循环调用虹软python接口会崩掉

[AloveDL]

我目前在做一个实时的人脸对比程序,我用了两个线程，第一个线程实时显示结果，第二个线程做人脸对比，年龄预测等功能，但是第二个线程总是崩掉，有时候也会报Fatal Python error: GC object already tracked这个错误，不知道什么原因
下面是我第二个线程循环调用的函数

 def compare_face(self, ):
 try:
 if self.tmp_frame is not None:
 new_img = self.tmp_frame.copy()
 new_img = load_image(new_img)

 image_ubytes = new_img.imageData.ctypes.data_as(POINTER(c_ubyte))
 detect_faces = asf_struct.ASFMultiFaceInfo()
 single_face = asf_struct.ASFSingleFaceInfo()
 feature = asf_struct.ASF_FaceFeature()
 detect_face_ret = asf_func.detect_face(
 self.image_engine,
 new_img.width,
 new_img.height,
 asf_common.ASVL_PAF_RGB24_B8G8R8,
 image_ubytes,
 byref(detect_faces)
 )
 if detect_face_ret != 0:
 print("检测人脸失败：%s" % (detect_face_ret))
 return 0
 # self.lock.release()

 if detect_faces.faceNum == 0:
 return 0
 else:
 single_face.faceRect.left = detect_faces.faceRect[0].left
 single_face.faceRect.top = detect_faces.faceRect[0].top
 single_face.faceRect.right = detect_faces.faceRect[0].right
 single_face.faceRect.bottom = detect_faces.faceRect[0].bottom
 single_face.faceOrient = detect_faces.faceOrient[0]
 single_face.faceDataInfo = detect_faces.faceDataInfoList[0]
 get_feature_ret = asf_func.get_feature(
 self.image_engine,
 new_img.width,
 new_img.height,
 asf_common.ASVL_PAF_RGB24_B8G8R8,
 image_ubytes,
 byref(single_face),
 asf_common.ASF_RegisterOrNot.ASF_RECOGNITION.value[0],
 0,
 byref(feature)
 )
 self.pos = [single_face.faceRect.left, single_face.faceRect.top, single_face.faceRect.right,
 single_face.faceRect.bottom]
 if get_feature_ret != 0:
 print("人脸特征提取失败：%s" % (get_feature_ret))

 for id, feature_data in enumerate(self.feature_set):
 confidenceLevel = c_float()
 feature1 = asf_struct.ASF_FaceFeature()
 feature1.featureSize = feature.featureSize
 featureptr = np.array(feature_data).ctypes.data_as(POINTER(c_ubyte))
 for i in range(feature.featureSize):
 featureptr[i] = feature_data[i]
 feature1.feature = featureptr
 compare_ret = asf_func.compare(
 self.image_engine,
 byref(feature),
 byref(feature1),
 byref(confidenceLevel),
 asf_common.ASF_CompareModel.ASF_LIFE_PHOTO.value[0]
 )
 feature1 = None
 if compare_ret != 0:
 print("人脸对比失败：%s" % (get_feature_ret))
 if confidenceLevel.value >= 0.8:
 self.result_id = id + 1
 break
 process_ret = asf_func.process(
 self.image_engine,
 new_img.width,
 new_img.height,
 asf_common.ASVL_PAF_RGB24_B8G8R8,
 image_ubytes,
 byref(detect_faces),
 self.process_mask
 )
 if process_ret != 0:
 print("初始化人脸属性失败：%s" % (process_ret))
 live_info = asf_struct.ASF_LivenessInfo()
 live_ret = asf_func.get_livescore(
 self.image_engine,
 byref(live_info)
 )
 if live_ret != 0:
 print("活体检测失败：%s" % (live_ret))
 else:
 self.is_live = live_info.isLive[0]
 gender_info = asf_struct.ASFGenderInfo()
 gender_ret = asf_func.get_gender(
 self.image_engine,
 byref(gender_info)
 )
 if gender_ret != 0:
 print("性别检测失败：%s" % (gender_ret))
 else:
 self.gender = gender_info.genderArray[0]
 # 年龄检测
 age_info = asf_struct.ASFAgeInfo()
 age_ret = asf_func.get_age(
 self.image_engine,
 byref(age_info)
 )
 if age_ret != 0:
 print("年龄检测失败：%s" % (age_ret))
 else:
 self.age = age_info.ageArray[0]
 else:
 cv2.waitKey(30)
 except Exception as e:
 print(e)