cv::dnn::NMSBoxes 的非dnn版本

cv::dnn::NMSBoxes是OpenCV中深度学习模块（dnn）中提供的一种非极大值抑制（NMS）的函数，用于去除检测结果中的重复框。如果不使用OpenCV的深度学习模块，可以通过以下代码实现非dnn版本的NMSBoxes函数：

void NMSBoxes(const std::vector<cv::Rect>& boxes, const std::vector<float>& scores, float score_threshold, float nms_threshold, std::vector<int>& indices) {
    // 将boxes和scores打包成一个结构体数组
    struct BoxWithScore {
        cv::Rect box;
        float score;
    };
    std::vector<BoxWithScore> boxes_with_scores;
    for (int i = 0; i < boxes.size(); i++) {
        boxes_with_scores.push_back({ boxes[i], scores[i] });
    }

    // 按照得分从高到低排序
    std::sort(boxes_with_scores.begin(), boxes_with_scores.end(), [](const BoxWithScore& a, const BoxWithScore& b) {
        return a.score > b.score;
    });

    // 执行NMS
    indices.clear();
    for (int i = 0; i < boxes_with_scores.size(); i++) {
        if (boxes_with_scores[i].score < score_threshold) {
            break;
        }
        bool keep = true;
        for (int j = 0; j < indices.size(); j++) {
            if (keep) {
                int overlap_area = (boxes_with_scores[i].box & boxes_with_scores[indices[j]].box).area();
                int union_area = boxes_with_scores[i].box.area() + boxes_with_scores[indices[j]].box.area() - overlap_area;
                keep = overlap_area * 1.0 / union_area <= nms_threshold;
            }
            else {
                break;
            }
        }
        if (keep) {
            indices.push_back(i);
        }
    }
}

以上代码仅供参考，实际的实现可能会有所不同，具体实现方式可能因模型和框架而异。