ocr识别修改

This commit is contained in:
张松
2025-11-26 14:07:49 +08:00
parent 1dae19fb27
commit 8cd2bbf3d8
4 changed files with 401 additions and 323 deletions

View File

@@ -8,13 +8,17 @@ import com.alibaba.dashscope.app.ApplicationParam;
import com.alibaba.dashscope.app.ApplicationResult;
import com.alibaba.dashscope.app.RagOptions;
import com.aliyun.bailian20231229.Client;
import com.aliyun.bailian20231229.models.AddFileResponse;
import com.aliyun.bailian20231229.models.ApplyFileUploadLeaseResponse;
import com.aliyun.bailian20231229.models.ApplyFileUploadLeaseResponseBody;
import com.aliyun.bailian20231229.models.DescribeFileResponse;
import com.czg.exception.CzgException;
import lombok.extern.slf4j.Slf4j;
import org.springframework.stereotype.Component;
import java.io.InputStream;
import java.io.DataOutputStream;
import java.net.HttpURLConnection;
import java.net.URL;
import java.util.List;
import java.util.Map;
@@ -53,13 +57,14 @@ public class AliOcrUtil {
public static ApplyFileUploadLeaseResponseBody.ApplyFileUploadLeaseResponseBodyData applyFileUpload(InputStream stream, String fileName) {
String md5 = DigestUtil.md5Hex(stream);
public static ApplyFileUploadLeaseResponseBody.ApplyFileUploadLeaseResponseBodyData applyFileUpload(byte[] bytes, String fileName) {
String md5 = DigestUtil.md5Hex(bytes);
com.aliyun.bailian20231229.Client client = createClient();
com.aliyun.bailian20231229.models.ApplyFileUploadLeaseRequest applyFileUploadLeaseRequest = new com.aliyun.bailian20231229.models.ApplyFileUploadLeaseRequest()
.setFileName(fileName)
.setMd5(md5)
.setSizeInBytes("100000")
.setSizeInBytes(String.valueOf(bytes.length))
.setCategoryType("SESSION_FILE")
.setUseInternalEndpoint(false);
com.aliyun.teautil.models.RuntimeOptions runtime = new com.aliyun.teautil.models.RuntimeOptions();
@@ -76,40 +81,94 @@ public class AliOcrUtil {
}
}
public static String uploadFile(InputStream stream, String fileName) throws Exception {
public static String getSessionId(byte[] bytes, String fileName) {
String leaseId = null;
try {
leaseId = uploadFile(bytes, fileName);
} catch (Exception e) {
throw new RuntimeException(e);
}
com.aliyun.bailian20231229.Client client = createClient();
com.aliyun.bailian20231229.models.AddFileRequest addFileRequest = new com.aliyun.bailian20231229.models.AddFileRequest()
.setLeaseId(leaseId)
.setParser("DASHSCOPE_DOCMIND")
.setCategoryId("default")
.setCategoryType("SESSION_FILE");
com.aliyun.teautil.models.RuntimeOptions runtime = new com.aliyun.teautil.models.RuntimeOptions();
java.util.Map<String, String> headers = new java.util.HashMap<>();
try {
// 复制代码运行请自行打印 API 的返回值
AddFileResponse addFileResponse = client.addFileWithOptions("llm-9zg04s7wlbvi32tq", addFileRequest, headers, runtime);
String fileId = addFileResponse.getBody().getData().getFileId();
while (!getFileStatus(fileId)) {
Thread.sleep(100);
}
return fileId;
} catch (Exception error) {
throw new RuntimeException(error);
}
}
ApplyFileUploadLeaseResponseBody.ApplyFileUploadLeaseResponseBodyData applyInfo = applyFileUpload(stream, fileName);
