去除FFMPEG后端

.github/copilot-instructions.md

@@ -0,0 +1,44 @@
+# Copilot / AI agent instructions for this repo
+
+目标：帮助 AI 编码代理快速上手此环视(surround-view)项目，聚焦可验证的代码模式、运行流程和常见修改点。
+
+- **大体架构**：数据流为 Capture -> Undistort -> Project -> Flip -> Stitch -> WhiteBalance -> Display。主要线程/模块：
+  - `surround_view.capture_thread.CaptureThread`：采集摄像头帧（USB/CSI，可传 `api_preference`）。
+  - `surround_view.fisheye_camera.FisheyeCameraModel`：负责 `undistort()`, `project()`, `flip()` 和相机参数读写（yaml 文件）。
+  - `surround_view.process_thread.CameraProcessingThread`：每个摄像头的处理线程，调用相机模型并写入 `ProjectedImageBuffer`。
+  - `surround_view.birdview.BirdView`：融合/拼接、亮度匹配与白平衡，输出最终鸟瞰图。
+  - 同步/缓冲：`surround_view.imagebuffer.Buffer`、`MultiBufferManager` 与 `ProjectedImageBuffer`，通过 `sync()`/信号等待实现多流同步。
+
+- **关键运行脚本（推荐顺序）**：
+  1. `python run_calibrate_camera.py`（标定/生成 camera yaml）
+  2. `python run_get_projection_maps.py -camera <front|back|left|right> -scale ... -shift ...`（选择投影点）
+  3. `python run_get_weight_matrices.py`（生成拼接权重/掩码）
+  4. `python run_live_demo.py`（车载/实时 demo）
+  - 调试摄像头采集：`python test_cameras.py`
+
+- **项目约定 / 易错点（请严格遵循）**：
+  - 摄像头名称固定为 `front, back, left, right`，对应 yaml/param 中的配置；device id 需在本机环境中确认（参见 `test_cameras.py`）。
+  - yaml 文件（`yaml/*.yaml`）里保存的字段：`camera_matrix`, `dist_coeffs`, `resolution`, `project_matrix`, `scale_xy`, `shift_xy`。修改后用 `FisheyeCameraModel.save_data()` 写回。
+  - `param_settings.py` 用 cm 为单位构造 birdview 参数（`total_w/total_h`、`xl/xr/yt/yb` 等），birdview 的切片函数（`FI, FM, ...`）依赖这些常量。
+  - 镜头“翻转”策略在 `FisheyeCameraModel.flip()` 明确：`front` 不变，`back` 180°, `left/right` 用转置+翻转处理。
+  - 缓冲区 `Buffer.add(..., drop_if_full=True)` 会丢帧；修 bug 时注意是否因为丢帧导致状态不一致。
+
+- **拼接与亮度调整实现要点**（参见 `surround_view/birdview.py` 与 `surround_view/utils.py`）：
+  - 权重计算：`get_weight_mask_matrix()` 提取重叠区域并基于点到多边形距离计算平滑权重 G。
+  - 亮度匹配通过 `mean_luminance_ratio()` 聚合重叠区域平均亮度并以几何/经验方式融合（见 `make_luminance_balance()`）。
+  - 白平衡用 `make_white_balance()` 简单按通道均值缩放。
+
+- **如何修改/扩展（示例）**：
+  - 若加入新相机或改变布局：更新 `param_settings.py` 的 `camera_names`、`project_shapes` 与 `xl/xr/yt/yb`；更新 yaml 参数并通过 `FisheyeCameraModel` 测试 `undistort()`/`project()` 输出。
+  - 若替换采集后端（gstreamer / v4l2），优先修改 `surround_view.utils.gstreamer_pipeline()` 与 `CaptureThread.connect_camera()` 的 `api_preference` 分支，确保 `cap.isOpened()` 行为一致。
+
+- **测试 / 调试建议**：
+  - 使用 `test_cameras.py` 确认设备索引与支持分辨率。
+  - 单线程走查：直接在 REPL 导入 `FisheyeCameraModel`，用已存在的 yaml 文件运行 `undistort()`、`project()` 并保存图片来验证投影矩阵。
+  - 帧同步问题优先检查 `MultiBufferManager.sync()` 与 `ProjectedImageBuffer.sync()` 的 `do_sync` 与 `sync_devices` 配置。
+
+- **风格/约定**：
+  - 代码使用 Python 3、OpenCV、PyQt5，多线程依赖 `QThread` + Qt 同步原语；避免用 `time.sleep()` 作为同步手段。
+  - 尽量在修改后用现有脚本跑完整流程（calibrate -> get_projection -> get_weight -> live_demo）以捕获参数联动问题。
+
+如果需要，我可以把这些点合并成更短或更详尽的版本，或者加入常见问题/故障排查列表；或者你希望我直接提交此文件为 PR？请告诉我想要的调整或遗漏的内容。

