self.stitcher.extract_tiles()
return True
stereo_1.jpg
BIN
View File
Binary file not shown.
Side by Side

After

Width:  |  Height:  |  Size: 195 KiB

stereo_1.mp4
BIN
View File
Binary file not shown.
streo.py
+163
View File
@@ -0,0 +1,163 @@
import numpy as np
import cv2
from matplotlib import pyplot as plt
import orisys
sensor = orisys.Sensor(vid_src='stereo_1.mp4',config_name='./config/stereo.json',mirror=False)
print(sensor.boxes)
sensor.get_img()
tiles = sensor.stitcher.tile_list
print(tiles)
imgL = tiles[0]
imgR = tiles[1]
# imgL = cv.imread('tsukuba_l.png', cv.IMREAD_GRAYSCALE)
# imgR = cv.imread('tsukuba_r.png', cv.IMREAD_GRAYSCALE)
plt.subplot(1,2,1)
plt.imshow(imgL,'gray')
plt.subplot(1,2,2)
plt.imshow(imgR,'gray')
plt.show()
# Feature matching for stereo computation
def match_features(img1, img2, method='SIFT', max_matches=50):
"""
Detect and match features between two images.
Args:
img1: Left image (grayscale)
img2: Right image (grayscale)
method: 'SIFT' or 'ORB'
max_matches: Maximum number of matches to show
Returns:
img_matches: Image with matches drawn
matches: List of matches
kp1, kp2: Keypoints in img1 and img2
"""
# Convert to grayscale if needed
if len(img1.shape) == 3:
img1_gray = cv2.cvtColor(img1, cv2.COLOR_BGR2GRAY)
else:
img1_gray = img1.copy()
if len(img2.shape) == 3:
img2_gray = cv2.cvtColor(img2, cv2.COLOR_BGR2GRAY)
else:
img2_gray = img2.copy()
# Initialize detector
if method == 'SIFT':
detector = cv2.SIFT_create()
elif method == 'ORB':
detector = cv2.ORB_create()
else:
raise ValueError("method must be 'SIFT' or 'ORB'")
# Detect keypoints and descriptors
kp1, des1 = detector.detectAndCompute(img1_gray, None)
kp2, des2 = detector.detectAndCompute(img2_gray, None)
if des1 is None or des2 is None or len(des1) == 0 or len(des2) == 0:
print(f"Warning: No features detected with {method}")
return None, [], kp1, kp2
# Match features
if method == 'SIFT':
# Check if we have enough descriptors for knnMatch (need at least k=2)
min_descriptors = min(len(des1), len(des2))
if min_descriptors < 2:
print(f"Warning: Not enough descriptors ({min_descriptors}) for knnMatch, using regular matching")
# Fallback to BFMatcher
matcher = cv2.BFMatcher(cv2.NORM_L2, crossCheck=False)
matches = matcher.match(des1, des2)
matches = sorted(matches, key=lambda x: x.distance)[:max_matches]
else:
# Use FLANN matcher for SIFT
try:
FLANN_INDEX_KDTREE = 1
index_params = dict(algorithm=FLANN_INDEX_KDTREE, trees=5)
search_params = dict(checks=50)
matcher = cv2.FlannBasedMatcher(index_params, search_params)
matches = matcher.knnMatch(des1, des2, k=2)
# Apply Lowe's ratio test
good_matches = []
for match_pair in matches:
if len(match_pair) == 2:
m, n = match_pair
if m.distance < 0.7 * n.distance:
good_matches.append(m)
matches = sorted(good_matches, key=lambda x: x.distance)[:max_matches]
except cv2.error:
# Fallback to BFMatcher if FLANN fails
print("Warning: FLANN matcher failed, using BFMatcher instead")
matcher = cv2.BFMatcher(cv2.NORM_L2, crossCheck=False)
matches = matcher.match(des1, des2)
matches = sorted(matches, key=lambda x: x.distance)[:max_matches]
else: # ORB
# Use BF matcher for ORB
matcher = cv2.BFMatcher(cv2.NORM_HAMMING, crossCheck=True)
matches = matcher.match(des1, des2)
matches = sorted(matches, key=lambda x: x.distance)[:max_matches]
# Draw matches
img_matches = cv2.drawMatches(
img1_gray, kp1, img2_gray, kp2, matches, None,
flags=cv2.DrawMatchesFlags_NOT_DRAW_SINGLE_POINTS
)
print(f"Found {len(matches)} good matches using {method}")
return img_matches, matches, kp1, kp2
# Perform feature matching
print("\n=== Feature Matching ===")
img_matches, matches, kp1, kp2 = match_features(imgL, imgR, method='SIFT', max_matches=50)
if img_matches is not None:
plt.figure(figsize=(15, 6))
plt.imshow(img_matches)
plt.title(f'Feature Matches (SIFT) - {len(matches)} matches')
plt.axis('off')
plt.tight_layout()
plt.show()
# Also create a side-by-side visualization with lines connecting matches
h1, w1 = imgL.shape[:2]
h2, w2 = imgR.shape[:2]
h_max = max(h1, h2)
combined = np.zeros((h_max, w1 + w2), dtype=np.uint8)
combined[:h1, :w1] = imgL
combined[:h2, w1:w1+w2] = imgR
combined_color = combined
# Draw lines connecting matched features
for match in matches[:30]: # Show top 30 matches
pt1 = tuple(map(int, kp1[match.queryIdx].pt))
pt2 = tuple(map(int, kp2[match.trainIdx].pt))
pt2 = (pt2[0] + w1, pt2[1])
cv2.line(combined_color, pt1, pt2, (0, 255, 0), 1)
cv2.circle(combined_color, pt1, 3, (0, 0, 255), -1)
cv2.circle(combined_color, pt2, 3, (255, 0, 0), -1)
plt.figure(figsize=(15, 6))
plt.imshow(cv2.cvtColor(combined_color, cv2.COLOR_BGR2RGB))
plt.title(f'Stereo Feature Matching - {len(matches[:30])} matches shown')
plt.axis('off')
plt.tight_layout()
plt.show()
# Stereo disparity computation
print("\n=== Stereo Disparity Computation ===")
stereo = cv2.StereoBM.create(numDisparities=32, blockSize=15)
disparity = stereo.compute(imgL, imgR)
plt.figure(figsize=(12, 5))
plt.imshow(disparity, cmap='jet')
plt.colorbar(label='Disparity')
plt.title('Stereo Disparity Map')
plt.axis('off')
plt.tight_layout()
plt.show()