【阿里云高校计划】车辆保险应用 day4 【拨云见日】

【阿里云高校计划】车辆保险应用 day4 【拨云见日】

有一说一，在五天阿里云视觉训练营中，身为菜鸟的我只学会了调用外部API和使用maven，我很开心，感谢阿里。

这次只在后端实现，前后端实在不会关联。希望以后学成之后可以实现。

实施前准备工作

一、本地图片上传为OSS

1.开通oss

其实在上周就开通了，但是本人较笨，搞了半天没搞好，这周末终于有空重新搞起了！

2.创建Bucket

Bucket的配置我只改了名称和区域，区域是因为

二、开通目标检测服务

三、查看所需API

这里我们用到阿里云视觉智能开放平台提供的三个功能：

• 车辆部件识别
• 车辆损伤识别
• 车险图片分类

1.车辆部件识别

检测图片中车辆部件的位置以及名称。

2.车辆损伤识别

针对常见小汽车车型，识别车辆外观受损部件及损伤类型，可识别数十种车辆部件、五大类外观损伤。（刮擦、凹陷、开裂、褶皱、穿孔）

3.车险图片分类

对输入的车险图片进行分类。

具体实施

一、本地图片上传至OSS的uploadPic类

1.在maven中导入所需依赖

<dependency>
    <groupId>com.aliyun.oss</groupId>
    <artifactId>aliyun-sdk-oss</artifactId>
    <version>3.8.0</version>
</dependency>

2.编写UploadPic类

package com.example.demo;

import com.aliyun.oss.OSS;
import com.aliyun.oss.OSSClientBuilder;

import java.io.File;
import java.net.URL;
import java.security.SecureRandom;
import java.util.Date;
import java.lang.*;
import java.util.Scanner;

public class UploadPic {
    public static String UploadPic(){
        // Endpoint以杭州为例，其它Region请按实际情况填写。
        String endpoint = "oss-cn-shanghai.aliyuncs.com";
        // 阿里云主账号AccessKey。
        String accessKeyId = "*************";
        String accessKeySecret = "*************";
        //本地文件名
        System.out.println("请输入本地图片path:");
        Scanner scanner = new Scanner(System.in);
        String fileName = scanner.nextLine();
        String bucketName = "auto-insurance-pic";
        // 获取文件的后缀名
        String suffixName = fileName.substring(fileName.lastIndexOf("."));

        // 生成上传文件名
        String objectName = System.currentTimeMillis() + "" + new SecureRandom().nextInt(0x0400) + suffixName;


        // 创建OSSClient实例。
        OSS ossClient = new OSSClientBuilder().build(endpoint, accessKeyId, accessKeySecret);


        // 如果需要上传时设置存储类型与访问权限，请参考以下示例代码。
        // ObjectMetadata metadata = new ObjectMetadata();
        // metadata.setHeader(OSSHeaders.OSS_STORAGE_CLASS, StorageClass.Standard.toString());
        // metadata.setObjectAcl(CannedAccessControlList.Private);
        // putObjectRequest.setMetadata(metadata);

        // 上传文件。
        ossClient.putObject(bucketName, objectName, new File(fileName));
        // 设置URL过期时间为1小时。
        Date expiration = new Date(System.currentTimeMillis() + 3600 * 1000);
        // 生成以GET方法访问的签名URL，访客可以直接通过浏览器访问相关内容。
        URL url = ossClient.generatePresignedUrl(bucketName, objectName, expiration);

        // 关闭OSSClient。
        ossClient.shutdown();

        return url.toString();
    }
}

3.运行结果

运行成功！

二、车辆部件识别RecognizeVehicleParts类

1.在maven中导入所需依赖

因为属于目标检测，而根据文档中的导入指令有错，所以我来到了阿里的maven仓库下查找并选择了最新的版本进行导入。

阿里maven私有仓库服务：https://maven.aliyun.com/mvn/search

<dependency>
	<groupId>com.aliyun</groupId>
	<artifactId>aliyun-java-sdk-core</artifactId>
	<version>4.4.8</version>
</dependency>
<dependency>
	<groupId>com.alibaba</groupId>
	<artifactId>fastjson</artifactId>
	<version>1.2.52</version>
</dependency>
<dependency>
	<groupId>com.aliyun</groupId>
	<artifactId>aliyun-java-sdk-objectdet</artifactId>
	<version>1.0.7</version>
</dependency>

