From a0a7f6930af6b8a391b876ee46b5c217eb23c1d5 Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?=E9=BE=99=E6=BE=B3?= Date: Mon, 6 Jan 2025 15:50:11 +0800 Subject: [PATCH] =?UTF-8?q?=E7=A3=81=E7=9B=98=E7=A9=BA=E9=97=B4=E3=80=81?= =?UTF-8?q?=E5=AE=B9=E9=94=99=E4=BC=98=E5=8C=96?= MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit --- odm_preprocess.py | 82 ++++++++++++++++++----- requirements.txt | 2 +- utils/odm_monitor.py | 151 ++++++++++++++++++++++++++----------------- 3 files changed, 158 insertions(+), 77 deletions(-) diff --git a/odm_preprocess.py b/odm_preprocess.py index 949c063..de8b891 100644 --- a/odm_preprocess.py +++ b/odm_preprocess.py @@ -3,6 +3,7 @@ import shutil from datetime import timedelta from dataclasses import dataclass from typing import Dict, Tuple +import psutil # 需要添加到 requirements.txt import matplotlib.pyplot as plt import pandas as pd @@ -53,6 +54,9 @@ class ImagePreprocessor: def __init__(self, config: PreprocessConfig): self.config = config + # 检查磁盘空间 + self._check_disk_space() + # 清理并重建输出目录 if os.path.exists(config.output_dir): self._clean_output_dir() @@ -91,6 +95,45 @@ class ImagePreprocessor: print(f"创建输出目录时发生错误: {str(e)}") raise + def _get_directory_size(self, path): + """获取目录的总大小（字节）""" + total_size = 0 + for dirpath, dirnames, filenames in os.walk(path): + for filename in filenames: + file_path = os.path.join(dirpath, filename) + try: + total_size += os.path.getsize(file_path) + except (OSError, FileNotFoundError): + continue + return total_size + + def _check_disk_space(self): + """检查磁盘空间是否足够""" + # 获取输入目录大小 + input_size = self._get_directory_size(self.config.image_dir) + + # 获取输出目录所在磁盘的剩余空间 + output_drive = os.path.splitdrive( + os.path.abspath(self.config.output_dir))[0] + if not output_drive: # 处理Linux/Unix路径 + output_drive = '/' + + disk_usage = psutil.disk_usage(output_drive) + free_space = disk_usage.free + + # 计算所需空间（输入大小的1.5倍） + required_space = input_size * 1.5 + + if free_space < required_space: + error_msg = ( + f"磁盘空间不足!\n" + f"输入目录大小: {input_size / (1024**3):.2f} GB\n" + f"所需空间: {required_space / (1024**3):.2f} GB\n" + f"可用空间: {free_space / (1024**3):.2f} GB\n" + f"在驱动器 {output_drive}" + ) + raise RuntimeError(error_msg) + def extract_gps(self) -> pd.DataFrame: """提取GPS数据""" self.logger.info("开始提取GPS数据") @@ -217,9 +260,9 @@ class ImagePreprocessor: for grid_id, points in grid_points.items(): output_dir = os.path.join( - self.config.output_dir, - f"grid_{grid_id[0]}_{grid_id[1]}", - "project", + self.config.output_dir, + f"grid_{grid_id[0]}_{grid_id[1]}", + "project", "images" ) @@ -229,7 +272,8 @@ class ImagePreprocessor: src = os.path.join(self.config.image_dir, point["file"]) dst = os.path.join(output_dir, point["file"]) shutil.copy(src, dst) - self.logger.info(f"网格 ({grid_id[0]},{grid_id[1]}) 包含 {len(points)} 张图像") + self.logger.info( + f"网格 ({grid_id[0]},{grid_id[1]}) 包含 {len(points)} 张图像") def merge_tif(self, grid_points: Dict[tuple, pd.DataFrame], produce_dem: bool): """合并所有网格的影像产品""" @@ -248,13 +292,17 @@ class ImagePreprocessor: self.logger.info("开始合并OBJ模型") merger = MergeObj(self.config.output_dir) merger.merge_grid_obj(grid_points, translations) - + def post_process(self, successful_grid_points: