ESP32-S3 摄像头图传：固件 + Slint 客户端 + 共享协议

两块 ESP32-S3-WROOM-1-N16R8 图传板（DVP 接口接 OV2640/OV3660/OV5640），向队伍客户端与裁判终端推 UDP 图传。固件是跑在 ESP-IDF 上的 Rust，用官方 esp32-camera 组件采集；客户端是 Rust + Slint 桌面程序。三个 crate 各自独立构建： . firmware，交叉编译到 Xtensa protocol/ 线路格式与帧重组，两端共用，无 ESP-IDF 依赖，有单元测试 client/ Slint 桌面客户端视频走 UDP 而非 TCP：实测传感器产出 22 fps，TCP MJPEG 只送出 1.9 fps，且只占用链路能力的 1%——链路是空的，帧堵在队头阻塞后面等一个实时画面等不起的重传计时器。改为独立数据报、不重传、缺片丢整帧后达到 20.8 fps。发现是双向的：板子每 2 秒广播 Hello，客户端广播 Subscribe，板子据此定向推流。板子身份由 MAC 派生，重启不变，所以客户端可以把绑定存进配置。裁判端同时收两块板子（每个发送方一个重组器），队伍端绑定其中一块并只显示自己那一边的血量。装甲板接口已预留（ArmorSource trait + ARMOR_PORT），当前无数据源，界面显示占位符。 vendor 了 esp32-camera 2.1.7：组件注册表的版本求解器找不到它，且需要打一处补丁（见 components/esp32-camera/target/esp32s3/ll_cam.c）。 .gitignore 三条 target/ 都锚定在根目录，因为 protocol/ 与 client/ 各自也是独立 crate；而 components/esp32-camera/target/ 是上游的按芯片分目录，里面是源码，必须保留。
2026-10-05 18:22:19 +08:00
commit 4f893ce140
134 changed files with 41128 additions and 0 deletions
@@ -0,0 +1,110 @@
+#!/usr/bin/env python3
+"""Minimal UDP listener for the camera stream, used to verify the firmware.
+
+Not the real client - just enough to confirm datagrams arrive, that they
+reassemble into a valid JPEG, and to measure the frame rate independently of
+the Slint app.
+"""
+import socket, sys, time, collections
+
+HOST = "192.168.1.139"
+PORT = 5600
+DURATION = float(sys.argv[1]) if len(sys.argv) > 1 else 10.0
+
+# Must match protocol/src/lib.rs: MAGIC, HEADER_LEN, MAX_FRAGMENT_PAYLOAD,
+# DEFAULT_PORT. Duplicated here on purpose - this is a throwaway diagnostic, and
+# making it depend on the Rust crate would defeat the point of it being an
+# independent check of the wire format.
+HEADER = 20
+MAGIC = 0x4D564331
+FRAG = 1400
+
+sock = socket.socket(socket.AF_INET, socket.SOCK_DGRAM)
+sock.setsockopt(socket.SOL_SOCKET, socket.SO_REUSEADDR, 1)
+sock.bind(("", PORT))
+
+# Announce ourselves, exactly as the Slint client will.
+sock.sendto(b"\x00", (HOST, PORT))
+sock.settimeout(0.5)
+
+frag_bytes = FRAG
+frames = {}
+stats = collections.Counter()
+sizes = []
+start = time.time()
+last_print = start
+
+while time.time() - start < DURATION:
+    try:
+        data, _ = sock.recvfrom(2048)
+    except socket.timeout:
+        continue
+
+    stats["datagrams"] += 1
+
+    if len(data) < HEADER:
+        stats["short"] += 1
+        continue
+
+    magic = int.from_bytes(data[0:4], "little")
+    if magic != MAGIC:
+        stats["bad_magic"] += 1
+        continue
+
+    frame_id = int.from_bytes(data[4:8], "little")
+    frag_index = int.from_bytes(data[8:10], "little")
+    frag_count = int.from_bytes(data[10:12], "little")
+    total_len = int.from_bytes(data[12:16], "little")
+    width = int.from_bytes(data[16:18], "little")
+    height = int.from_bytes(data[18:20], "little")
+
+    payload = data[HEADER:]
+
+    if frame_id not in frames:
+        if frames:
+            stats["superseded"] += 1
+        frames[frame_id] = {
+            "buf": bytearray(total_len),
+            "seen": 0,
+            "count": frag_count,
+            "total": total_len,
+            "w": width,
+            "h": height,
+            "t0": time.time(),
+        }
+
+    f = frames[frame_id]
+    off = frag_index * frag_bytes
+    if off + len(payload) <= total_len:
+        f["buf"][off:off + len(payload)] = payload
+        f["seen"] += 1
+    else:
+        stats["overflow"] += 1
+
+    if f["seen"] >= f["count"]:
+        jpeg = bytes(f["buf"])
+        if jpeg[:2] == b"\xff\xd8" and jpeg[-2:] == b"\xff\xd9":
+            stats["good"] += 1
+            sizes.append((f["total"], f["w"], f["h"]))
+        else:
+            stats["corrupt"] += 1
+        del frames[frame_id]
+
+    now = time.time()
+    if now - last_print >= 2.0:
+        elapsed = now - last_print
+        print(
+            f"  {stats['good']/elapsed:5.1f} fps  "
+            f"datagrams={stats['datagrams']:5d}  "
+            f"superseded={stats['superseded']:4d}  corrupt={stats['corrupt']:3d}  "
+            f"good={stats['good']:4d}"
+        )
+        last_print = now
+
+sock.close()
+print()
+print("totals:", dict(stats))
+if sizes:
+    avg = sum(s[0] for s in sizes) / len(sizes)
+    print(f"avg frame: {avg:.0f} bytes, {sizes[0][1]}x{sizes[0][2]}")
+    print(f"throughput: {sum(s[0] for s in sizes)/DURATION/1024:.1f} kB/s")