import cv2 import numpy as np import math from shapely.geometry import Polygon from rknnlite.api import RKNNLite import os # ------------------- 全局配置变量 ------------------- # 模型相关 CLASSES = ['clamp'] nmsThresh = 0.4 objectThresh = 0.35 # 可视化与保存控制（全局变量，可外部修改） DRAW_RESULT = True # 是否在输出图像上绘制旋转框 SAVE_PATH = None # 保存路径，如 "./result.jpg"；设为 None 则不保存 # RKNN 单例 _rknn_instance = None # ------------------- RKNN 管理函数 ------------------- def init_rknn(model_path): """只加载一次 RKNN 模型""" global _rknn_instance if _rknn_instance is None: _rknn_instance = RKNNLite(verbose=False) ret = _rknn_instance.load_rknn(model_path) if ret != 0: print(f"[ERROR] Failed to load RKNN model: {ret}") return None ret = _rknn_instance.init_runtime(core_mask=RKNNLite.NPU_CORE_0) if ret != 0: print(f"[ERROR] Failed to init runtime: {ret}") return None return _rknn_instance def release_rknn(): """释放 RKNN 对象""" global _rknn_instance if _rknn_instance: _rknn_instance.release() _rknn_instance = None # ------------------- 工具函数 ------------------- def letterbox_resize(image, size, bg_color=114): target_width, target_height = size image_height, image_width, _ = image.shape scale = min(target_width / image_width, target_height / image_height) new_width, new_height = int(image_width * scale), int(image_height * scale) image_resized = cv2.resize(image, (new_width, new_height), interpolation=cv2.INTER_AREA) canvas = np.ones((target_height, target_width, 3), dtype=np.uint8) * bg_color offset_x, offset_y = (target_width - new_width) // 2, (target_height - new_height) // 2 canvas[offset_y:offset_y + new_height, offset_x:offset_x + new_width] = image_resized return canvas, scale, offset_x, offset_y class DetectBox: def __init__(self, classId, score, xmin, ymin, xmax, ymax, angle): self.classId = classId self.score = score self.xmin = xmin self.ymin = ymin self.xmax = xmax self.ymax = ymax self.angle = angle def rotate_rectangle(x1, y1, x2, y2, a): cx, cy = (x1 + x2) / 2, (y1 + y2) / 2 cos_a, sin_a = math.cos(a), math.sin(a) pts = [(x1, y1), (x1, y2), (x2, y2), (x2, y1)] return [[int(cx + (xx - cx) * cos_a - (yy - cy) * sin_a), int(cy + (xx - cx) * sin_a + (yy - cy) * cos_a)] for xx, yy in pts] def intersection(g, p): g = Polygon(np.array(g).reshape(-1, 2)) p = Polygon(np.array(p).reshape(-1, 2)) if not g.is_valid or not p.is_valid: return 0 inter = g.intersection(p).area union = g.area + p.area - inter return 0 if union == 0 else inter / union def NMS(detectResult): predBoxs = [] sort_detectboxs = sorted(detectResult, key=lambda x: x.score, reverse=True) for i in range(len(sort_detectboxs)): if sort_detectboxs[i].classId == -1: continue p1 = rotate_rectangle(sort_detectboxs[i].xmin, sort_detectboxs[i].ymin, sort_detectboxs[i].xmax, sort_detectboxs[i].ymax, sort_detectboxs[i].angle) predBoxs.append(sort_detectboxs[i]) for j in range(i + 1, len(sort_detectboxs)): if sort_detectboxs[j].classId == sort_detectboxs[i].classId: p2 = rotate_rectangle(sort_detectboxs[j].xmin, sort_detectboxs[j].ymin, sort_detectboxs[j].xmax, sort_detectboxs[j].ymax, sort_detectboxs[j].angle) if intersection(p1, p2) > nmsThresh: sort_detectboxs[j].classId = -1 return predBoxs def sigmoid(x): x = np.clip(x, -709, 709) # 防止 exp 溢出 return np.where(x >= 0, 1 / (1 + np.exp(-x)), np.exp(x) / (1 + np.exp(x))) def softmax(x, axis=-1): exp_x = np.exp(x - np.max(x, axis=axis, keepdims=True)) return exp_x / np.sum(exp_x, axis=axis, keepdims=True) def process(out, model_w, model_h, stride, angle_feature, index, scale_w=1, scale_h=1): class_num = len(CLASSES) angle_feature = angle_feature.reshape(-1) xywh = out[:, :64, :] conf = sigmoid(out[:, 64:, :]) conf = conf.reshape(-1) boxes = [] for ik in range(model_h * model_w * class_num): if conf[ik] > objectThresh: w = ik % model_w h = (ik % (model_w * model_h)) // model_w c = ik // (model_w * model_h) xywh_ = xywh[0, :, (h * model_w) + w].reshape(1, 4, 16, 1) data = np.arange(16).reshape(1, 1, 16, 1) xywh_ = softmax(xywh_, 2) xywh_ = np.sum(xywh_ * data, axis=2).reshape(-1) xywh_add = xywh_[:2] + xywh_[2:] xywh_sub = (xywh_[2:] - xywh_[:2]) / 2 angle = (angle_feature[index + (h * model_w) + w] - 0.25) * math.pi cos_a, sin_a = math.cos(angle), math.sin(angle) xy = xywh_sub[0] * cos_a - xywh_sub[1] * sin_a, xywh_sub[0] * sin_a + xywh_sub[1] * cos_a xywh1 = np.array([xy[0] + w + 0.5, xy[1] + h + 0.5, xywh_add[0], xywh_add[1]]) xywh1 *= stride xmin = (xywh1[0] - xywh1[2] / 2) * scale_w ymin = (xywh1[1] - xywh1[3] / 2) * scale_h xmax = (xywh1[0] + xywh1[2] / 2) * scale_w ymax = (xywh1[1] + xywh1[3] / 2) * scale_h boxes.append(DetectBox(c, conf[ik], xmin, ymin, xmax, ymax, angle)) return boxes # ------------------- 主推理函数 ------------------- def detect_two_box_angle(model_path, rgb_frame): """ 输入模型路径和 RGB 图像（numpy array），输出夹角和结果图像。 可视化和保存由全局变量 DRAW_RESULT 和 SAVE_PATH 控制。 """ global _rknn_instance, DRAW_RESULT, SAVE_PATH if not isinstance(rgb_frame, np.ndarray) or rgb_frame is None: print(f"[ERROR] detect_two_box_angle 接收到错误类型: {type(rgb_frame)}") return None, np.zeros((640, 640, 3), np.uint8) # 注意：输入是 BGR（因为 cv2.imread 返回 BGR），但内部会转为 RGB 给模型 img = rgb_frame.copy() img_resized, scale, offset_x, offset_y = letterbox_resize(img, (640, 640)) infer_img = np.expand_dims(cv2.cvtColor(img_resized, cv2.COLOR_BGR2RGB), 0) try: rknn = init_rknn(model_path) if rknn is None: return None, img results = rknn.inference([infer_img]) except Exception as e: print(f"[ERROR] RKNN 推理失败: {e}") return None, img outputs = [] for x in results[:-1]: index, stride = 0, 0 if x.shape[2] == 20: stride, index = 32, 20*4*20*4 + 20*2*20*2 elif x.shape[2] == 40: stride, index = 16, 20*4*20*4 elif x.shape[2] == 80: stride, index = 8, 0 feature = x.reshape(1, 65, -1) outputs += process(feature, x.shape[3], x.shape[2], stride, results[-1], index) predbox = NMS(outputs) print(f"[DEBUG] 检测到 {len(predbox)} 个框") if len(predbox) < 2: print("检测少于两个目标，无法计算夹角。") return None, img predbox = sorted(predbox, key=lambda x: x.score, reverse=True) box1, box2 = predbox[:2] output_img = img.copy() if DRAW_RESULT else img # 若不绘制，则直接用原图 if DRAW_RESULT: for box in [box1, box2]: xmin = int((box.xmin - offset_x) / scale) ymin = int((box.ymin - offset_y) / scale) xmax = int((box.xmax - offset_x) / scale) ymax = int((box.ymax - offset_y) / scale) points = rotate_rectangle(xmin, ymin, xmax, ymax, box.angle) cv2.polylines(output_img, [np.array(points, np.int32)], True, (0, 255, 0), 2) def main_direction(box): w, h = (box.xmax - box.xmin)/scale, (box.ymax - box.ymin)/scale direction = box.angle if w >= h else box.angle + np.pi/2 return direction % np.pi dir1 = main_direction(box1) dir2 = main_direction(box2) diff = abs(dir1 - dir2) diff = min(diff, np.pi - diff) angle_deg = np.degrees(diff) # 保存结果（如果需要） if SAVE_PATH: save_dir = os.path.dirname(SAVE_PATH) if save_dir: # 非空目录才创建 os.makedirs(save_dir, exist_ok=True) cv2.imwrite(SAVE_PATH, output_img) return angle_deg, output_img # ------------------- 示例调用 ------------------- if __name__ == "__main__": MODEL_PATH = "./obb.rknn" IMAGE_PATH = "./11.jpg" # === 全局控制开关 === DRAW_RESULT = True # 是否绘制框 SAVE_PATH = "./result11.jpg" # 保存路径，设为 None 则不保存 frame = cv2.imread(IMAGE_PATH) if frame is None: print(f"[ERROR] 无法读取图像: {IMAGE_PATH}") else: angle_deg, output_image = detect_two_box_angle(MODEL_PATH, frame) if angle_deg is not None: print(f"检测到的角度差: {angle_deg:.2f}°") else: print("未能成功检测到目标或计算角度差")