Dict[tuple, pd.DataFrame], grid_points: Dict[tuple, pd.DataFrame], translations: Dict[tuple, tuple]): - if len(successful_grid_points) < len(grid_points): - self.logger.warning( - f"有 {len(grid_points) - len(successful_grid_points)} 个网格处理失败，" - f"将只合并成功处理的 {len(successful_grid_points)} 个网格" - ) + if len(successful_grid_points) == 1: + self.logger.info(f"只有一个网格{successful_grid_points.keys()}，无需合并") + self.logger.info(f"生产结果请在{successful_grid_points.keys()[0]}目录下查看") + return + elif len(successful_grid_points) < len(grid_points): + self.logger.warning( + f"有 {len(grid_points) - len(successful_grid_points)} 个网格处理失败，" + f"将只合并成功处理的 {len(successful_grid_points)} 个网格" + ) self.merge_tif(successful_grid_points, self.config.produce_dem) if self.config.mode != "快拼模式": self.merge_ply(successful_grid_points) @@ -271,10 +319,12 @@ class ImagePreprocessor: self.copy_images(grid_points) self.logger.info("预处理任务完成") - successful_grid_points = self.odm_monitor.process_all_grids(grid_points, self.config.produce_dem) + successful_grid_points = self.odm_monitor.process_all_grids( + grid_points, self.config.produce_dem) + + self.post_process(successful_grid_points, + grid_points, translations) - self.post_process(successful_grid_points, grid_points, translations) - except Exception as e: self.logger.error(f"处理过程中发生错误: {str(e)}", exc_info=True) raise @@ -283,8 +333,8 @@ class ImagePreprocessor: if __name__ == "__main__": # 创建配置 config = PreprocessConfig( - image_dir=r"/home/cug/datasets/error3/project/images", - output_dir=r"/home/cug/ODM_output/error3", + image_dir=r"E:\datasets\UAV\1619\project\images", + output_dir=r"G:\ODM_output\1619", cluster_eps=0.01, cluster_min_samples=5, @@ -300,7 +350,7 @@ if __name__ == "__main__": filter_dense_distance_threshold=10, filter_time_threshold=timedelta(minutes=5), - grid_size=800, + grid_size=1000, grid_overlap=0.05, diff --git a/requirements.txt b/requirements.txt index ff696df..c7f738b 100644 --- a/requirements.txt +++ b/requirements.txt @@ -4,5 +4,5 @@ scikit-learn matplotlib piexif geopy -psutil +psutil>=5.8.0 docker>=6.1.3 diff --git a/utils/odm_monitor.py b/utils/odm_monitor.py index 59d13b2..80c2001 100644 --- a/utils/odm_monitor.py +++ b/utils/odm_monitor.py @@ -11,14 +11,27 @@ class NotOverlapError(Exception): """图像重叠度不足异常""" pass + class DockerNotRunError(Exception): """Docker未启动异常""" pass + class DockerShareError(Exception): """Docker目录共享异常""" pass + +class OutOfMemoryError(Exception): + """内存不足异常""" + pass + + +class StrangeValuesError(Exception): + """异常值异常""" + pass + + class ODMProcessMonitor: """ODM处理监控器""" @@ -29,7 +42,7 @@ class ODMProcessMonitor: def _check_success(self, grid_dir: str) -> bool: """检查ODM是否执行成功 - + 检查项目: 1. 必要的文件夹是否存在 2. 正射影像是否生成且有效 @@ -39,41 +52,42 @@ class ODMProcessMonitor: success_markers = ['odm_orthophoto'] if self.mode != "快拼模式": success_markers.extend(['odm_texturing', 'odm_georeferencing']) - + if not all(os.path.exists(os.path.join(grid_dir, 'project', marker)) for marker in success_markers): self.logger.error("必要的文件夹未生成") return False - + # 检查正射影像文件 - ortho_path = os.path.join(grid_dir, 'project', 'odm_orthophoto', 'odm_orthophoto.original.tif') + ortho_path = os.path.join( + grid_dir, 'project', 'odm_orthophoto', 