import json import os import numpy as np import scipy from scipy.spatial.distance import cdist from sklearn.cluster import AgglomerativeClustering from tool.utils import copy_joint_tracklet,mymovefile,cal_boxes_avg_area,is_interaction_middle_region,is_in_reconnect_region,distance_boxes import copy from tool.utils import load_reconn_filter_region,is_in_crowd_region,is_in_crowd_region42 def check_time(row,col,tracklets): # if (tracklets[col]["start_frame_id"] - tracklets[row]["end_frame_id"]) < 2: # return False # if (tracklets[col]["start_frame_id"] - tracklets[row]["end_frame_id"]) > 500: # return False return True def check_tracklets(tracklets): end_region_ids = [2,4,6,8] tracklets.sort(key=function) remove_tracklet_idx = [] re_track_id_list = [] keep_tracklet_idx = [] track_id_list = [] for idx in range(len(tracklets)): if idx == 0: if tracklets[idx]["end_region_id"] in end_region_ids: keep_tracklet_idx.append(idx) for ind in keep_tracklet_idx: track_id_list.append(tracklets[ind]["track_id"]) for idx_temp in range(len(tracklets)): if idx_temp not in keep_tracklet_idx: re_track_id_list.append(tracklets[idx_temp]["track_id"]) return track_id_list,re_track_id_list idx_ = idx + 1 if idx_ == len(tracklets): for ind in keep_tracklet_idx: track_id_list.append(tracklets[ind]["track_id"]) for idx_temp in range(len(tracklets)): if idx_temp not in keep_tracklet_idx: re_track_id_list.append(tracklets[idx_temp]["track_id"]) return track_id_list,re_track_id_list offset = tracklets[idx]["end_frame_id"] - tracklets[idx_]["start_frame_id"] if offset > -400 and offset < 52: if check_space(idx,idx_,tracklets): if idx not in keep_tracklet_idx: keep_tracklet_idx.append(idx) if idx_ not in keep_tracklet_idx: keep_tracklet_idx.append(idx_) if tracklets[keep_tracklet_idx[-1]]["end_region_id"] in end_region_ids: for ind in keep_tracklet_idx: track_id_list.append(tracklets[ind]["track_id"]) for idx_temp in range(len(tracklets)): if idx_temp not in keep_tracklet_idx: re_track_id_list.append(tracklets[idx_temp]["track_id"]) return track_id_list,re_track_id_list else: if idx in keep_tracklet_idx: remove_tracklet_idx.append(idx_) else: remove_tracklet_idx.append(idx) else: if idx in keep_tracklet_idx: remove_tracklet_idx.append(idx_) else: remove_tracklet_idx.append(idx) print("error*****************") raise def check_space(idx,idx_,tracklets): tracklet_0 = tracklets[idx] boxes_0 = tracklet_0["box_list"] tracklet_1 = tracklets[idx_] boxes_1 = tracklet_1["box_list"] a = distance_boxes(boxes_0[0],boxes_0[-1]) b = distance_boxes(boxes_0[0],boxes_1[0]) c = distance_boxes(boxes_0[-1],boxes_1[0]) # print("a:",a) # print("b:",b) # print("c:",c) if (c > 500): return False if(b - a) > -30: if (boxes_0[0][0] > boxes_0[-1][0] and (boxes_0[-1][0] - boxes_1[0][0]) > -40): return True if (boxes_0[0][0] < boxes_0[-1][0] and (boxes_1[0][0] - boxes_0[-1][0]) > -40): return True return False def gate_cost_matrix(cost_matrix, tracklets): for row, track_0 in enumerate(tracklets): for col, track_1 in enumerate(tracklets): if row == col: cost_matrix[row,col] = 2 continue # if check_time(row,col,tracklets) == False: # cost_matrix[row,col] = 2 # continue # if check_space(row,col,tracklets) == False: # cost_matrix[row,col] = 2 # continue return cost_matrix def embedding_distance(tracklet_0, metric='cosine'): cost_matrix = np.zeros((len(tracklet_0), len(tracklet_0)), dtype=np.float64) track_features_0 = np.asarray([track["avg_feature"] for