2.编写RecognizeVehicleParts类

package com.example.demo;

import com.aliyuncs.DefaultAcsClient;
import com.aliyuncs.IAcsClient;
import com.aliyuncs.exceptions.ClientException;
import com.aliyuncs.exceptions.ServerException;
import com.aliyuncs.profile.DefaultProfile;
import com.example.demo.UploadPic;
import com.google.gson.Gson;
import java.util.*;
import com.aliyuncs.objectdet.model.v20191230.*;

public class RecognizeVehicleParts {

    public static void main(String[] args) {
        DefaultProfile profile = DefaultProfile.getProfile("cn-shanghai", "accessKeyId", "accessKeySecret");
        IAcsClient client = new DefaultAcsClient(profile);

        RecognizeVehiclePartsRequest request = new RecognizeVehiclePartsRequest();
        request.setRegionId("cn-shanghai");
        request.setImageURL(UploadPic.UploadPic());

        try {
            RecognizeVehiclePartsResponse response = client.getAcsResponse(request);
            System.out.println(new Gson().toJson(response));
        } catch (ServerException e) {
            e.printStackTrace();
        } catch (ClientException e) {
            System.out.println("ErrCode:" + e.getErrCode());
            System.out.println("ErrMsg:" + e.getErrMsg());
            System.out.println("RequestId:" + e.getRequestId());
        }
    }
}

2.运行结果

返回值为：

{"requestId":"BBFB102D-5EAC-483F-9A94-9DA79A06E1F6","data":{"elements":[{"score":0.98788995,"type":"left_tail_light","boxes":[132,274,862,607]},{"score":0.952229,"type":"left_rear_wing","boxes":[4,162,365,750]},{"score":0.74785864,"type":"rear_bumper","boxes":[60,456,987,760]}],"originShapes":[768,1024]}}

返回部件类型为：left_tail_light 左后灯，left_rear_wing 左后翼，rear_bumper尾部保险杠

识别成功！（其实我是想识别左后灯，结果连其他两个也识别出来了！）

再来一张！！

返回值为：

{"requestId":"61E13B25-82AD-4E4C-A133-ADA281C1A4EE","data":{"elements":[{"score":0.9911305,"type":"right_front_door","boxes":[1,0,350,559]},{"score":0.9898079,"type":"right_front_tire","boxes":[335,191,727,656]},{"score":0.98655057,"type":"right_front_wing","boxes":[296,2,749,490]},{"score":0.97444427,"type":"right_doorsill","boxes":[1,497,359,651]},{"score":0.9604019,"type":"front_bumper","boxes":[671,168,779,465]}],"originShapes":[768,1024]}}

返回部件类型为：right_front_door 右前门，right_front_tire 右前轮，right_front_wing右前翼，right_doorsill右底大边，front_bumper前保险杠