.vscode/tasks.json

@@ -0,0 +1,12 @@
+{
+    // See https://go.microsoft.com/fwlink/?LinkId=733558
+    // for the documentation about the tasks.json format
+    "version": "2.0.0",
+    "tasks": [
+        {
+            "label": "kill python",
+            "type": "shell",
+            "command": "pkill -f python3"
+        }
+    ]
+}

run.sh

@@ -0,0 +1,18 @@
+#!/bin/bash
+
+# 获取当前脚本的 PID
+SELF_PID=$$
+
+echo "【1/3】终止其他 run.sh（跳过自身 PID=$SELF_PID）..."
+# 使用 pgrep + grep -v 排除自己
+pgrep -f "run.sh" | grep -v "^$SELF_PID$" | xargs kill -9 2>/dev/null
+
+echo "【2/3】终止 run_live_demo.py ..."
+pkill -9 -f "run_live_demo.py" 2>/dev/null
+
+sleep 1
+
+
+echo "【3/3】启动新实例..."
+exec python3 ./run_live_demo.py
+

run_live_demo.py

@@ -8,14 +8,15 @@ import surround_view.param_settings as settings
 
 yamls_dir = os.path.join(os.getcwd(), "yaml")
 camera_ids = [0, 1, 2, 3]
-flip_methods = [0, 2, 0, 2]
+flip_methods = [0, 0, 0,0]
+
 names = settings.camera_names
 cameras_files = [os.path.join(yamls_dir, name + ".yaml") for name in names]
 camera_models = [FisheyeCameraModel(camera_file, name) for camera_file, name in zip(cameras_files, names)]
 
 
 def main():
-    capture_tds = [CaptureThread(camera_id, flip_method)
+    capture_tds = [CaptureThread(camera_id, flip_method,resolution=(1920, 1080))
                    for camera_id, flip_method in zip(camera_ids, flip_methods)]
     capture_buffer_manager = MultiBufferManager()
     for td in capture_tds:
@@ -38,17 +39,18 @@ def main():
     while True:
         img = cv2.resize(birdview.get(), (300, 400))
         cv2.imshow("birdview", img)
+
         key = cv2.waitKey(1) & 0xFF
         if key == ord("q"):
             break
 
-        for td in capture_tds:
-            print("camera {} fps: {}\n".format(td.device_id, td.stat_data.average_fps), end="\r")
+        # for td in capture_tds:
+        #     print("camera {} fps: {}\n".format(td.device_id, td.stat_data.average_fps), end="\r")
 
-        for td in process_tds:
-            print("process {} fps: {}\n".format(td.device_id, td.stat_data.average_fps), end="\r")
+        # for td in process_tds:
+        #     print("process {} fps: {}\n".format(td.device_id, td.stat_data.average_fps), end="\r")
 
-        print("birdview fps: {}".format(birdview.stat_data.average_fps))
+        # print("birdview fps: {}".format(birdview.stat_data.average_fps))
 
 
     for td in process_tds:

saveimg.py

@@ -10,7 +10,7 @@ running = True
 
 def video_thread():
     global frame, running
-    cap = cv2.VideoCapture(0, cv2.CAP_ANY)
+    cap = cv2.VideoCapture(1, cv2.CAP_ANY)
     cap.set(cv2.CAP_PROP_FOURCC,cv2.VideoWriter_fourcc(*"YUYV"))
     cap.set(cv2.CAP_PROP_FRAME_WIDTH, 1920)
     cap.set(cv2.CAP_PROP_FRAME_HEIGHT, 1080)
@@ -18,7 +18,7 @@ def video_thread():
 
     if not cap.isOpened():
         print("[ERROR] Cannot open camera", file=sys.stderr)
-        running = Falseq
+        running = False
         return
 