track in tracklet_0], dtype=np.float64) track_features_1 = np.asarray([track["avg_feature"] for track in tracklet_0], dtype=np.float64) cost_matrix = np.maximum(0.0, cdist(track_features_0, track_features_1, metric)) # Nomalized features by ceiling to 0 return cost_matrix def get_cost_matrix(tracklet_0): cost_matrix = embedding_distance(tracklet_0) cost_matrix = gate_cost_matrix(cost_matrix,tracklet_0) return cost_matrix def get_match(cluster_labels): cluster_dict = dict() cluster = list() result=[] for i, l in enumerate(cluster_labels): if l in list(cluster_dict.keys()): cluster_dict[l].append(i) else: cluster_dict[l] = [i] for idx in cluster_dict: cluster.append(cluster_dict[idx]) print("cluster:",cluster) for connect in cluster: if len(connect) > 1: result.append(connect) return result def check_tracklet(tracklet): is_box_list = tracklet["is_box"] box_list = tracklet["box_list"] is_feature_list = tracklet["is_feature"] feature_list = tracklet["feature_list"] start_frame_id = tracklet["start_frame_id"] end_frame_id = tracklet["end_frame_id"] start_region_id = tracklet["start_region_id"] end_region_id = tracklet["end_region_id"] go_through_region = tracklet["go_through_region"] if end_frame_id - start_frame_id != len(is_box_list) - 1: print("error box number track id:",tracklet["track_id"]) print("box size {},frame size {}".format(len(is_box_list),(end_frame_id - start_frame_id))) raise boxes_num = len([i for i in is_box_list if i == 1]) feature_num = len([i for i in is_feature_list if i == 1]) if feature_num > boxes_num: print("feature box num is error track_id:",tracklet["track_id"]) raise if boxes_num != len(box_list): print("boxes_num num is error track_id",tracklet["track_id"]) raise if feature_num != len(feature_list): print("feature_num is error track_id",tracklet["track_id"]) raise if start_region_id != go_through_region[0]: print("go_through_region start_region_id is error track_id",tracklet["track_id"]) raise # if end_region_id != go_through_region[-1]: # print("go_through_region end_region_id is error track_id",tracklet["track_id"]) # raise def generate_new_tracklet(tracklet_a,tracklet_b,filter_region,cam_name): tracklet_new = {"track_id":0,"is_box":[],"box_list":[],"is_feature":[],"feature_list":[], "start_frame_id":0,"end_frame_id":0,"start_region_id":0,"end_region_id":0,"go_through_region":[]} tracklet_new["track_id"] = tracklet_a["track_id"] tracklet_new["start_frame_id"] = tracklet_a["start_frame_id"] tracklet_new["end_frame_id"] = tracklet_b["end_frame_id"] tracklet_new["start_region_id"] = tracklet_a["start_region_id"] tracklet_new["end_region_id"] = tracklet_b["end_region_id"] tracklet_new["go_through_region"] = tracklet_a["go_through_region"] is_box_list_b = tracklet_b["is_box"] box_list_b = tracklet_b["box_list"] is_feature_list_b = tracklet_b["is_feature"] feature_list_b = tracklet_b["feature_list"] start_frame_id_b = tracklet_b["start_frame_id"] go_through_region_b = tracklet_b["go_through_region"] for region_id in go_through_region_b: if region_id not in tracklet_new["go_through_region"]: tracklet_new["go_through_region"].append(region_id) is_box_list_a = tracklet_a["is_box"] box_list_a = tracklet_a["box_list"] is_feature_list_a = tracklet_a["is_feature"] feature_list_a = tracklet_a["feature_list"] start_frame_id_a = tracklet_a["start_frame_id"] end_frame_id_a = tracklet_a["end_frame_id"] have_box_idx_a = 0 have_feature_idx_a = 0 have_box_idx_b = 0 have_feature_idx_b = 0 offset = tracklet_new["end_frame_id"] - tracklet_new["start_frame_id"] print("offset:",offset) for idx in range(offset+1): frame_id = start_frame_id_a + idx #print("frame_id:{},idx:{}".format(frame_id,idx)) if frame_id >= start_frame_id_b: #print("append b") idx_ = frame_id - start_frame_id_b is_box = is_box_list_b[idx_] if is_box == 1: box = box_list_b[have_box_idx_b] ret = is_in_crowd_region(box,filter_region,cam_name) #print("bbret:{},area :{}".format(ret,(box[2]*box[3]))) if box[2]*box[3] > 500 and ret == False: tracklet_new["is_box"].append(is_box) tracklet_new["box_list"].append(box_list_b[have_box_idx_b]) have_box_idx_b = have_box_idx_b + 1 if is_feature_list_b[idx_] == 1: if is_in_crowd_region42(box,filter_region,cam_name)and cam_name =="c042": tracklet_new["is_feature"].append(0) have_feature_idx_b = have_feature_idx_b + 1 else: tracklet_new["is_feature"].append(is_feature_list_b[idx_]) tracklet_new["feature_list"].append(feature_list_b[have_feature_idx_b]) have_feature_idx_b = have_feature_idx_b + 1 else: tracklet_new["is_feature"].append(is_feature_list_b[idx_]) else: have_box_idx_b = have_box_idx_b + 1 tracklet_new["is_box"].append(0) tracklet_new["is_feature"].append(0) else: tracklet_new["is_box"].append(is_box) if is_feature_list_b[idx_] == 1: print("*********error") raise tracklet_new["is_feature"].append(is_feature_list_b[idx_]) tracklet_new["feature_list"].append(feature_list_b[have_feature_idx_b]) have_feature_idx_b = have_feature_idx_b + 1 else: tracklet_new["is_feature"].append(is_feature_list_b[idx_]) continue if frame_id > end_frame_id_a and frame_id < start_frame_id_b: tracklet_new["is_box"].append(0) tracklet_new["is_feature"].append(0) if frame_id <= end_frame_id_a: #print("append a") p = [i for i in is_feature_list_a if i == 1] is_box = is_box_list_a[idx] if is_box == 1: box = box_list_a[have_box_idx_a] ret = is_in_crowd_region(box,filter_region,cam_name) #print("aaret:{},area :{}".format(ret,(box[2]*box[3]))) if box[2]*box[3] > 500 and ret == False: tracklet_new["box_list"].append(box_list_a[have_box_idx_a]) tracklet_new["is_box"].append(is_box) have_box_idx_a = have_box_idx_a + 1 if is_feature_list_a[idx] == 1: aa = is_in_crowd_region42(box,filter_region,cam_name) #print("have_feature_idx_a:",aa) if aa and cam_name =="c042": tracklet_new["is_feature"].append(0) have_feature_idx_a = have_feature_idx_a + 1 else: tracklet_new["is_feature"].append(is_feature_list_a[idx]) tracklet_new["feature_list"].append(feature_list_a[have_feature_idx_a]) have_feature_idx_a = have_feature_idx_a + 1 else: tracklet_new["is_feature"].append(is_feature_list_a[idx]) else: have_box_idx_a = have_box_idx_a + 1 tracklet_new["is_box"].append(0) tracklet_new["is_feature"].append(0) else: tracklet_new["is_box"].append(is_box) if is_feature_list_a[idx] == 1: print("*********error") raise tracklet_new["is_feature"].append(is_feature_list_a[idx]) tracklet_new["feature_list"].append(feature_list_a[have_feature_idx_a]) have_feature_idx_a = have_feature_idx_a + 1 else: tracklet_new["is_feature"].append(is_feature_list_a[idx]) check_tracklet(tracklet_new) return tracklet_new def function(date): return date['start_frame_id'] def merger_tracklet(track_id_list,tracklets,filter_region,cam_name): #tracklets = copy.deepcopy(tracklets_) candidate = [] for track_id in track_id_list: for tracklet in tracklets: if track_id == tracklet["track_id"]: candidate.append(tracklet) candidate.sort(key=function) if