如何从Lepton 3.5辐射热像仪提取温度最值及均值

已成功读取掩膜矩形区域的像素值数组，现有实现代码如下：

heatmap_gray = cv2.cvtColor(frame, cv2.COLOR_RGB2GRAY)
heatmap = cv2.applyColorMap(heatmap_gray, cv2.COLORMAP_HOT)

# Binary threshold
_, binary_thresh = cv2.threshold(heatmap_gray, threshold, 255, cv2.THRESH_BINARY)

# Image opening: Erosion followed by dilation
kernel = np.ones((3, 3), np.uint8)
image_erosion = cv2.erode(binary_thresh, kernel, iterations=1)
image_opening = cv2.dilate(image_erosion, kernel, iterations=1)

# Get contours from the image obtained by opening operation
contours, _ = cv2.findContours(image_opening, 1, 2) 

image_with_rectangles = np.copy(frame)

for contour in contours:
    # rectangle over each contour
    x, y, w, h = cv2.boundingRect(contour)

    # Pass if the area of rectangle is not large enough
    if (w) * (h) < area_of_box:
        continue

    # Mask is boolean type of matrix.
    mask = np.zeros_like(heatmap_gray)
    cv2.drawContours(mask, contour, -1, 255, -1)
    max = convert_to_temperature(np.amax(heatmap_gray+mask))
    print(heatmap_gray+mask)

当前已获取到掩膜区域的像素值数组，需要提取该区域的温度值并计算最大值、最小值和均值。

解决方案 #

1. 正确提取掩膜区域像素：原代码中heatmap_gray + mask的方式错误，应通过布尔索引筛选掩膜覆盖的有效像素：

   # 提取掩膜区域内的像素值
   masked_pixels = heatmap_gray[mask == 255]
   

2. 批量转换为温度值：将筛选出的像素值批量转换为温度（确保convert_to_temperature支持数组输入，若不支持可结合np.vectorize包装）：

   # 转换为温度数组
   temperature_values = convert_to_temperature(masked_pixels)
   

3. 计算统计值：直接对温度数组计算最大值、最小值和均值：

   # 计算统计值
   temp_max = np.max(temperature_values)
   temp_min = np.min(temperature_values)
   temp_mean = np.mean(temperature_values)
   

修改后的完整循环代码 #

替换原循环内的逻辑为：

for contour in contours:
    x, y, w, h = cv2.boundingRect(contour)
    if (w) * (h) < area_of_box:
        continue

    mask = np.zeros_like(heatmap_gray)
    cv2.drawContours(mask, contour, -1, 255, -1)
    
    # 提取有效像素并转换为温度
    masked_pixels = heatmap_gray[mask == 255]
    temperature_values = convert_to_temperature(masked_pixels)
    
    # 计算并输出统计值
    temp_max = np.max(temperature_values)
    temp_min = np.min(temperature_values)
    temp_mean = np.mean(temperature_values)
    
    print(f"区域温度最大值: {temp_max}, 最小值: {temp_min}, 均值: {temp_mean}")

注意事项 #

• 确保convert_to_temperature函数正确匹配Lepton 3.5的像素-温度转换规则：若你将原始16位热图转为了8位灰度图，需确认转换时的缩放系数，避免温度计算偏差。
• 若convert_to_temperature仅支持单个值输入，可使用np.vectorize(convert_to_temperature)包装后处理数组。

内容的提问来源于stack exchange，提问作者RChristan