     while running:
@@ -26,7 +26,13 @@ def video_thread():
         if not ret:
             break
         frame = f.copy()
-        cv2.imshow('Live Feed (Local Display)', f)
+        # print(frame.shape)
+        # frame = cv2.resize(frame, (1280, 720))
+        # print(frame.shape)
+        cv2.namedWindow('AHD Video', cv2.WND_PROP_FULLSCREEN)
+        cv2.setWindowProperty('AHD Video', cv2.WND_PROP_FULLSCREEN, cv2.WINDOW_FULLSCREEN)
+        cv2.imshow('AHD Video', f)
+        # cv2.showFullscreen('Live Feed (Local Display)', f)
         if cv2.waitKey(1) & 0xFF == ord('q'):
             running = False
             break

surround_view/birdview.py

@@ -104,6 +104,7 @@ def LIII(left_image):
 
 
 def LM(left_image):
+
     return left_image[yt:yb, :]
 
 
@@ -116,6 +117,7 @@ def RIV(right_image):
 
 
 def RM(right_image):
+    
     return right_image[yt:yb, :]
 
 
@@ -194,7 +196,10 @@ class BirdView(BaseThread):
         return self.image[yt:yb, xl:xr]
 
     def stitch_all_parts(self):
+       
         front, back, left, right = self.frames
+        print(front.shape, back.shape, left.shape, right.shape)
+
         np.copyto(self.F, FM(front))
         np.copyto(self.B, BM(back))
         np.copyto(self.L, LM(left))
@@ -326,7 +331,8 @@ class BirdView(BaseThread):
             self.processing_mutex.lock()
 
             self.update_frames(self.proc_buffer_manager.get().values())
-            self.make_luminance_balance().stitch_all_parts()
+            self.stitch_all_parts()  # 直接拼接原始投影图像
+            # self.make_luminance_balance().stitch_all_parts()
             self.make_white_balance()
             self.copy_car_image()
             self.buffer.add(self.image.copy(), self.drop_if_full)

surround_view/capture_thread.py

@@ -1,9 +1,7 @@
 import cv2
 from PyQt5.QtCore import qDebug
-
 from .base_thread import BaseThread
 from .structures import ImageFrame
-from .utils import gstreamer_pipeline
 
 
 
@@ -11,11 +9,10 @@ class CaptureThread(BaseThread):
 
     def __init__(self,
                  device_id,
-                #  flip_method=2,
+                 flip_method=0,
                  drop_if_full=True,
                  api_preference=cv2.CAP_ANY,
                  resolution=None,
-                #  use_gst=None,
                  parent=None):
         """
         device_id: device number of the camera.
@@ -27,8 +24,7 @@ class CaptureThread(BaseThread):
         """
         super(CaptureThread, self).__init__(parent)
         self.device_id = device_id
-        # self.flip_method = flip_method
-        # self.use_gst = None
+        self.flip_method = flip_method
         self.drop_if_full = drop_if_full
         self.api_preference = api_preference
         self.resolution = resolution
@@ -61,15 +57,25 @@ class CaptureThread(BaseThread):
                 continue
 
             # retrieve frame and add it to buffer
-            _, frame = self.cap.retrieve()
+            _, frame = self.cap.read()
+          
+            # Skip empty frames (e.g., when camera capture times out)
+            if frame is None or frame.size == 0:
+                continue
+                
+            # Apply image flip if needed
+            if self.flip_method == 2:
+                frame = cv2.rotate(frame, cv2.ROTATE_180)
+                
+            # frame = cv2.resize(frame, (960, 640))
             img_frame = ImageFrame(self.clock.msecsSinceStartOfDay(), frame)
             self.buffer_manager.get_device(self.device_id).add(img_frame, self.drop_if_full)
 
             # update statistics
-            self.update_fps(self.processing_time)
-            self.stat_data.frames_processed_count += 1
-            # inform GUI of updated statistics
-            self.update_statistics_gui.emit(self.stat_data)
+            # self.update_fps(self.processing_time)
+            # self.stat_data.frames_processed_count += 1
+            # # inform GUI of updated statistics
+            # self.update_statistics_gui.emit(self.stat_data)
 
         qDebug("Stopping capture thread...")
 
