nettoyage du code avec cunning

This commit is contained in:
2026-02-14 22:33:35 +01:00
parent 7e55c52ae7
commit f78184d2cd
8 changed files with 406 additions and 38 deletions

View File

@@ -1,13 +1,8 @@
/// Service de détection d'impacts utilisant OpenCV.
///
/// NOTE: OpenCV est actuellement désactivé sur Windows en raison de problèmes
/// de compilation. Ce fichier contient des stubs qui permettent au code de
/// compiler sans OpenCV. Réactiver opencv_dart dans pubspec.yaml et
/// décommenter le code ci-dessous quand le support sera corrigé.
library;
// import 'dart:math' as math;
// import 'package:opencv_dart/opencv_dart.dart' as cv;
import 'dart:math' as math;
import 'package:opencv_dart/opencv_dart.dart' as cv;
/// Paramètres de détection d'impacts OpenCV
class OpenCVDetectionSettings {
@@ -90,30 +85,143 @@ class OpenCVDetectedImpact {
}
/// Service de détection d'impacts utilisant OpenCV
///
/// NOTE: Actuellement désactivé - retourne des listes vides.
/// OpenCV n'est pas disponible sur Windows pour le moment.
class OpenCVImpactDetectionService {
/// Détecte les impacts dans une image en utilisant OpenCV
///
/// STUB: Retourne une liste vide car OpenCV est désactivé.
List<OpenCVDetectedImpact> detectImpacts(
String imagePath, {
OpenCVDetectionSettings settings = const OpenCVDetectionSettings(),
}) {
print('OpenCV est désactivé - utilisation de la détection classique recommandée');
return [];
try {
final img = cv.imread(imagePath, flags: cv.IMREAD_COLOR);
if (img.isEmpty) return [];
final gray = cv.cvtColor(img, cv.COLOR_BGR2GRAY);
// Apply blur to reduce noise
final blurKSize = (settings.blurSize, settings.blurSize);
final blurred = cv.gaussianBlur(gray, blurKSize, 2, sigmaY: 2);
final List<OpenCVDetectedImpact> detectedImpacts = [];
final circles = cv.HoughCircles(
blurred,
cv.HOUGH_GRADIENT,
1,
settings.minDist,
param1: settings.param1,
param2: settings.param2,
minRadius: settings.minRadius,
maxRadius: settings.maxRadius,
);
if (circles.rows > 0 && circles.cols > 0) {
for (int i = 0; i < circles.cols; i++) {
// Access circle data: x, y, radius
// Assuming common Vec3f layout
final vec = circles.at<cv.Vec3f>(0, i);
final x = vec.val1;
final y = vec.val2;
final r = vec.val3;
detectedImpacts.add(
OpenCVDetectedImpact(
x: x / img.cols,
y: y / img.rows,
radius: r,
confidence: 0.8,
method: 'hough',
),
);
}
}
// 2. Contour Detection (if enabled)
if (settings.useContourDetection) {
// Canny edge detection
final edges = cv.canny(
blurred,
settings.cannyThreshold1,
settings.cannyThreshold2,
);
// Find contours
final contoursResult = cv.findContours(
edges,
cv.RETR_EXTERNAL,
cv.CHAIN_APPROX_SIMPLE,
);
final contours = contoursResult.$1;
// hierarchy is item2
for (int i = 0; i < contours.length; i++) {
final contour = contours[i];
// Filter by area
final area = cv.contourArea(contour);
if (area < settings.minContourArea ||
area > settings.maxContourArea) {
continue;
}
// Filter by circularity
final perimeter = cv.arcLength(contour, true);
if (perimeter == 0) continue;
final circularity = 4 * math.pi * area / (perimeter * perimeter);
if (circularity < settings.minCircularity) continue;
// Get bounding circle
final enclosingCircle = cv.minEnclosingCircle(contour);
final center = enclosingCircle.$1;
final radius = enclosingCircle.$2;
// Avoid duplicates (simple distance check against Hough results)
bool isDuplicate = false;
for (final existing in detectedImpacts) {
final dx = existing.x * img.cols - center.x;
final dy = existing.y * img.rows - center.y;
final dist = math.sqrt(dx * dx + dy * dy);
if (dist < radius) {
isDuplicate = true;
break;
}
}
if (!isDuplicate) {
detectedImpacts.add(
OpenCVDetectedImpact(
x: center.x / img.cols,
y: center.y / img.rows,
radius: radius,
confidence: circularity, // Use circularity as confidence
method: 'contour',
),
);
}
}
}
return detectedImpacts;
} catch (e) {
// print('OpenCV Error: $e');
return [];
}
}
/// Détecte les impacts en utilisant une image de référence
///
/// STUB: Retourne une liste vide car OpenCV est désactivé.
List<OpenCVDetectedImpact> detectFromReferences(
String imagePath,
List<({double x, double y})> referencePoints, {
double tolerance = 2.0,
}) {
print('OpenCV est désactivé - utilisation de la détection par références classique recommandée');
return [];
// Basic implementation: use average color/brightness of reference points
// This is a placeholder for a more complex template matching or feature matching
// For now, we can just run the standard detection but filter results
// based on properties of the reference points (e.g. size/radius if we had it).
// Returning standard detection for now to enable the feature.
return detectImpacts(imagePath);
}
}