识别正确！超乎预期

附文档的type说明

Type说明

名字	英文名字	部件编码	标注编码	是否是34部件	是否计划支持
前保险杠	front_bumper	PR01	11	是	是
后保险杠	rear_bumper	PR02	13	是	是
左前大灯	left_light	PR03	71	是	是
右前大灯	right_light	PR04	72	是	是
中网	grille	PR05	b11	是	是
中网	grille	PR05	B11	是	是
前机盖	hood	PR06	21	是	是
左前门	left_front_door	PR07	51	是	是
右前门	right_front_door	PR08	52	是	是
左后门	left_rear_door	PR09	53	是	是
右后门	right_rear_door	PR010	54	是	是
左前翼子板	left_front_wing	PR11	41	是	是
右前翼子板	right_front_wing	PR12	42	是	是
左后翼子板	left_rear_wing	PR13	43	是	是
右后翼子板	right_rear_wing	PR14	44	是	是
后机盖	decklid	PR015	23	是	是
前挡风玻璃	front_windshield	PR16	61	是	是
后挡风玻璃	rear_windshield	PR17	63	是	是
左后视镜	left_mirror	PR18	93	是	是
右后视镜	right_mirror	PR19	94	是	是
左尾灯	left_tail_light	PR20	73	是	是
右尾灯	right_tail_light	PR21	74	是	是
左雾灯	left_foglight	PR22	b13	是	是
左雾灯	left_foglight	PR22	B13	是	是
右雾灯	right_foglight	PR23	b14	是	是
右雾灯	right_foglight	PR23	B14	是	是
格栅	grates	PR24	b12	是	是
格栅	grates	PR24	B12	是	是
右前车窗	right_front_window	PR25	82	是	是
左前车窗	left_front_window	PR26	81	是	是
右后车窗	right_rear_window	PR27	84	是	是
左后车窗	right_tail_light	PR28	83	是	是
右底大边	right_doorsill	PR29	92	是	是
左底大边	left_doorsill	PR30	91	是	是
右前门把手	right_front_doorknob	PR31	502	是	是
左前门把手	left_front_doorknob	PR32	501	是	是
左后门把手	left_rear_doorknob	PR34	503	是	是
右后门把手	right_tail_light	PR33	504	是	是
左前轮胎	left_front_tire	PR3601	31	是	是
左后轮胎	left_rear_tire	PR3602	33	是	是
右前轮胎	right_front_tire	PR3603	32	是	是
右后轮胎	right_tail_light	PR3604	34	是	是
轮胎	right_tail_light	PR36	3	是	是
右前轮眉	right_front_WheelBrow	PR1203	352	否	是
左前轮眉	left_front_WheelBrow	PR1103	351	否	是
右后轮眉	right_rear_WheelBrow	PR1403	354	否	是
左后轮眉	left_rear_WheelBrow	PR1303	353	否	是
左前钢圈	left_front_ring	PR3501	316	否	是
左后钢圈	left_rear_ring	PR3502	336	否	是
右前钢圈	right_tail_light	PR3503	326	否	是
右后钢圈	right_rear_ring	PR3504	346	否	是
钢圈	ring	PR35	306	否	是
左前门饰条	left_front_door_Panel	PR0701	6511	否	是
右前门饰条	right_front_door_Panel	PR0801	6512	否	是
左后门饰条	left_rear_door_Panel	PR0901	6513	否	是
右后门饰条	right_rear_door_Panel	PR1001	6514	否	是
左前门亮条	left_front_door_Wisp	PR0701	6521	否	是
右前门亮条	right_front_door_Wisp	PR0801	6522	否	是
左后门亮条	left_rear_door_Wisp	PR0901	6523	否	是
右后门亮条	right_rear_door_Wisp	PR1001	6524	否	是
左前门饰板	left_front_door_Plaque	PR0703	166	否	是
右前门饰板	right_front_door_Plaque	PR0803	266	否	是
左后门饰板	left_rear_door_Plaque	PR0903	366	否	是
右后门饰板	right_rear_door_Plaque	PR1003	466	否	是
前保险杠饰条	front_bumper_Panel	PR0101	1651	否	是
后保险杠饰条	rear_bumper_Panel	PR0201	3651	否	是
前保险杠亮条	front_bumper_Wisp	PR0102	1652	否	是
后保险杠亮条	rear_bumper_Wisp	PR0202	3652	否	是
前保险杠护板	front_bumper_Backplate	PR0103	6621	否	是
后保险杠护板	rear_bumper_Backplate	PR0203	6623	否	是
前保险杠导流板	front_bumper_Deflector	PR0104	6611	否	是
后保险杠导流板	rear_bumper_Deflector	PR0204	6613	否	是
后保反光板	BumperLight	PR0205	761	否	是
车顶	Roof	待定	25	否	否
A柱	Apillar	待定	26	否	否
挡泥板	Fender	待定	36	否	否
反光镜灯	MirrorLight	待定	762	否	否
翼子板灯	WingLight	待定	763	否	否
牌照灯	PlateLight	待定	764	否	否
车窗三角玻璃	TriangleWindow	待定	85	否	否
天窗	RoofWindow	待定	86	否	否
字标	WordMark	待定	67	否	否
牵引钩盖板	DragCover	待定	681	否	否
喷水嘴盖板	LightWaterCover	待定	682	否	否
电眼	ElectricEye	待定	683	否	否
油箱盖	TankCover	待定	684	否	否
导流板（无方向）	Deflector	待定	661	否	否
门把手（无方向）	Deflector	待定	50	否	否
车标	logo	待定	b16,B16	否	否
饰板（无方向）	Plaque	待定	66	否	否
保险杠（无方向）	Roof	待定	1	否	否
车牌	plate	待定	b17,B17	否	否
雾灯（无方向）	foglight	待定	b15,B15	否	否
亮条（无方向）	Wisp	待定	652	否	否
翼子板（无方向）	wing	待定	4	否	否
饰条（无方向）	Panel	待定	651	否	否
底大边（无方向）	doorsill	待定	9	否	否
门（无方向）	door	待定	5	否	否
挡风玻璃	windshield	待定	6	否	否
大灯（无方向）	light	待定	70	否	否
挡泥板	Fender	待定	36	否	否
轮眉（无方向）	WheelBrow	待定	35	否	否
看不出什么部件	无	待定	00	无	无