'odm_orthophoto.original.tif') if not os.path.exists(ortho_path): self.logger.error("正射影像文件未生成") return False - + # 检查文件大小 file_size_mb = os.path.getsize(ortho_path) / (1024 * 1024) # 转换为MB if file_size_mb < 1: self.logger.error(f"正射影像文件过小: {file_size_mb:.2f}MB") return False - + try: # 打开影像文件 ds = gdal.Open(ortho_path) if ds is None: self.logger.error("无法打开正射影像文件") return False - + # 读取第一个波段 band = ds.GetRasterBand(1) - + # 获取统计信息 stats = band.GetStatistics(False, True) if stats is None: self.logger.error("无法获取影像统计信息") return False - + min_val, max_val, mean, std = stats - + # 计算空值比例 no_data_value = band.GetNoDataValue() array = band.ReadAsArray() @@ -81,20 +95,20 @@ class ODMProcessMonitor: no_data_ratio = np.sum(array == no_data_value) / array.size else: no_data_ratio = 0 - + # 检查空值比例是否过高（超过50%） if no_data_ratio > 0.5: self.logger.error(f"正射影像空值比例过高: {no_data_ratio:.2%}") return False - + # 检查影像是否全黑或全白 if max_val - min_val < 1: self.logger.error("正射影像可能无效：像素值范围过小") return False - + ds = None # 关闭数据集 return True - + except Exception as e: self.logger.error(f"检查正射影像时发生错误: {str(e)}") return False @@ -112,7 +126,8 @@ class ODMProcessMonitor: while current_try < max_retries: current_try += 1 - self.logger.info(f"第 {current_try} 次尝试处理网格 ({grid_id[0]},{grid_id[1]})") + self.logger.info( + f"第 {current_try} 次尝试处理网格 ({grid_id[0]},{grid_id[1]})") try: # 构建Docker命令 @@ -146,54 +161,59 @@ class ODMProcessMonitor: docker_command += "--rerun-all" self.logger.info(docker_command) - + result = subprocess.run( docker_command, shell=True, stdout=subprocess.PIPE, stderr=subprocess.PIPE) stdout, stderr = result.stdout.decode( 'utf-8'), result.stderr.decode('utf-8') - self.logger.error(f"==========stderr==========: {stderr}") - - # 检查是否有错误 stdout_lines = stdout.strip().split('\n') - last_lines = stdout_lines[-10:] if len(stdout_lines) > 10 else stdout_lines - - # 检查Docker是否未运行 - if any("docker not run" in line.lower() for line in last_lines) or \ - any("docker daemon" in line.lower() for line in last_lines) or \ - any("cannot connect to the docker daemon" in line.lower() for line in last_lines): - raise DockerNotRunError("Docker服务未启动") + last_lines = '\n'.join( + stdout_lines[-50:] if len(stdout_lines) > 10 else stdout_lines) + self.logger.info(f"==========stdout==========: {last_lines}") - # 检查目录共享问题 - if any("not share" in line.lower() for line in last_lines) or \ - any("permission denied" in line.lower() for line in last_lines) or \ - any("access is denied" in line.lower() for line in last_lines): - raise DockerShareError("Docker无法访问目录") - - # 检查是否有重叠度不足错误 - if any("not overlap" in line.lower() for line in last_lines): - raise NotOverlapError("检测到图像重叠度不足错误") - - # 检查执行结果 - if self._check_success(grid_dir): - self.logger.info(f"网格 ({grid_id[0]},{grid_id[1]}) 处理成功") - return True, "" - - if current_try < max_retries: - self.logger.warning(f"网格处理失败，准备第 {current_try + 1} 次重试") + if stderr: + self.logger.error(f"docker run指令执行失败") + self.logger.error(f"==========stderr==========: {stderr}") + if "error during connect" in stderr or "The system cannot find the file specified" in stderr: + raise DockerNotRunError + elif "user declined directory sharing" in stderr: + raise DockerShareError + else: + raise