len(candidate) < 2: print("error candidate") raise tracklet_new = {} for indx in range(len(candidate)): current_tracklet = candidate[indx] if indx == 0: tracklet_new = candidate[indx] continue else: tracklet_new = generate_new_tracklet(tracklet_new,current_tracklet,filter_region,cam_name) return tracklet_new def joint_tracklets(tracklets,track_id_list,cam_name,base_path,filter_region):#tracklet_0,cam_name): cost_matrix = get_cost_matrix(tracklets) #print("cost_matrix:",cost_matrix) score = 0.0 # if cam_name == "c042": # score = 0.7 # else: # score = 0.7 cluster_labels = AgglomerativeClustering(n_clusters=None, distance_threshold=0.7, affinity='precomputed', linkage='complete').fit_predict(cost_matrix) print("cluster_labels:",type(cluster_labels)) print("cluster_labels:",cluster_labels) labels = get_match(cluster_labels) print("labels:",labels) for tracket_conn in labels: connection_tracklet = [] for idx in tracket_conn: connection_tracklet.append(tracklets[idx]) print("******") for track_i in tracket_conn: print("track_id:",track_id_list[track_i]) keep_track_id,remove_track_id = check_tracklets(connection_tracklet) print("keep_tracklet_id:",keep_track_id) print("remove_track_id:",remove_track_id) if len(keep_track_id) >= 2: res = merger_tracklet(keep_track_id,tracklets,filter_region,cam_name) new_tracklet_path = os.path.join(base_path,cam_name+"/connect/") new_tracklet_file = os.path.join(new_tracklet_path,str(res["track_id"]) + ".json") jsObj = json.dumps(res) if not os.path.exists(new_tracklet_path): os.makedirs(new_tracklet_path) with open(new_tracklet_file, "w") as f: f.write(jsObj) f.close() move_joint_tracklet(keep_track_id,cam_name,base_path,"/remove/",res["track_id"]) if len(keep_track_id) == 1: pass if len(remove_track_id) >= 2 and len(keep_track_id) >= 1: while True: conn_tracklets = [] for track_id in remove_track_id: for tracklet in tracklets: if track_id == tracklet["track_id"]: conn_tracklets.append(tracklet) keep_track_id,remove_track_id = check_tracklets(conn_tracklets) if (len(keep_track_id) >=2): res = merger_tracklet(keep_track_id,tracklets,filter_region,cam_name) new_tracklet_path = os.path.join(base_path,cam_name+"/connect/") new_tracklet_file = os.path.join(new_tracklet_path,str(res["track_id"]) + ".json") jsObj = json.dumps(res) if not os.path.exists(new_tracklet_path): os.makedirs(new_tracklet_path) with open(new_tracklet_file, "w") as f: f.write(jsObj) f.close() move_joint_tracklet(keep_track_id,cam_name,base_path,"/remove/",res["track_id"]) print("keep_tracklet_id:",keep_track_id) print("remove_track_id:",remove_track_id) if(len(remove_track_id) < 2) or (len(keep_track_id) < 2): break def move_joint_tracklet(remove_tracklets,cam_name,base_path,type_remove,merger_track_id): print("move_joint_tracklet cam {} move {} type {}".format(cam_name,len(remove_tracklets),type_remove)) print("move list:",remove_tracklets) srcpath = os.path.join(base_path,cam_name+"/result/") despath = os.path.join(base_path,cam_name+type_remove) for track_id in remove_tracklets: srcfile = os.path.join(srcpath,str(track_id) + ".json") desfile = os.path.join(despath,str(merger_track_id) +"-"+str(track_id)+ ".json") mymovefile(srcfile,desfile) def cam_joint_tracklets(all_tracklet,new_all_mask,reconn_regions,base_path): filter_region = load_reconn_filter_region("../re_region") for cam_name in all_tracklet: if cam_name == "c041": middle_regions = [10] start_region_ids = [1,3,5,7] end_region_ids = [2,4,6,8] tracklets = [] for tracklet in all_tracklet[cam_name]: feature_array = np.array(tracklet["feature_list"]) avg_feature = np.mean(feature_array,axis=0) tracklet['avg_feature'] = avg_feature.tolist() start_region_id = tracklet["start_region_id"] end_region_id = tracklet["end_region_id"] start_frame_id = tracklet["start_frame_id"] end_frame_id = tracklet["end_frame_id"] if end_region_id in start_region_ids: #在开始区域结束 if end_frame_id != 2000: if cal_boxes_avg_area(tracklet["box_list"]) > 700: tracklets.append(tracklet) continue if end_region_id in middle_regions: #在中间区域结束 if end_frame_id != 2000: if cal_boxes_avg_area(tracklet["box_list"]) > 700: tracklets.append(tracklet) continue if(end_region_id in end_region_ids) and (start_region_id not in start_region_ids):#不在开始区域创建 if start_frame_id != 0 and end_frame_id != 2000: if cal_boxes_avg_area(tracklet["box_list"]) > 700: tracklets.append(tracklet) continue if start_frame_id != 0: #第一个框在中间区域 box = tracklet["box_list"][0] if is_interaction_middle_region(box,cam_name,new_all_mask): if cal_boxes_avg_area(tracklet["box_list"]) > 700: tracklets.append(tracklet) continue if is_in_reconnect_region(tracklet["box_list"][0],reconn_regions,cam_name): #在自定义拼接区域 if cal_boxes_avg_area(tracklet["box_list"]) > 800: if end_region_id == 6 or end_region_id == 4: continue tracklets.append(tracklet) continue track_id_list = [] for tracklet in tracklets: track_id_list.append(tracklet["track_id"]) print("exception tracklet len:",len(track_id_list)) print("exception tracklet:",track_id_list) joint_tracklets(tracklets,track_id_list,cam_name,base_path,filter_region) copy_joint_tracklet(cam_name,base_path,"/connect/","/result/") if cam_name == "c042": middle_regions = [10] start_region_ids = [1,3,5,7] end_region_ids = [2,4,6,8] tracklets = [] for tracklet in all_tracklet[cam_name]: feature_array = np.array(tracklet["feature_list"]) avg_feature = np.mean(feature_array,axis=0) tracklet['avg_feature'] = avg_feature.tolist() start_region_id = tracklet["start_region_id"] end_region_id = tracklet["end_region_id"] start_frame_id = tracklet["start_frame_id"] end_frame_id = tracklet["end_frame_id"] # if end_region_id in start_region_ids: #在开始区域结束 # if end_frame_id != 2000: # if cal_boxes_avg_area(tracklet["box_list"]) > 700: # tracklets.append(tracklet) # continue if end_region_id in middle_regions: #在中间区域结束 if end_frame_id != 2000: if cal_boxes_avg_area(tracklet["box_list"]) > 700: tracklets.append(tracklet) continue if(end_region_id in end_region_ids) and (start_region_id not in start_region_ids):#不在开始区域创建 if start_frame_id != 0 and end_frame_id != 2000: if cal_boxes_avg_area(tracklet["box_list"]) > 700: tracklets.append(tracklet) continue if is_in_reconnect_region(tracklet["box_list"][0],reconn_regions,cam_name): #在自定义拼接区域 if cal_boxes_avg_area(tracklet["box_list"]) > 800: tracklets.append(tracklet) continue track_id_list = [] for tracklet in tracklets: track_id_list.append(tracklet["track_id"]) print("exception tracklet len:",len(track_id_list)) print("exception tracklet:",track_id_list) joint_tracklets(tracklets,track_id_list,cam_name,base_path,filter_region) copy_joint_tracklet(cam_name,base_path,"/connect/","/result/") if cam_name == "c043": middle_regions = [10] start_region_ids = [1,3,5,7] end_region_ids = [2,4,6,8] tracklets = [] for tracklet in all_tracklet[cam_name]: feature_array = np.array(tracklet["feature_list"]) avg_feature = np.mean(feature_array,axis=0) tracklet['avg_feature'] = avg_feature.tolist() start_region_id = tracklet["start_region_id"] end_region_id = tracklet["end_region_id"] start_frame_id = tracklet["start_frame_id"] end_frame_id = tracklet["end_frame_id"] if end_region_id in start_region_ids: #在开始区域结束 if end_frame_id != 2000: if cal_boxes_avg_area(tracklet["box_list"]) > 700: tracklets.append(tracklet) continue if end_region_id in middle_regions: #在中间区域结束 if end_frame_id != 2000: if cal_boxes_avg_area(tracklet["box_list"]) > 700: tracklets.append(tracklet) continue if(end_region_id in end_region_ids) and (start_region_id not in start_region_ids):#不在开始区域创建 if start_frame_id != 0 and end_frame_id != 2000: if cal_boxes_avg_area(tracklet["box_list"]) > 700: tracklets.append(tracklet) continue if start_frame_id != 0: #第一个框在中间区域 box = tracklet["box_list"][0] if is_interaction_middle_region(box,cam_name,new_all_mask): if cal_boxes_avg_area(tracklet["box_list"]) > 700: tracklets.append(tracklet) continue if is_in_reconnect_region(tracklet["box_list"][0],reconn_regions,cam_name): #在自定义拼接区域 if cal_boxes_avg_area(tracklet["box_list"]) > 800: if end_region_id == 6 or end_region_id == 4: continue tracklets.append(tracklet) continue track_id_list = [] for tracklet in tracklets: track_id_list.append(tracklet["track_id"]) print("exception tracklet len:",len(track_id_list)) print("exception tracklet:",track_id_list) joint_tracklets(tracklets,track_id_list,cam_name,base_path,filter_region) copy_joint_tracklet(cam_name,base_path,"/connect/","/result/") if cam_name == "c044": middle_regions = [10] start_region_ids = [1,3,5,7] end_region_ids = [2,4,6,8] tracklets = [] for tracklet in all_tracklet[cam_name]: feature_array = np.array(tracklet["feature_list"]) avg_feature = np.mean(feature_array,axis=0) tracklet['avg_feature'] = avg_feature.tolist() start_region_id = tracklet["start_region_id"] end_region_id = tracklet["end_region_id"] start_frame_id = tracklet["start_frame_id"] end_frame_id = tracklet["end_frame_id"] if end_region_id in start_region_ids: #在开始区域结束 if end_frame_id != 2000: if cal_boxes_avg_area(tracklet["box_list"]) > 700: tracklets.append(tracklet) continue if end_region_id in middle_regions: #在中间区域结束 if end_frame_id != 2000: if cal_boxes_avg_area(tracklet["box_list"]) > 700: tracklets.append(tracklet) continue if(end_region_id in end_region_ids) and (start_region_id not in start_region_ids):#不在开始区域创建 if start_frame_id != 0 and end_frame_id != 2000: if cal_boxes_avg_area(tracklet["box_list"]) > 700: tracklets.append(tracklet) continue if start_frame_id != 0: #第一个框在中间区域 box = tracklet["box_list"][0] if is_interaction_middle_region(box,cam_name,new_all_mask): if cal_boxes_avg_area(tracklet["box_list"]) > 700: tracklets.append(tracklet) continue # if is_in_reconnect_region(tracklet["box_list"][0],reconn_regions,cam_name): #在自定义拼接区域 # if cal_boxes_avg_area(tracklet["box_list"]) > 800: # if end_region_id == 6 or end_region_id == 4: # continue # tracklets.append(tracklet) # continue track_id_list = [] for tracklet in tracklets: track_id_list.append(tracklet["track_id"]) print("exception tracklet len:",len(track_id_list)) print("exception tracklet:",track_id_list) joint_tracklets(tracklets,track_id_list,cam_name,base_path,filter_region) copy_joint_tracklet(cam_name,base_path,"/connect/","/result/") if cam_name == "c045": middle_regions = [10] start_region_ids = [1,3,5,7] end_region_ids = [2,4,6,8] tracklets = [] for