三、车辆损坏识别类

1.在maven中导入所需依赖

见本文二-1

2.编写RecognizeVehicleDamage类

package com.example.demo;

import com.aliyuncs.DefaultAcsClient;
import com.aliyuncs.IAcsClient;
import com.aliyuncs.exceptions.ClientException;
import com.aliyuncs.exceptions.ServerException;
import com.aliyuncs.profile.DefaultProfile;
import com.google.gson.Gson;
import java.util.*;
import com.aliyuncs.objectdet.model.v20191230.*;

public class RecognizeVehicleDamage {

    public static void main(String[] args) {
        DefaultProfile profile = DefaultProfile.getProfile("cn-shanghai", "accessKeyId", "accessKeySecret");
        IAcsClient client = new DefaultAcsClient(profile);

        RecognizeVehicleDamageRequest request = new RecognizeVehicleDamageRequest();
        request.setRegionId("cn-shanghai");
        request.setImageURL(UploadPic.UploadPic());

        try {
            RecognizeVehicleDamageResponse response = client.getAcsResponse(request);
            System.out.println(new Gson().toJson(response));
        } catch (ServerException e) {
            e.printStackTrace();
        } catch (ClientException e) {
            System.out.println("ErrCode:" + e.getErrCode());
            System.out.println("ErrMsg:" + e.getErrMsg());
            System.out.println("RequestId:" + e.getRequestId());
        }
    }
}

3.识别结果

返回值为：

{"requestId":"7FFBD390-7019-4B85-9FFA-779C912A9CEB","data":{"elements":[{"score":0.414995,"type":"1","scores":[0.414995,0.0,0.0,0.0,0.0,0.0,0.0,0.0,0.0,0.0,0.0,0.0,0.0,0.0,0.0],"boxes":[343,390,473,542]},{"score":0.408405,"type":"1","scores":[0.408405,0.0,0.0,0.0,0.0,0.0,0.0,0.0,0.0,0.0,0.0,0.0,0.0,0.0,0.0],"boxes":[541,442,659,545]},{"score":0.348472,"type":"1","scores":[0.348472,0.0,0.0,0.0,0.0,0.0,0.0,0.0,0.0,0.0,0.0,0.0,0.0,0.0,0.0],"boxes":[273,293,423,400]},{"score":0.378637,"type":"2","scores":[0.0,0.378637,0.0,0.0,0.0,0.0,0.0,0.0,0.0,0.0,0.0,0.0,0.0,0.0,0.0],"boxes":[261,26,496,142]},{"score":0.873101,"type":"5","scores":[0.0,0.0,0.0,0.0,0.873101,0.0,0.0,0.0,0.0,0.0,0.0,0.0,0.0,0.0,0.0],"boxes":[91,4,555,267]},{"score":0.815785,"type":"6","scores":[0.0,0.0,0.0,0.0,0.0,0.815785,0.0,0.0,0.0,0.0,0.0,0.0,0.0,0.0,0.0],"boxes":[564,270,869,441]},{"score":0.845525,"type":"8","scores":[0.0,0.0,0.0,0.0,0.0,0.0,0.0,0.845525,0.0,0.0,0.0,0.0,0.0,0.0,0.0],"boxes":[230,234,529,313]},{"score":0.411336,"type":"11","scores":[0.0,0.0,0.0,0.0,0.0,0.0,0.0,0.0,0.0,0.0,0.411336,0.0,0.0,0.0,0.0],"boxes":[632,425,922,563]},{"score":0.334054,"type":"11","scores":[0.0,0.0,0.0,0.0,0.0,0.0,0.0,0.0,0.0,0.0,0.334054,0.0,0.0,0.0,0.0],"boxes":[538,91,733,194]},{"score":0.333818,"type":"11","scores":[0.0,0.0,0.0,0.0,0.0,0.0,0.0,0.0,0.0,0.0,0.333818,0.0,0.0,0.0,0.0],"boxes":[694,157,899,286]},{"score":0.32519,"type":"11","scores":[0.0,0.0,0.0,0.0,0.0,0.0,0.0,0.0,0.0,0.0,0.32519,0.0,0.0,0.0,0.0],"boxes":[523,87,902,278]}]}}