Exception(f"Docker运行失败，需要人工排查错误") + # TODO 处理时间组删除，删多了的情况 else: - self.logger.error(f"网格 ({grid_id[0]},{grid_id[1]}) 处理失败，已达到最大重试次数") - return False, f"网格 ({grid_id[0]},{grid_id[1]}) 处理失败，已重试{max_retries}次" + self.logger.info("docker run指令执行成功") + if "ODM app finished" in last_lines: + self.logger.info("ODM处理完成") + if self._check_success(grid_dir): + self.logger.info( + f"网格 ({grid_id[0]},{grid_id[1]}) 处理成功") + return True, "" + else: + self.logger.error( + f"虽然ODM处理完成，但是生产产品质量可能不合格，需要人工检查") + raise Exception(f"虽然ODM处理完成，但是生产产品质量可能不合格，需要人工检查") + elif "enough overlap" in last_lines: + raise NotOverlapError + elif "out of memory" in last_lines: + raise OutOfMemoryError + elif "strange values" in last_lines: + raise StrangeValuesError + else: + raise Exception(f"ODM处理失败，需要人工排查错误") except NotOverlapError: if use_lowest_quality: - self.logger.warning("检测到'not overlap'错误，移除lowest quality参数后重试") + self.logger.warning( + "检测到not overlap错误，移除lowest quality参数后重试") use_lowest_quality = False continue else: - self.logger.error("即使移除lowest quality参数后仍然出现'not overlap'错误") - return False, "图像重叠度不足" - + self.logger.error( + "即使移除lowest quality参数后仍然出现not overlap错误") + return False, "图像重叠度不足，请检查数据集的采样间隔情况" + except DockerNotRunError: self.logger.error("Docker服务未启动") return False, "Docker没有启动，请启动Docker" @@ -202,18 +222,26 @@ class ODMProcessMonitor: self.logger.error("Docker无法访问目录") return False, "Docker无法访问数据目录或输出目录，请检查目录权限和共享设置" + except OutOfMemoryError: + self.logger.error("内存不足，请减少输入图像的数量") + return False, "内存不足" + except StrangeValuesError: + # TODO 怎么处理异常值 + self.logger.error("重建过程中出现异常值") + return False, "检测到异常值，请检查输入图像" + return False, f"网格 ({grid_id[0]},{grid_id[1]}) 处理失败" def process_all_grids(self, grid_points: Dict[tuple, pd.DataFrame], produce_dem: bool) -> Dict[tuple, pd.DataFrame]: """处理所有网格 - + Returns: Dict[tuple, pd.DataFrame]: 成功处理的网格点数据字典 """ self.logger.info("开始执行网格处理") successful_grid_points = {} failed_grids = [] - + for grid_id, points in grid_points.items(): grid_dir = os.path.join( self.output_dir, f'grid_{grid_id[0]}_{grid_id[1]}' @@ -230,27 +258,30 @@ class ODMProcessMonitor: if success: successful_grid_points[grid_id] = points else: - self.logger.error(f"网格 ({grid_id[0]},{grid_id[1]}) 处理失败: {error_msg}") + self.logger.error( + f"网格 ({grid_id[0]},{grid_id[1]}) 处理失败: {error_msg}") failed_grids.append((grid_id, error_msg)) - + except Exception as e: error_msg = str(e) - self.logger.error(f"处理网格 ({grid_id[0]},{grid_id[1]}) 时发生异常: {error_msg}") + self.logger.error( + f"处理网格 ({grid_id[0]},{grid_id[1]}) 时发生异常: {error_msg}") failed_grids.append((grid_id, error_msg)) - + # 汇总处理结果 total_grids = len(grid_points) failed_count = len(failed_grids) success_count = len(successful_grid_points) - - self.logger.info(f"网格处理完成。总计: {total_grids}, 成功: {success_count}, 失败: {failed_count}") + + self.logger.info( + f"网格处理完成。总计: {total_grids}, 成功: {success_count}, 失败: {failed_count}") if failed_grids: self.logger.error("失败的网格:") for grid_id, error_msg in failed_grids: - self.logger.error(f"网格 ({grid_id[0]},{grid_id[1]}): {error_msg}") - + self.logger.error( + f"网格 ({grid_id[0]},{grid_id[1]}): {error_msg}") + if len(successful_grid_points) == 0: raise Exception("所有网格处理都失败，无法继续处理") - return successful_grid_points