public static boolean getFileStatus(String fileId) {
com.aliyun.bailian20231229.Client client = createClient();
com.aliyun.teautil.models.RuntimeOptions runtime = new com.aliyun.teautil.models.RuntimeOptions();
java.util.Map<String, String> headers = new java.util.HashMap<>();
try {
// 复制代码运行请自行打印 API 的返回值
DescribeFileResponse describeFileResponse = client.describeFileWithOptions("llm-9zg04s7wlbvi32tq", fileId, headers, runtime);
log.info("file status: {}", describeFileResponse.getBody());
return describeFileResponse.getBody().getData() != null && "FILE_IS_READY".equals(describeFileResponse.getBody().getData().getStatus());
} catch (Exception error) {
throw new RuntimeException(error);
}
}
public static String uploadFile(byte[] bytes, String fileName) throws Exception {
ApplyFileUploadLeaseResponseBody.ApplyFileUploadLeaseResponseBodyData applyInfo = applyFileUpload(bytes, fileName);
// 获取预签名要求的所有头
Map<String, String> headers = (Map<String, String>) applyInfo.getParam().getHeaders();
// 补充必须的 Content-Type
if (!headers.containsKey("Content-Type")) {
headers.put("Content-Type", "application/octet-stream");
}
HttpRequest request = HttpRequest.put(applyInfo.getParam().getUrl());
// 必须一模一样地设置所有 headers否则签名失效
// 设置所有 header
for (Map.Entry<String, String> entry : headers.entrySet()) {
request.header(entry.getKey(), entry.getValue());
}
// 设置文件内容
request.body(stream.readAllBytes());
request.body(bytes);
HttpResponse resp = request.execute();
log.info(resp.body());
log.info("upload status: {}", resp.getStatus());
if (resp.getStatus() != 200) {
throw new CzgException("OSS 上传失败: " + resp.body());
}
return applyInfo.getFileUploadLeaseId();
}
public static String appCall(InputStream stream, String fileName) {
public static String appCall(byte[] bytes, String fileName) {
String id = null;
try {
id = uploadFile(stream, fileName);
id = getSessionId(bytes, fileName);
} catch (Exception e) {
throw new RuntimeException(e);
}
ApplicationParam param = ApplicationParam.builder()
ApplicationParam param = ApplicationParam.builder()
.apiKey("sk-2343af4413834ad1ab43b036e3a903de")
.appId("3493340ef5e146c487364395fbca7bf3")
.prompt("开始处理")
.appId("cd612ac509a4499f8ac68a656532d4ae")
.prompt("你是一名票据OCR结构化专家请从我提供的票据图片中智能提取信息并只输出JSON不得添加解释、不补充不存在内容、不得返回空字符串、字段缺失填null、字段不可省略、数字一律用字符串使用以下固定JSON结构{\"documentType\":\"\",\"orderNumber\":\"\",\"date\":\"\",\"customerName\":\"\",\"operator\":\"\",\"items\":[{\"conName\":\"\",\"spec\":\"\",\"unitName\":\"\",\"inOutNumber\":\"\",\"purchasePrice\":\"\",\"subTotal\":\"\"}],\"totalAmount\":\"\",\"remark\":\"\"}。字段映射规则documentType对应单据类型/销售单/采购单/出货单orderNumber对应单号/编号/Nodate对应日期/开单日期customerName对应客户名称/收货单位/供应商operator对应业务员/经办人/制单人/操作员items.conName对应品名/名称items.spec对应规格/型号items.unitName对应单位items.inOutNumber对应数量items.purchasePrice对应单价items.subTotal对应金额/小计totalAmount对应总金额/合计金额remark对应备注。严禁生成图片中不存在的字段内容看不清或未出现的字段必须为null不允许推测或补全不得生成多余字段items只能根据识别到的行生成不得虚构。必须能识别旋转、倾斜、模糊、撕裂、光照差异、列顺序混乱、无表格线等情况并尽量恢复信息。最终输出必须是纯JSON不得包含任何非JSON字符。")
.ragOptions(RagOptions.builder()
.sessionFileIds(List.of(id))
.build())