← Back to Skills Marketplace
45
Downloads
0
Stars
0
Active Installs
1
Versions
Install in OpenClaw
/install jf-open-pro-device-human-detection
Description
杰峰设备人形检测技能(开发版)。支持人形检测开关设置、人形检测灵敏度设置、云台设备人形追踪开关设置等功能。
README (SKILL.md)
jf-open-pro-device-human-detection - 杰峰设备人形检测技能(开发版)
技能描述
支持杰峰设备的人形检测和追踪功能,基于杰峰开放平台 OpenAPI 实现:
- 人形检测开关设置 - 开启/关闭人形检测报警
- 人形检测灵敏度设置 - 调节检测灵敏度(低/中/高)
- 人形追踪开关设置 - 开启/关闭云台自动追踪人形
- 人形追踪灵敏度设置 - 调节追踪灵敏度
- 追踪返回时间设置 - 设置无人时返回守望位的时间
触发词
- 查询人形检测配置 / 设置人形检测开关 / 开启人形检测 / 关闭人形检测
- 设置人形灵敏度 / 查询人形追踪配置 / 设置人形追踪开关
- 开启人形追踪 / 关闭人形追踪 / 人形跟踪设置
前置条件
必需配置
- 签名算法 - 使用杰峰官方移位加密算法生成 signature
- 时间戳算法 - counter(7 位) + timeMillis(13 位),实时生成
- 设备绑定 - 设备需先绑定到开放平台账号
- 云台设备 - 人形追踪功能需要云台设备支持
环境变量
| 变量名 | 说明 | 默认值 | 必需 |
|---|---|---|---|
JF_UUID |
开放平台用户 uuid | - | ✅ |
JF_APP_KEY |
开放平台应用 appKey | - | ✅ |
JF_APP_SECRET |
开放平台应用密钥 | - | ✅ |
JF_MOVE_CARD |
移动卡标识(用于签名) | 2 |
✅ |
JF_DEVICE_SN |
设备序列号 | - | ✅ |
JF_DEVICE_TOKEN |
设备接口访问令牌 | - | ✅ |
JF_ENDPOINT |
API 接入地址 | api-cn.jftechws.com |
❌ |
API 接口
| 功能 | 地址 | 方法 | 需要 Token | 需要在线 |
|---|---|---|---|---|
| 获取人形检测配置 | POST /gwp/v3/rtc/device/getconfig/{token} |
POST | ✅ | ✅ |
| 设置人形检测配置 | POST /gwp/v3/rtc/device/setconfig/{token} |
POST | ✅ | ✅ |
| 获取人形追踪配置 | POST /gwp/v3/rtc/device/getconfig/{token} |
POST | ✅ | ✅ |
| 设置人形追踪配置 | POST /gwp/v3/rtc/device/setconfig/{token} |
POST | ✅ | ✅ |
核心功能
1. 人形检测配置(Detect.HumanDetection)
配置项说明:
| 字段 | 类型 | 说明 | 取值 |
|---|---|---|---|
Enable |
boolean | 人形检测开关 | true=开启,false=关闭 |
Sensitivity |
int | 检测灵敏度 | 0=低,1=中,2=高,3=灵敏度数量 |
PedFdrAlg |
int | 人形人脸算法类型 | 0=单人形,1=人形 + 人脸,2=人形 + 人脸识别,3=人形 + 车形,4=人形 + 车形 + 人脸,5=宠物 |
ObjectType |
int | 检测目标类型 | 0=检测人,1=检测物体 |
ShowRule |
boolean | 是否叠加人形规则框 | true=是,false=否 |
ShowTrack |
boolean | 是否叠加移动人形框 | true=是,false=否 |
PushInterval |
int | 单人脸推图间隔(毫秒) | -1=不推图,其他=间隔时间 |
警戒规则(PedRule):
| 字段 | 类型 | 说明 |
|---|---|---|
Enable |
boolean | 规则是否使能 |
RuleType |
int | 规则类型(0=警戒线,1=警戒区域) |
RuleRegion |
object | 警戒区域参数 |
RuleLine |
object | 警戒线参数 |
警戒区域参数:
PtsNum: 区域点数(3-8)AlarmDirect: 检测方向(0=进入,1=离开,2=双向)Pts: 坐标点数组(0-8192)Sensitivity: 该区域灵敏度
2. 人形追踪配置(Detect.DetectTrack)
配置项说明:
| 字段 | 类型 | 说明 | 取值 |
|---|---|---|---|
Enable |
int | 人形追踪开关 | 0=关闭,1=开启 |
Sensitivity |
int | 追踪灵敏度 | 0=低,1=中,2=高 |
ReturnTime |
int | 返回默认位置时间(秒) | 0=不返回,1-600=指定时间 |
注意: 人形追踪不是报警,只是云台自动跟随人形移动。画面需要正放,识别出人形才能生效。
使用示例
环境准备
# 设置环境变量
export JF_UUID="uuidxxxx"
export JF_APP_KEY="appkeyxxxx"
export JF_APP_SECRET="appsecretxxxx"
export JF_MOVE_CARD=0
export JF_DEVICE_SN="snxxx1"
export JF_DEVICE_TOKEN="NTQ0NzQ3YmE3MXwyYzFk..."
