Flask+Python+MySQL环境下YOLOv5检测程序报错及自动关闭问题

1. 解决AttributeError: 'tuple' object has no attribute 'encode'错误 #

问题根源 #

get_product_data函数查询到数据时返回元组(message, image_data)，但detect_objects中直接对返回值调用encode()方法，元组没有该方法；同时函数返回值类型不统一，无数据时会引发未定义变量的错误。

修复代码 #

修改get_product_data函数

def get_product_data(object_name):
    cnx = mysql.connector.connect(user='root', password='', host='localhost', database='product')
    cursor = cnx.cursor()
    # 改用参数化查询避免SQL注入
    query = "SELECT * FROM information WHERE product_name = %s"
    cursor.execute(query, (object_name,))
    data = cursor.fetchone()
    
    select_query = "SELECT image_data FROM information WHERE product_name = %s"
    cursor.execute(select_query, (object_name,))
    result = cursor.fetchone()
    
    cursor.close()
    cnx.close()

    if data:
        message = f"<br>產品名稱：{data[1]}<br>價格：{data[2]}<br>折扣價格：{data[3]}<br>折扣：{data[4]}<br>"
        image_data = result[0] if result else None
        return message, image_data
    else:
        return None, None

修改detect_objects中的调用逻辑

# 替换原有的分判断逻辑，统一处理
if object_name not in object_names:
    object_names.append(object_name)
    message, _ = get_product_data(object_name)
    if message:
        yield message.encode()

2. 解决程序偶尔自动关闭的问题 #

问题根源 #

• stop_video中调用cap.release()释放摄像头后，后续无法重新打开导致程序崩溃
• 摄像头初始化无异常处理，启动失败直接报错
• 循环中的break导致只处理第一个检测物体，引发逻辑异常

修复代码 #

修改stop_video和resume_video接口

@app.route('/stop_video', methods=['POST'])
def stop_video():
    global video_paused
    video_paused = True
    # 不直接释放摄像头，保留重启能力
    return '已停止'

@app.route('/resume_video', methods=['POST'])
def resume_video():
    global video_paused, cap, height, width, baseline
    video_paused = False
    # 摄像头已释放则重新初始化
    if not cap.isOpened():
        cap = cv2.VideoCapture(0)
        success, frame = cap.read()
        if success:
            height, width, _ = frame.shape
            baseline = height // 2 + 100
    return '已繼續'

移除detect_objects中的多余break

for obj in detected_objects:
    obj_class = int(obj[-1])
    if obj_class in [0, 1, 2]:
        obj_center_x = (obj[0] + obj[2]) / 2 + 100
        obj_center_y = (obj[1] + obj[3]) / 2 + 100
        if obj_center_y > baseline:
            object_name = ""
            if obj_class == 0:
                object_name = "柚香"
            elif obj_class == 1:
                object_name = "福樂"
            elif obj_class == 2:
                object_name = "美祿"

            if object_name not in object_names:
                object_names.append(object_name)
                message, _ = get_product_data(object_name)
                if message:
                    yield message.encode()
# 移除此处的break，确保遍历所有检测物体

3. 其他优化点 #

• 简化product_page逻辑，避免重复判断：

@app.route('/product_page')
def product_page():
    object_name = request.args.get('object_name')
    message = None
    image_base64 = None
    if object_name:
        message, image_data = get_product_data(object_name)
        if image_data:
            image_base64 = base64.b64encode(image_data).decode('utf-8')
    return render_template('product_page.html', message=message, image_data=image_base64)

内容的提问来源于stack exchange，提问作者TanNeil