tracklet in all_tracklet[cam_name]: feature_array = np.array(tracklet["feature_list"]) avg_feature = np.mean(feature_array,axis=0) tracklet['avg_feature'] = avg_feature.tolist() start_region_id = tracklet["start_region_id"] end_region_id = tracklet["end_region_id"] start_frame_id = tracklet["start_frame_id"] end_frame_id = tracklet["end_frame_id"] if end_region_id in start_region_ids: #在开始区域结束 if end_frame_id != 2000: if cal_boxes_avg_area(tracklet["box_list"]) > 700: tracklets.append(tracklet) continue if end_region_id in middle_regions: #在中间区域结束 if end_frame_id != 2000: if cal_boxes_avg_area(tracklet["box_list"]) > 700: tracklets.append(tracklet) continue if(end_region_id in end_region_ids) and (start_region_id not in start_region_ids):#不在开始区域创建 if start_frame_id != 0 and end_frame_id != 2000: if cal_boxes_avg_area(tracklet["box_list"]) > 700: tracklets.append(tracklet) continue if start_frame_id != 0: #第一个框在中间区域 box = tracklet["box_list"][0] if is_interaction_middle_region(box,cam_name,new_all_mask): if cal_boxes_avg_area(tracklet["box_list"]) > 700: tracklets.append(tracklet) continue # if is_in_reconnect_region(tracklet["box_list"][0],reconn_regions,cam_name): #在自定义拼接区域 # if cal_boxes_avg_area(tracklet["box_list"]) > 800: # if end_region_id == 6 or end_region_id == 4: # continue # tracklets.append(tracklet) # continue track_id_list = [] for tracklet in tracklets: track_id_list.append(tracklet["track_id"]) print("exception tracklet len:",len(track_id_list)) print("exception tracklet:",track_id_list) joint_tracklets(tracklets,track_id_list,cam_name,base_path,filter_region) copy_joint_tracklet(cam_name,base_path,"/connect/","/result/") if cam_name == "c046": middle_regions = [10] start_region_ids = [1,3,5,7] end_region_ids = [2,4,6,8] tracklets = [] for tracklet in all_tracklet[cam_name]: feature_array = np.array(tracklet["feature_list"]) avg_feature = np.mean(feature_array,axis=0) tracklet['avg_feature'] = avg_feature.tolist() start_region_id = tracklet["start_region_id"] end_region_id = tracklet["end_region_id"] start_frame_id = tracklet["start_frame_id"] end_frame_id = tracklet["end_frame_id"] if end_region_id in start_region_ids: #在开始区域结束 if end_frame_id != 2000: if cal_boxes_avg_area(tracklet["box_list"]) > 700: tracklets.append(tracklet) continue if end_region_id in middle_regions: #在中间区域结束 if end_frame_id != 2000: if cal_boxes_avg_area(tracklet["box_list"]) > 700: tracklets.append(tracklet) continue if(end_region_id in end_region_ids) and (start_region_id not in start_region_ids):#不在开始区域创建 if start_frame_id != 0 and end_frame_id != 2000: if cal_boxes_avg_area(tracklet["box_list"]) > 700: tracklets.append(tracklet) continue if start_frame_id != 0: #第一个框在中间区域 box = tracklet["box_list"][0] if is_interaction_middle_region(box,cam_name,new_all_mask): if cal_boxes_avg_area(tracklet["box_list"]) > 700: tracklets.append(tracklet) continue # if is_in_reconnect_region(tracklet["box_list"][0],reconn_regions,cam_name): #在自定义拼接区域 # if cal_boxes_avg_area(tracklet["box_list"]) > 800: # if end_region_id == 6 or end_region_id == 4: # continue # tracklets.append(tracklet) # continue track_id_list = [] for tracklet in tracklets: track_id_list.append(tracklet["track_id"]) print("exception tracklet len:",len(track_id_list)) print("exception tracklet:",track_id_list) joint_tracklets(tracklets,track_id_list,cam_name,base_path,filter_region) copy_joint_tracklet(cam_name,base_path,"/connect/","/result/")