export JF_ENDPOINT="api-cn.jftechws.com"
1. 查询人形检测配置
cd ~/.openclaw/workspace/skills/developer/jf-open-pro-device-human-detection/scripts
python3 human_detection.py --action get-human-detect-config
2. 开启/关闭人形检测
# 开启人形检测
python3 human_detection.py --action set-human-detect-switch --enable true
# 关闭人形检测
python3 human_detection.py --action set-human-detect-switch --enable false
3. 设置人形检测灵敏度
# 设置低灵敏度
python3 human_detection.py --action set-human-detect-sensitivity --level 0
# 设置中灵敏度
python3 human_detection.py --action set-human-detect-sensitivity --level 1
# 设置高灵敏度
python3 human_detection.py --action set-human-detect-sensitivity --level 2
4. 查询人形追踪配置
python3 human_detection.py --action get-human-track-config
5. 开启/关闭人形追踪
# 开启人形追踪
python3 human_detection.py --action set-human-track-switch --enable true
# 关闭人形追踪
python3 human_detection.py --action set-human-track-switch --enable false
6. 设置人形追踪灵敏度
# 设置低灵敏度
python3 human_detection.py --action set-human-track-sensitivity --level 0
# 设置中灵敏度
python3 human_detection.py --action set-human-track-sensitivity --level 1
# 设置高灵敏度
python3 human_detection.py --action set-human-track-sensitivity --level 2
7. 设置追踪返回时间
# 设置 10 秒后返回
python3 human_detection.py --action set-track-return-time --seconds 10
# 设置不返回
python3 human_detection.py --action set-track-return-time --seconds 0
# 设置 5 分钟后返回(300 秒)
python3 human_detection.py --action set-track-return-time --seconds 300
灵敏度说明
人形检测灵敏度
| 级别 | 值 | 说明 | 适用场景 |
|---|---|---|---|
| 低灵敏度 | 0 | 检测较宽松,误报少 | 人员流动频繁区域 |
| 中灵敏度 | 1 | 平衡检测和误报 | 一般区域(默认) |
| 高灵敏度 | 2 | 检测更敏感,易触发 | 重要安防区域 |
人形追踪灵敏度
| 级别 | 值 | 说明 | 适用场景 |
|---|---|---|---|
| 低灵敏度 | 0 | 追踪较平缓 | 目标移动缓慢 |
| 中灵敏度 | 1 | 追踪速度适中 | 一般场景(默认) |
| 高灵敏度 | 2 | 追踪响应快 | 目标移动快速 |
人形人脸算法类型
| 类型 | 值 | 说明 |
|---|---|---|
| 单人形检测 | 0 | 仅检测人形(默认) |
| 人形加人脸检测 | 1 | 检测人形并检测人脸 |
| 人形加人脸识别 | 2 | 检测人形并识别人脸身份 |
| 人形加车形检测 | 3 | 检测人形和车辆 |
| 人形加车形加人脸 | 4 | 检测人形、车形和人脸 |
| 宠物检测 | 5 | 检测宠物 |
状态码
平台状态码
| code | 说明 | 处理建议 |
|---|---|---|
| 2000 | 成功 | - |
| 28007 | Header 参数错误 | 检查 uuid、appKey、timeMillis、signature |
| 40103 | 无效 Token | deviceToken 过期,重新获取 |
| 50000 | 服务器内部错误 | 联系杰峰技术支持 |
设备状态码(Ret)
| Ret | 说明 |
|---|---|
| 100 | 成功 |
| 106 | 用户名或密码错误 |
注意事项
- deviceToken 有效期 - 24 小时,过期需重新获取
- 设备在线要求 - 配置类操作需要设备在线
- 云台设备 - 人形追踪功能仅云台设备支持
- 画面正放 - 人形追踪需要画面正放才能生效
- 识别生效 - 需先识别出人形,追踪功能才能生效
- 坐标范围 - 警戒区域坐标需缩放到 0-8192 范围
- 北京时间 - 建议使用北京时间(UTC+8)进行时间查询
- 算法创建 - 多通道设备可能部分通道不创建算法(AlgoCreate 字段)
相关文件
| 文件 | 说明 |
|---|---|
SKILL.md |
技能文档 |
scripts/human_detection.py |
Python 执行脚本 |
scripts/crypto.py |
签名/时间戳加密工具(复用) |
参考文档
Usage Guidance
Install only if you operate the target JFTech devices and are comfortable managing surveillance-related settings. Use only official JFTech regional endpoints, keep JF credentials and device tokens private, confirm the device serial/token before running set actions, and consider reading the current configuration before making changes.
