【Raspberry Pi】電子工作入門⑤～OpenCVの物体認識～【Python】【AI入門】

# Otsu + 反転 → 輪郭抽出
# Picamera2 + OpenCV

from picamera2 import Picamera2
import cv2, time

def show_small(winname, img, width=640):
    """画像を等比縮小して表示（学習用・任意）"""
    h, w = img.shape[:2]
    scale = width / w
    cv2.imshow(winname, cv2.resize(img, (width, int(h * scale))))

def main():
    # カメラ初期化（解像度640x360）
    cam = Picamera2()
    cam.configure(cam.create_preview_configuration(main={"size": (640, 360)}))
    cam.start()
    time.sleep(0.2)  # 起動安定

    # フル視野を使用（デジタルズーム防止）
    try:
        sw, sh = cam.camera_properties["PixelArraySize"]
        cam.set_controls({"ScalerCrop": (0, 0, int(sw), int(sh))})
    except:
        cam.set_controls({"ScalerCrop": (0, 0, 2592, 1944)})

    while True:
        # フレーム取得 → RGB→BGR→GRAY
        frame = cv2.cvtColor(cam.capture_array(), cv2.COLOR_RGB2BGR)
        gray  = cv2.cvtColor(frame, cv2.COLOR_BGR2GRAY)

        # 二値化（黒物体を白にするため Otsu + INV）
        _, binary = cv2.threshold(
            gray, 0, 255,
            cv2.THRESH_BINARY_INV + cv2.THRESH_OTSU
        )

        # 最外輪郭のみ抽出
        contours, _ = cv2.findContours(
            binary, cv2.RETR_EXTERNAL, cv2.CHAIN_APPROX_SIMPLE
        )

        # 輪郭描画
        vis = frame.copy()
        cv2.drawContours(vis, contours, -1, (0,0,255), 2)
        cv2.putText(
            vis, f"Contours: {len(contours)}", (10,28),
            cv2.FONT_HERSHEY_SIMPLEX, 0.8, (0,200,255), 2
        )

        # 縮小表示（見やすさのため）
        show_small("Binary (INV+Otsu)", binary, 480)
        show_small("Detected Contours", vis, 640)

        # s：保存 / q：終了
        key = cv2.waitKey(1) & 0xFF
        if key == ord('s'):
            cv2.imwrite("contours_frame.jpg", vis)
            print("Saved: contours_frame.jpg")
        if key == ord('q'):
            break

    cv2.destroyAllWindows()
    cam.stop()

if __name__ == "__main__":
    main()

# 図形検出（Otsu+反転 → 輪郭近似 → 円形度補強）
# Picamera2 + OpenCV

from picamera2 import Picamera2
import cv2, numpy as np, time

def show_small(winname, img, width=640):
    """等比で縮小表示（学習用・任意）"""
    h, w = img.shape[:2]
    cv2.imshow(winname, cv2.resize(img, (width, int(h * (width / w)))))

def classify_shape(cnt):
    """輪郭から図形名と近似多角形を返す（小ノイズは除外）"""
    area = cv2.contourArea(cnt)
    if area < 80: return None, None  # 面積でノイズ除外
    peri = cv2.arcLength(cnt, True)
    approx = cv2.approxPolyDP(cnt, 0.02 * peri, True)  # 近似精度=周長の2%
    v = len(approx)
    if v == 3:
        name = "Triangle"
    elif v == 4:
        x, y, w, h = cv2.boundingRect(approx)
        aspect = w / float(h)
        name = "Square" if 0.90 <= aspect <= 1.10 else "Rectangle"
    else:
        circ = (4.0 * np.pi * area / (peri * peri)) if peri > 0 else 0  # 1に近いほど円
        name = "Circle" if circ > 0.80 else "Polygon"
    return name, approx

def main():
    # カメラ初期化（640x480に変更：4:3で視野が自然）
    cam = Picamera2()
    cam.configure(cam.create_preview_configuration(main={"size": (640, 480)}))
    cam.start(); time.sleep(0.2)

    # フル視野を使用（デジタルズーム防止）
    try:
        sw, sh = cam.camera_properties["PixelArraySize"]
        cam.set_controls({"ScalerCrop": (0, 0, int(sw), int(sh))})
    except:
        cam.set_controls({"ScalerCrop": (0, 0, 2592, 1944)})

    while True:
        # 入力 → RGB→BGR→GRAY → ぼかし → Otsu+反転（黒図形→白）
        frame = cv2.cvtColor(cam.capture_array(), cv2.COLOR_RGB2BGR)
        gray  = cv2.cvtColor(frame, cv2.COLOR_BGR2GRAY)
        blur  = cv2.GaussianBlur(gray, (3, 3), 0.8)
        _, binary = cv2.threshold(blur, 0, 255, cv2.THRESH_BINARY_INV + cv2.THRESH_OTSU)

        # 輪郭（最外）抽出
        contours, _ = cv2.findContours(binary, cv2.RETR_EXTERNAL, cv2.CHAIN_APPROX_SIMPLE)

        # 図形分類と描画
        counts = {"Triangle": 0, "Square": 0, "Rectangle": 0, "Circle": 0, "Polygon": 0}
        vis = frame.copy()
        for c in contours:
            name, approx = classify_shape(c)
            if not name: continue
            counts[name] = counts.get(name, 0) + 1
            cv2.drawContours(vis, [approx], -1, (0, 0, 255), 2)  # 赤：輪郭
            x, y, w, h = cv2.boundingRect(approx)
            cv2.putText(vis, name, (x, y - 6), cv2.FONT_HERSHEY_SIMPLEX, 0.6, (50, 50, 255), 2)

        # 集計表示
        summary = f"Tri:{counts['Triangle']}  Sq:{counts['Square']}  Rect:{counts['Rectangle']}  Cir:{counts['Circle']}  Poly:{counts['Polygon']}"
        cv2.putText(vis, summary, (10, 28), cv2.FONT_HERSHEY_SIMPLEX, 0.8, (0, 180, 255), 2)

        # 表示（縮小は見やすさのため）
        show_small("Binary (INV+Otsu)", binary, 480)
        show_small("Shape Detection (Live)", vis, 640)

        key = cv2.waitKey(1) & 0xFF
        if key == ord('s'):
            cv2.imwrite("shape_live_frame.jpg", vis); print("Saved: shape_live_frame.jpg")
        if key == ord('q'):
            break

    cv2.destroyAllWindows(); cam.stop()

if __name__ == "__main__":
    main()

M = cv2.moments(cnt)
cx = int(M["m10"]/M["m00"])
cy = int(M["m01"]/M["m00"])