YOLOv8 构建的无人机红外图像下极小目标检测系统 | 极客日志

PythonAI算法

YOLOv8 构建的无人机红外图像下极小目标检测系统

本数据集包含 8302 张红外图像，涵盖飞机、无人机、直升机及鸟类四类极小目标。基于 YOLOv8 搭建的检测系统支持 Web 端上传与可视化，提供 VOC 与 YOLO 双格式标注。系统采用 Flask 框架实现轻量部署，具备实时推理与统计功能，适用于低空安防、机场净空监测等场景。代码结构清晰，便于二次开发与模型优化。

DockerOne发布于 2026/3/29更新于 2026/4/251 浏览

YOLOv8 构建的无人机红外图像下极小目标检测系统

基于 YOLOv8 的无人机红外极小目标检测系统

在低空安防、机场净空监测以及反无人机系统中，红外图像下的极小目标检测一直是个挑战。这里分享一个实战项目，包含 8302 张标注数据，涵盖飞机、无人机、直升机和鸟类四类目标，并基于 YOLOv8 搭建了一套完整的 Web 检测系统。

数据集概览

本数据集共包含 8302 张红外图像，采用 YOLO (.txt) 和 PASCAL VOC (.xml) 双格式标注。目标尺寸普遍较小（通常小于 32×32 像素），对模型的特征提取能力提出了较高要求。

类别	英文标识	标注框数量
飞机	Plane	2,163
无人机	Drone	3,120
直升机	Heli	2,217
鸟类	Bird	1,958
总计	—	9,458

样本展示如下：

图 1：典型红外背景下的无人机目标
图 2：复杂光照环境中的飞机与鸟类混合场景

系统架构与依赖

系统采用 Flask 提供 Web 服务，结合 Ultralytics YOLOv8 进行推理。整体结构清晰，便于部署到边缘服务器。

目录结构

infrared_small_target_detection_yolov8/
├── data/
│   └── dataset.yaml      # 数据集配置
├── models/
│   └── best.pt           # 训练好的权重
├── static/
│   ├── uploads/          # 用户上传目录
│   └── results/          # 结果保存目录
├── templates/
│   ├── index.html        # 上传界面
│   └── result.html       # 结果展示
├── app.py                # Flask 主程序
├── detect.py             # 核心检测逻辑
├── requirements.txt
└── README.md

环境准备

安装必要的依赖包即可启动：

pip install -r requirements.txt

requirements.txt 内容如下：

ultralytics==8.2.0
flask==2.3.3
opencv-python==4.8.0
numpy==1.24.3
Pillow==9.5.0
matplotlib

核心代码实现

train: ../data/images/train
val: ../data/images/val
test: ../data/images/test
nc: 4
names: ['Plane', 'Drone', 'Heli', 'Bird']

# detect.py
import cv2
import os
from ultralytics import YOLO

CLASS_NAMES = {
    0: "飞机",
    1: "无人机",
    2: "直升机",
    3: "鸟类"
}

def run_detection(input_path, output_dir="static/results", model_path="models/best.pt"):
    """对单张红外图像进行极小目标检测"""
    model = YOLO(model_path)
    img = cv2.imread(input_path)
    if img is None:
        raise ValueError("无法读取图像，请检查路径")
    
    # 设置置信度阈值为 0.4，可根据实际效果调整
    results = model(img, conf=0.4)
    annotated_img = img.copy()
    boxes = results[0].boxes
    class_counts = {name: 0 for name in CLASS_NAMES.values()}

    for box in boxes:
        x1, y1, x2, y2 = map(int, box.xyxy[0])
        conf = float(box.conf[0])
        cls_id = int(box.cls[0])
        
        if cls_id not in CLASS_NAMES:
            continue
        
        class_name = CLASS_NAMES[cls_id]
        class_counts[class_name] += 1
        
        # 绘制边界框和标签
        color = (0, 255, 0)
        cv2.rectangle(annotated_img, (x1, y1), (x2, y2), color, 2)
        label = f"{class_name} {conf:.2f}"
        cv2.putText(annotated_img, label, (x1, y1 - 10), cv2.FONT_HERSHEY_SIMPLEX, 0.5, color, 1)

    os.makedirs(output_dir, exist_ok=True)
    filename = os.path.basename(input_path)
    output_path = os.path.join(output_dir, f"det_{filename}")
    cv2.imwrite(output_path, annotated_img)
    return output_path, class_counts

# app.py
from flask import Flask, render_template, request, redirect, url_for
import os
from detect import run_detection

app = Flask(__name__)
UPLOAD_FOLDER = 'static/uploads'
RESULT_FOLDER = 'static/results'
ALLOWED_EXTENSIONS = {'png', 'jpg', 'jpeg'}

app.config['UPLOAD_FOLDER'] = UPLOAD_FOLDER
app.config['RESULT_FOLDER'] = RESULT_FOLDER

def allowed_file(filename):
    return '.' in filename and filename.rsplit('.', 1)[1].lower() in ALLOWED_EXTENSIONS

@app.route('/')
def index():
    return render_template('index.html')

@app.route('/upload', methods=['POST'])
def upload_file():
    if 'file' not in request.files:
        return redirect(request.url)
    file = request.files['file']
    if file.filename == '' or not allowed_file(file.filename):
        return redirect(request.url)

    filename = file.filename
    filepath = os.path.join(app.config['UPLOAD_FOLDER'], filename)
    os.makedirs(app.config['UPLOAD_FOLDER'], exist_ok=True)
    file.save(filepath)

    try:
        output_path, counts = run_detection(
            filepath, 
            output_dir=app.config['RESULT_FOLDER'], 
            model_path="models/best.pt"
        )
        result_filename = os.path.basename(output_path)
    except Exception as e:
        return f"<h2>检测失败：{str(e)}</h2>"

    non_zero_counts = {k: v for k, v in counts.items() if v > 0}
    return render_template('result.html', original=filename, detected=result_filename, counts=non_zero_counts)

if __name__ == '__main__':
    app.run(debug=True, host='0.0.0.0', port=5000)

<!DOCTYPE html>
<html>
<head><title>红外极小目标检测系统</title>
<style>
body{font-family: Arial;text-align: center;margin-top: 40px;}
.btn{padding: 8px 16px;margin: 5px;background: #007bff;color: white;border: none;cursor: pointer;}
.btn:hover{background: #0069d9;}
</style>
</head>
<body>
<h2>🔍 无人机红外极小目标检测系统</h2>
<form method="post" enctype="multipart/form-data" action="/upload">
<input type="file" name="file" accept="image/*" required><br><br>
<button type="submit" class="btn">上传并检测</button>
</form>
</body>
</html>

<!DOCTYPE html>
<html>
<head><title>检测结果</title>
<style>
body{font-family: Arial;margin: 20px;}
img{max-width: 95%;margin: 10px;border: 1px solid #ddd;}
.stats{background: #f8f9fa;padding: 15px;border-radius: 5px;margin: 20px 0;}
.item{display: inline-block;margin-right: 15px;}
</style>
</head>
<body>
<h2>📊 检测结果</h2>
<img src="{{ url_for('static', filename='uploads/' + original) }}" alt="原始图">
<img src="{{ url_for('static', filename='results/' + detected) }}" alt="检测结果">
<div class="stats">
<strong>检测到的目标：</strong> 
{% for name, count in counts.items() %} <span class="item">{{ name }}: {{ count }} 处</span> {% endfor %} 
</div>
<a href="/">← 返回上传</a>
</body>
</html>

python app.py