Spring Boot和Elasticsearch进行人脸数据的高效检索

import org.elasticsearch.index.query.QueryBuilders;
import org.elasticsearch.search.builder.SearchSourceBuilder;
import org.springframework.beans.factory.annotation.Autowired;
import org.springframework.data.domain.Page;
import org.springframework.data.domain.PageRequest;
import org.springframework.data.elasticsearch.core.ElasticsearchRestTemplate;
import org.springframework.data.elasticsearch.core.SearchHit;
import org.springframework.data.elasticsearch.core.mapping.IndexCoordinates;
import org.springframework.data.elasticsearch.core.query.NativeSearchQueryBuilder;
import org.springframework.stereotype.Service;
 
@Service
public class FaceSearchService {
 
    @Autowired
    private ElasticsearchRestTemplate elasticsearchRestTemplate;
 
    public Page<Face> searchFaces(float[] vector, String collectionId, int page, int size) {
        // 构建向量查询
        SearchSourceBuilder searchSourceBuilder = new SearchSourceBuilder();
        searchSourceBuilder.query(QueryBuilders.scriptScoreQuery(
                QueryBuilders.matchAllQuery(),
                new ScoreScript.Builder("mscore", ScoreScript.DEFAULT_CONTEXT)
                        .lang("painless")
                        .param("vector", vector)
                        .build()
        ));
 
        // 设置分页参数
        int from = PageRequest.of(page, size).getOffset();
        searchSourceBuilder.from(from);
        searchSourceBuilder.size(size);
 
        // 执行搜索
        NativeSearchQueryBuilder queryBuilder = new NativeSearchQueryBuilder();
        queryBuilder.withSourceFilter(new FetchSourceFilter(new String[]{"id"}, null));
        queryBuilder.withQuery(searchSourceBuilder.query());
        queryBuilder.withSort(new FieldSortBuilder("_score").order(SortOrder.DESC));
        queryBuilder.withPageable(PageRequest.of(page, size));
 
        // 执行搜索并转换结果
        Page<SearchHit<Face>> searchHits = elasticsearchRestTemplate.search(queryBuilder.build(), Face.class, IndexCoordinates.of(collectionId));
        List<Face> faces = searchHits.stream()
                .map(SearchHit::getContent)
                .collect(Collectors.toList());
 
        return new PageImpl<>(faces, PageRequest.of(page, size), searchHits.getTotalElements());
    }
}

这段代码展示了如何在Spring Boot应用中使用ElasticsearchRestTemplate来执行基于向量相似度的人脸数据搜索。其中Face类应该包含一个id属性，以便于之后的查询。ScoreScript是假设已经定义在Elasticsearch中的一个脚本，用于计算查询向量和文档向量之间的相似度分数。这个