Boxes	List	173,178,277,259	部件位置框信息，形式为【左上角点x坐标，右下角点y坐标，右上角点x坐标，左下角点y坐标】。
Score	Float	0.683465	损伤类型对应的概率值。
Scores	List	0.683465	对应所有15种损伤类型的损伤概率。
Type	String	1	损伤类型id。具体Type类型如下所示。1：轻微刮擦2：重度刮擦3：轻度变形4：中度变形5：重度变形6：crack破损孔洞7：翼子板和大灯缝隙8：翼子板保险杠缝隙9：大灯轻微刮擦10：大灯重度刮擦11：大灯破损12：后视镜轻微刮擦13：后视镜玻璃破损14：后视镜脱落15：挡风玻璃破损

识别成功！

四、车险图片分类

1.在maven中导入所需依赖

见本文二-1

2.编写ClassifyVehicleInsurance类

package com.example.demo;

import com.aliyuncs.DefaultAcsClient;
import com.aliyuncs.IAcsClient;
import com.aliyuncs.exceptions.ClientException;
import com.aliyuncs.exceptions.ServerException;
import com.aliyuncs.profile.DefaultProfile;
import com.google.gson.Gson;
import java.util.*;
import com.aliyuncs.objectdet.model.v20191230.*;

public class ClassifyVehicleInsurance {

    public static void main(String[] args) {
        DefaultProfile profile = DefaultProfile.getProfile("cn-shanghai", "accessKeyId", "accessKeySecret");
        IAcsClient client = new DefaultAcsClient(profile);

        ClassifyVehicleInsuranceRequest request = new ClassifyVehicleInsuranceRequest();
        request.setRegionId("cn-shanghai");
        request.setImageURL(UploadPic.UploadPic());

        try {
            ClassifyVehicleInsuranceResponse response = client.getAcsResponse(request);
            System.out.println(new Gson().toJson(response));
        } catch (ServerException e) {
            e.printStackTrace();
        } catch (ClientException e) {
            System.out.println("ErrCode:" + e.getErrCode());
            System.out.println("ErrMsg:" + e.getErrMsg());
            System.out.println("RequestId:" + e.getRequestId());
        }
    }
}

3.识别结果

返回值为：

{"requestId":"87BEADED-F581-4C38-9F5F-E6F8DF0A1BA5","data":{"threshold":0.0,"labels":[{"score":0.0046,"name":"others"},{"score":0.0164,"name":"detail"},{"score":0.1934,"name":"component"},{"score":0.0,"name":"vin"},{"score":8.0E-4,"name":"people"},{"score":2.0E-4,"name":"motor"},{"score":0.1439,"name":"semi-car"},{"score":0.0027,"name":"panoramic"},{"score":3.0E-4,"name":"license"},{"score":0.0169,"name":"CT-scan"},{"score":5.0E-4,"name":"truck"},{"score":0.0144,"name":"disassembly"},{"score":0.6059,"name":"scene"}]}}

Labels	Array		输出分类结果。
Name	String	others	label名称。枚举类型。枚举值包括： others：其他 detail：细节图 component：汽车部件图 vin：汽车vin码 people：人物 motor：发动机 semi-car：半车图 panoramic：全车图 license：行驶证 CT-scan：CT-扫描 truck：卡车 disassembly：拆解件 scene：现场图
Score	Float	0.0023	分类结果对应的概率值。

实施完成

完成之后我会上传源码

加入高校计划

本人是参加的达摩院特别版-视觉AI训练营第二期
训练营里面的小哥哥小姐姐说话超级好听，也有超多大佬，我超喜欢这里！