Capability Tags
Capability Assessment
Purpose & Capability
The documented purpose matches the implementation: it queries and changes JFTech device human-detection and PTZ tracking configuration via OpenAPI.
Instruction Scope
The examples directly enable, disable, and tune live camera monitoring/tracking behavior, but the documentation does not clearly warn users that these commands immediately alter device behavior or raise privacy and consent issues.
Install Mechanism
The artifact contains a SKILL.md and two Python scripts; there is no installer, package dependency list, background worker, or automatic execution path.
Credentials
The skill requires JF OpenAPI credentials and a device token, then builds the API host from JF_ENDPOINT without enforcing the documented JFTech regional hostnames, so misconfiguration or environment manipulation could redirect signed requests and device tokens.
Persistence & Privilege
There is no local persistence or privilege escalation, but successful setconfig calls persistently modify remote device detection/tracking settings.
How to Use
- Make sure OpenClaw is installed (local or Docker)
- Run the install command in chat:
/install jf-open-pro-device-human-detection - After installation, invoke the skill by name or use
/jf-open-pro-device-human-detection - Provide required inputs per the skill's parameter spec and get structured output
Version History
v1.0.0
Initial release of jf-open-pro-device-human-detection.
- Supports enabling/disabling human detection and adjusting detection sensitivity on JFTech devices.
- Adds configuration and control for PTZ device human tracking, including tracking sensitivity and return time settings.
- Provides commands to view and modify human detection/tracking settings via CLI scripts.
- Requires environment variables for JFTech OpenAPI authentication and device information.
- Includes detailed documentation and usage examples in SKILL.md.
Metadata
Frequently Asked Questions
What is jf-open-pro-device-human-detection?
杰峰设备人形检测技能(开发版)。支持人形检测开关设置、人形检测灵敏度设置、云台设备人形追踪开关设置等功能。 It is an AI Agent Skill for Claude Code / OpenClaw, with 45 downloads so far.
How do I install jf-open-pro-device-human-detection?
Run "/install jf-open-pro-device-human-detection" in the OpenClaw or Claude Code chat to install it in one step — no extra setup required.
Is jf-open-pro-device-human-detection free?
Yes, jf-open-pro-device-human-detection is completely free, licensed under MIT-0. You can download, install and use it at no cost.
Which platforms does jf-open-pro-device-human-detection support?
jf-open-pro-device-human-detection is cross-platform and runs anywhere OpenClaw / Claude Code is available (cross-platform).
Who created jf-open-pro-device-human-detection?
It is built and maintained by jftech (@jftech); the current version is v1.0.0.
More Skills