@@ -83,7 +89,6 @@ class CaptureThread(BaseThread):
             # try to set camera resolution
             if self.resolution is not None:
                 width, height = self.resolution
-
                 self.cap.set(cv2.CAP_PROP_FOURCC, cv2.VideoWriter_fourcc(*"YUYV"))
                 self.cap.set(cv2.CAP_PROP_FRAME_WIDTH, width)
                 self.cap.set(cv2.CAP_PROP_FRAME_HEIGHT, height)

surround_view/fisheye_camera.py

@@ -92,7 +92,7 @@ class FisheyeCameraModel(object):
             return cv2.transpose(image)[::-1]
 
         else:
-            return np.flip(cv2.transpose(image), 1)
+            return cv2.transpose(image)[::-1]
 
     def save_data(self):
         fs = cv2.FileStorage(self.camera_file, cv2.FILE_STORAGE_WRITE)

surround_view/process_thread.py

@@ -2,6 +2,7 @@ import cv2
 from PyQt5.QtCore import qDebug, QMutex
 
 from .base_thread import BaseThread
+from . import param_settings as settings
 
 
 class CameraProcessingThread(BaseThread):
@@ -47,10 +48,28 @@ class CameraProcessingThread(BaseThread):
 
             self.processing_mutex.lock()
             raw_frame = self.capture_buffer_manager.get_device(self.device_id).get()
+            
+            # Validate frame before processing
+            if raw_frame.image is None or raw_frame.image.size == 0:
+                self.processing_mutex.unlock()
+                continue
+                
             und_frame = self.camera_model.undistort(raw_frame.image)
             pro_frame = self.camera_model.project(und_frame)
             flip_frame = self.camera_model.flip(pro_frame)
             self.processing_mutex.unlock()
+            
+            # Check if the processed frame has valid dimensions
+            name = self.camera_model.camera_name
+            if name in settings.project_shapes:
+                # For left and right cameras, the flip operation changes the shape
+                if name in ['left', 'right']:
+                    expected_shape = settings.project_shapes[name] + (3,)
+                else:
+                    expected_shape = settings.project_shapes[name][::-1] + (3,)
+                if flip_frame.shape != expected_shape:
+                    print(f"Warning: {name} camera frame has unexpected shape {flip_frame.shape}, expected {expected_shape}")
+                    continue
 
             self.proc_buffer_manager.sync(self.device_id)
             self.proc_buffer_manager.set_frame_for_device(self.device_id, flip_frame)

yaml/left.yaml

@@ -16,7 +16,7 @@ resolution: !!opencv-matrix
    rows: 2
    cols: 1
    dt: i
-   data: [ 960, 640 ]
+   data: [ 1920, 1080 ]
 project_matrix: !!opencv-matrix
    rows: 3
    cols: 3

yaml/right.yaml

@@ -1,19 +1,37 @@
 %YAML:1.0
 ---
+camera_matrix: !!opencv-matrix
+   rows: 3
+   cols: 3
+   dt: d
+   data: [ 3.0245305983229298e+02, 0., 4.9664001463163459e+02, 0.,
+       3.2074618594392325e+02, 3.3119980984361649e+02, 0., 0., 1. ]
+dist_coeffs: !!opencv-matrix
+   rows: 4
+   cols: 1
+   dt: d
+   data: [ -4.3735601598704078e-02, 2.1692522970939803e-02,
+       -2.6388839028513571e-02, 8.4123126605702321e-03 ]
 resolution: !!opencv-matrix
    rows: 2
    cols: 1
    dt: i
    data: [ 1920, 1080 ]
-camera_matrix: !!opencv-matrix
+project_matrix: !!opencv-matrix
    rows: 3
    cols: 3
    dt: d
-   data: [ 5.3426649626056496e+02, 0., 9.4035331422643389e+02, 0.,
-       5.3438087838176853e+02, 5.6780561206232505e+02, 0., 0., 1. ]
-dist_coeffs: !!opencv-matrix
-   rows: 4
+   data: [ -7.0390891066994388e-01, -2.5544083216952904e+00,
+       7.0809808916259806e+02, -2.9600383808093766e-01,
+       -2.4971504395791286e+00, 6.3578234365104447e+02,
+       -5.6872782515522376e-04, -4.4482832729892769e-03, 1. ]
+scale_xy: !!opencv-matrix
+   rows: 2
    cols: 1
-   dt: d
-   data: [ -1.3928766380167932e-02, 5.8925693280706653e-04,
-       -1.1389190533024249e-03, -9.6700186038395686e-05 ]
+   dt: f
+   data: [ 6.99999988e-01, 8.00000012e-01 ]
+shift_xy: !!opencv-matrix
+   rows: 2
+   cols: 1
+   dt: f
+   data: [ -150., -100. ]