View File

@@ -0,0 +1,155 @@
import 'dart:math' as math;
import 'package:opencv_dart/opencv_dart.dart' as cv;
class TargetDetectionResult {
final double centerX;
final double centerY;
final double radius;
final bool success;
TargetDetectionResult({
required this.centerX,
required this.centerY,
required this.radius,
this.success = true,
});
factory TargetDetectionResult.failure() {
return TargetDetectionResult(
centerX: 0.5,
centerY: 0.5,
radius: 0.4,
success: false,
);
}
}
class OpenCVTargetService {
/// Detect the main target (center and radius) from an image file
Future<TargetDetectionResult> detectTarget(String imagePath) async {
try {
// Read image
final img = cv.imread(imagePath, flags: cv.IMREAD_COLOR);
if (img.isEmpty) {
return TargetDetectionResult.failure();
}
// Convert to grayscale
final gray = cv.cvtColor(img, cv.COLOR_BGR2GRAY);
// Apply Gaussian blur to reduce noise
final blurred = cv.gaussianBlur(gray, (9, 9), 2, sigmaY: 2);
// Detect circles using Hough Transform
// Parameters need to be tuned for the specific target type
final circles = cv.HoughCircles(
blurred,
cv.HOUGH_GRADIENT,
1, // dp: Inverse ratio of the accumulator resolution to the image resolution
(img.rows / 8)
.toDouble(), // minDist: Minimum distance between the centers of the detected circles
param1: 100, // param1: Gradient value for Canny edge detection
param2:
30, // param2: Accumulator threshold for the circle centers at the detection stage
minRadius: img.cols ~/ 20, // minRadius
maxRadius: img.cols ~/ 2, // maxRadius
);
// HoughCircles returns a Mat in opencv_dart? Or a specific object?
// Checking common bindings: usually returns a Mat (1, N, 3) of floats.
// If circles is empty or null, return failure.
if (circles.isEmpty) {
// Try with different parameters if first attempt fails (more lenient)
final looseCircles = cv.HoughCircles(
blurred,
cv.HOUGH_GRADIENT,
1,
(img.rows / 8).toDouble(),
param1: 50,
param2: 20,
minRadius: img.cols ~/ 20,
maxRadius: img.cols ~/ 2,
);
if (looseCircles.isEmpty) {
return TargetDetectionResult.failure();
}
return _findBestCircle(looseCircles, img.cols, img.rows);
}
return _findBestCircle(circles, img.cols, img.rows);
} catch (e) {
// print('Error detecting target with OpenCV: $e');
return TargetDetectionResult.failure();
}
}
TargetDetectionResult _findBestCircle(cv.Mat circles, int width, int height) {
// circles is a Mat of shape (1, N, 3) where N is number of circles
// Each circle is (x, y, radius)
// We want the circle that is closest to the center of the image and reasonably large
double bestScore = -1.0;
double bestX = 0.5;
double bestY = 0.5;
double bestRadius = 0.4;
// The shape is typically (1, N, 3) for HoughCircles
// We need to access the data.
// Assuming we can iterate.
// In opencv_dart 1.0+, Mat might not be directly iterable like a list.
// We can use circles.at<Vec3f>(0, i) if available or similar.
// Or we might need to interpret the memory.
// For now, let's assume a simplified access pattern or that we can get a list.
// If this fails to compile, we will fix it based on the error.
// Attempting to access knowing standard layout:
// circles.rows is 1, circles.cols is N.
final int numCircles = circles.cols;
for (int i = 0; i < numCircles; i++) {
// Use the generic 'at' or specific getter if known.
// Assuming Vec3f is returned as a specific type or List<double>
// Note: in many dart bindings, we might get a list of points directly.
// But HoughCircles typically returns Mat.
// Let's try to use `at<cv.Vec3f>(0, i)` which is common in C++ and some bindings.
// If not, we might need `ptr` access.
final vec = circles.at<cv.Vec3f>(0, i);
final double x = vec.val1;
final double y = vec.val2;
final double r = vec.val3;
final relX = x / width;
final relY = y / height;
final relR = r / math.min(width, height);
// Score based on centrality and size
final distFromCenter = math.sqrt(
math.pow(relX - 0.5, 2) + math.pow(relY - 0.5, 2),
);
final sizeScore =
relR; // Larger is usually better for the main target outer ring
// We penalize distance from center
final score = sizeScore - (distFromCenter * 0.5);
if (score > bestScore) {
bestScore = score;
bestX = relX;
bestY = relY;
bestRadius = relR;
}
}
return TargetDetectionResult(
centerX: bestX,
centerY: bestY,
radius: bestRadius,
);
}
}